From 118fd3fd282104779c31d0e385cc201c0853ca54 Mon Sep 17 00:00:00 2001
From: Mateusz Ozga <mateusz.ozga@amd.com>
Date: Tue, 12 Nov 2024 15:27:50 +0000
Subject: [PATCH] Apply cshuffle to bwd_weight_cshuffle operator

---
 .../grouped_conv_bwd_weight_xdl_bf16.cpp      |   50 +-
 .../grouped_conv_bwd_weight_xdl_fp16.cpp      |   20 +-
 ..._conv_bwd_weight_xdl_fp16_comp_bf8_fp8.cpp |   49 +-
 ...e_grouped_conv_bwd_weight_xdl_cshuffle.hpp | 1083 +++++++++++++----
 .../transform_conv_bwd_weight_to_gemm_v2.hpp  |  269 ++++
 ...e_grouped_conv_bwd_weight_xdl_instance.hpp |  227 ++--
 .../grouped_convolution_backward_weight.hpp   |  146 ++-
 ...rouped_convolution_backward_weight_xdl.inc |  764 ++++++++++--
 .../grouped_conv1d_bwd_weight/CMakeLists.txt  |   16 +-
 ...bf16_f32_bf16_default_pipev2_instance.cpp} |   18 +-
 ..._bf16_f32_bf16_default_pipev5_instance.cpp |   39 +
 ...nwk_bf16_f32_bf16_pad0_pipev2_instance.cpp |   39 +
 ...nwk_bf16_f32_bf16_pad0_pipev5_instance.cpp |   39 +
 ..._gkxc_gnwk_f16_default_pipev2_instance.cpp |   38 +
 ..._gkxc_gnwk_f16_default_pipev5_instance.cpp |   38 +
 ...wc_gkxc_gnwk_f16_pad0_pipev2_instance.cpp} |   17 +-
 ...nwc_gkxc_gnwk_f16_pad0_pipev5_instance.cpp |   38 +
 ..._gkxc_gnwk_f32_default_pipev2_instance.cpp |   38 +
 ..._gkxc_gnwk_f32_default_pipev5_instance.cpp |   38 +
 ...wc_gkxc_gnwk_f32_pad0_pipev2_instance.cpp} |   17 +-
 ...nwc_gkxc_gnwk_f32_pad0_pipev5_instance.cpp |   38 +
 .../grouped_conv2d_bwd_weight/CMakeLists.txt  |   36 +-
 ...bf16_f32_bf16_default_pipev2_instance.cpp} |   18 +-
 ..._bf16_f32_bf16_default_pipev5_instance.cpp |   40 +
 ...hwk_bf16_f32_bf16_pad0_pipev2_instance.cpp |   40 +
 ...hwk_bf16_f32_bf16_pad0_pipev5_instance.cpp |   40 +
 ...kyxc_gnhwk_f16_default_pipev2_instance.cpp |   38 +
 ...kyxc_gnhwk_f16_default_pipev5_instance.cpp |   38 +
 ..._gkyxc_gnhwk_f16_pad0_pipev2_instance.cpp} |   17 +-
 ...c_gkyxc_gnhwk_f16_pad0_pipev5_instance.cpp |   39 +
 ...kyxc_gnhwk_f32_default_pipev2_instance.cpp |   39 +
 ...kyxc_gnhwk_f32_default_pipev5_instance.cpp |   39 +
 ..._gkyxc_gnhwk_f32_pad0_pipev2_instance.cpp} |   17 +-
 ...c_gkyxc_gnhwk_f32_pad0_pipev5_instance.cpp |   39 +
 ...yxc_nhwgk_bf16_default_pipev2_instance.cpp |   39 +
 ...yxc_nhwgk_bf16_default_pipev5_instance.cpp |   39 +
 ...bf16_f32_bf16_default_pipev2_instance.cpp} |   16 +-
 ..._bf16_f32_bf16_default_pipev5_instance.cpp |   40 +
 ...wgk_bf16_f32_bf16_pad0_pipev2_instance.cpp |   40 +
 ...wgk_bf16_f32_bf16_pad0_pipev5_instance.cpp |   40 +
 ...gkyxc_nhwgk_bf16_pad0_pipev2_instance.cpp} |   16 +-
 ..._gkyxc_nhwgk_bf16_pad0_pipev5_instance.cpp |   40 +
 ...kyxc_nhwgk_f16_default_pipev2_instance.cpp |   40 +
 ...kyxc_nhwgk_f16_default_pipev5_instance.cpp |   40 +
 ..._gkyxc_nhwgk_f16_pad0_pipev2_instance.cpp} |   17 +-
 ...c_gkyxc_nhwgk_f16_pad0_pipev5_instance.cpp |   39 +
 ...kyxc_nhwgk_f32_default_pipev2_instance.cpp |   39 +
 ...kyxc_nhwgk_f32_default_pipev5_instance.cpp |   39 +
 ..._gkyxc_nhwgk_f32_pad0_pipev2_instance.cpp} |   17 +-
 ...c_gkyxc_nhwgk_f32_pad0_pipev5_instance.cpp |   39 +
 .../grouped_conv3d_bwd_weight/CMakeLists.txt  |   40 +-
 ...bf16_f32_bf16_default_pipev2_instance.cpp} |   18 +-
 ..._bf16_f32_bf16_default_pipev5_instance.cpp |   39 +
 ...hwk_bf16_f32_bf16_pad0_pipev2_instance.cpp |   39 +
 ...hwk_bf16_f32_bf16_pad0_pipev5_instance.cpp |   39 +
 ...yxc_gndhwk_f16_default_pipev2_instance.cpp |   38 +
 ...yxc_gndhwk_f16_default_pipev5_instance.cpp |   38 +
 ...kzyxc_gndhwk_f16_pad0_pipev2_instance.cpp} |   17 +-
 ...gkzyxc_gndhwk_f16_pad0_pipev5_instance.cpp |   38 +
 ...yxc_gndhwk_f32_default_pipev2_instance.cpp |   37 +
 ...yxc_gndhwk_f32_default_pipev5_instance.cpp |   37 +
 ...kzyxc_gndhwk_f32_pad0_pipev2_instance.cpp} |   18 +-
 ...gkzyxc_gndhwk_f32_pad0_pipev5_instance.cpp |   37 +
 ...xc_ndhwgk_bf16_default_pipev2_instance.cpp |   39 +
 ...xc_ndhwgk_bf16_default_pipev5_instance.cpp |   39 +
 ...bf16_f32_bf16_default_pipev2_instance.cpp} |   16 +-
 ..._bf16_f32_bf16_default_pipev5_instance.cpp |   40 +
 ...wgk_bf16_f32_bf16_pad0_pipev2_instance.cpp |   40 +
 ...wgk_bf16_f32_bf16_pad0_pipev5_instance.cpp |   40 +
 ...zyxc_ndhwgk_bf16_pad0_pipev2_instance.cpp} |   15 +-
 ...kzyxc_ndhwgk_bf16_pad0_pipev5_instance.cpp |   39 +
 ..._comp_bf8_fp8_default_pipev2_instance.cpp} |   20 +-
 ...6_comp_bf8_fp8_default_pipev5_instance.cpp |   49 +
 ..._f16_comp_bf8_fp8_pad0_pipev2_instance.cpp |   47 +
 ..._f16_comp_bf8_fp8_pad0_pipev5_instance.cpp |   47 +
 ...yxc_ndhwgk_f16_default_pipev2_instance.cpp |   39 +
 ...yxc_ndhwgk_f16_default_pipev5_instance.cpp |   39 +
 ...kzyxc_ndhwgk_f16_pad0_pipev2_instance.cpp} |   17 +-
 ...gkzyxc_ndhwgk_f16_pad0_pipev5_instance.cpp |   39 +
 ...yxc_ndhwgk_f32_default_pipev2_instance.cpp |   39 +
 ...yxc_ndhwgk_f32_default_pipev5_instance.cpp |   39 +
 ...kzyxc_ndhwgk_f32_pad0_pipev2_instance.cpp} |   17 +-
 ...gkzyxc_ndhwgk_f32_pad0_pipev5_instance.cpp |   39 +
 test/grouped_convnd_bwd_weight/CMakeLists.txt |    2 +-
 ...rouped_convnd_bwd_weight_interface_xdl.cpp |  441 +++++--
 85 files changed, 4741 insertions(+), 837 deletions(-)
 rename library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/{device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instance.cpp => device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instance.cpp} (75%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/{device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instance.cpp => device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instance.cpp} (72%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/{device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instance.cpp => device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instance.cpp} (72%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instance.cpp} (76%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instance.cpp} (73%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instance.cpp} (73%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instance.cpp} (80%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instance.cpp} (76%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instance.cpp} (73%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/{device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instance.cpp => device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instance.cpp} (73%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instance.cpp} (75%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instance.cpp} (72%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instance.cpp} (72%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instance.cpp} (80%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instance.cpp} (76%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev2_instance.cpp} (78%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instance.cpp} (73%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instance.cpp
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instance.cpp
 rename library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/{device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instance.cpp => device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instance.cpp} (73%)
 create mode 100644 library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev5_instance.cpp

diff --git a/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_bf16.cpp b/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_bf16.cpp
index 80b9724930..aa2b9f6051 100644
--- a/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_bf16.cpp
+++ b/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_bf16.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "common.hpp"
 
@@ -39,33 +39,33 @@ using DeviceConvBwdWeightInstance =
         WeiElementOp,         // WeiElementwiseOperation
         OutElementOp,         // OutElementwiseOperation
         ConvBwdWeightDefault, // ConvolutionBackwardWeightSpecialization
-        256,                  // BlockSize
-        128,                  // MPerBlock
-        128,                  // NPerBlock
-        4,                    // K0PerBlock
+        64,                   // BlockSize
+        16,                   // MPerBlock
+        16,                   // NPerBlock
+        32,                   // K0PerBlock
         8,                    // K1
-        32,                   // MPerXdl
-        32,                   // NPerXdl
-        2,                    // MXdlPerWave
-        2,                    // NXdlPerWave
-        S<1, 4, 16, 4>,       // ABlockTransferThreadClusterLengths_K0_M_K1
-        S<0, 3, 1, 2>,        // ABlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // ABlockTransferSrcAccessOrder
-        2,                    // ABlockTransferSrcVectorDim
-        8,                    // ABlockTransferSrcScalarPerVector
-        2,                    // ABlockTransferDstScalarPerVector_K1
-        true,                 // ABlockLdsAddExtraM
-        S<1, 4, 16, 4>,       // BBlockTransferThreadClusterLengths_K0_N_K1
-        S<0, 3, 1, 2>,        // BBlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // BBlockTransferSrcAccessOrder
-        2,                    // BBlockTransferSrcVectorDim
-        8,                    // BBlockTransferSrcScalarPerVector
-        2,                    // BBlockTransferDstScalarPerVector_K1
-        true,                 // BBlockLdsAddExtraN
+        16,                   // MPerXdl
+        16,                   // NPerXdl
+        1,                    // MXdlPerWave
+        1,                    // NXdlPerWave
+        S<4, 16, 1>,          // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<2, 0, 1>,           // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,           // ABlockTransferSrcAccessOrder
+        1,                    // ABlockTransferSrcVectorDim
+        1,                    // ABlockTransferSrcScalarPerVector
+        4,                    // ABlockTransferDstScalarPerVector_K1
+        false,                // ABlockLdsAddExtraM
+        S<4, 16, 1>,          // BBlockTransferThreadClusterLengths_K0_N_K1
+        S<2, 0, 1>,           // BBlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,           // BBlockTransferSrcAccessOrder
+        1,                    // BBlockTransferSrcVectorDim
+        1,                    // BBlockTransferSrcScalarPerVector
+        4,                    // BBlockTransferDstScalarPerVector_K1
+        false,                // BBlockLdsAddExtraN
         1,                    // CShuffleMXdlPerWavePerShuffle
         1,                    // CShuffleNXdlPerWavePerShuffle
-        S<1, 32, 1, 4>,       // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
-        128 / (sizeof(WeiDataType) * CHAR_BIT)>; // CBlockTransferScalarPerVector_NWaveNPerXdl
+        S<1, 8, 1, 8>,        // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
+        1>;                   // CBlockTransferScalarPerVector_NWaveNPerXdl
 
 template <ck::index_t NDimSpatial>
 using HostConvBwdWeightInstance = ck::tensor_operation::host::ReferenceConvBwdWeight<NDimSpatial,
diff --git a/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16.cpp b/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16.cpp
index 5c1c97d615..2c01e57e8c 100644
--- a/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16.cpp
+++ b/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "common.hpp"
 
@@ -41,26 +41,26 @@ using DeviceConvBwdWeightInstance =
         256,                  // BlockSize
         128,                  // MPerBlock
         128,                  // NPerBlock
-        4,                    // K0PerBlock
+        32,                   // K0PerBlock
         8,                    // K1
         32,                   // MPerXdl
         32,                   // NPerXdl
         2,                    // MXdlPerWave
         2,                    // NXdlPerWave
-        S<1, 4, 16, 4>,       // ABlockTransferThreadClusterLengths_K0_M_K1
-        S<0, 3, 1, 2>,        // ABlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // ABlockTransferSrcAccessOrder
+        S<4, 16, 1>,          // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<2, 0, 1>,           // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,           // ABlockTransferSrcAccessOrder
         2,                    // ABlockTransferSrcVectorDim
         8,                    // ABlockTransferSrcScalarPerVector
         2,                    // ABlockTransferDstScalarPerVector_K1
-        true,                 // ABlockLdsAddExtraM
-        S<1, 4, 16, 4>,       // BBlockTransferThreadClusterLengths_K0_N_K1
-        S<0, 3, 1, 2>,        // BBlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // BBlockTransferSrcAccessOrder
+        false,                // ABlockLdsAddExtraM
+        S<4, 16, 1>,          // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<2, 0, 1>,           // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,           // ABlockTransferSrcAccessOrder
         2,                    // BBlockTransferSrcVectorDim
         8,                    // BBlockTransferSrcScalarPerVector
         2,                    // BBlockTransferDstScalarPerVector_K1
-        true,                 // BBlockLdsAddExtraN
+        false,                // BBlockLdsAddExtraN
         1,                    // CShuffleMXdlPerWavePerShuffle
         1,                    // CShuffleNXdlPerWavePerShuffle
         S<1, 32, 1, 4>,       // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
diff --git a/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16_comp_bf8_fp8.cpp b/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16_comp_bf8_fp8.cpp
index fca6cdd7ab..68ec627736 100644
--- a/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16_comp_bf8_fp8.cpp
+++ b/example/20_grouped_conv_bwd_weight/grouped_conv_bwd_weight_xdl_fp16_comp_bf8_fp8.cpp
@@ -40,35 +40,38 @@ using DeviceConvBwdWeightInstance =
         WeiElementOp,         // WeiElementwiseOperation
         OutElementOp,         // OutElementwiseOperation
         ConvBwdWeightDefault, // ConvolutionBackwardWeightSpecialization
-        256,                  // BlockSize
-        128,                  // MPerBlock
-        128,                  // NPerBlock
-        4,                    // K0PerBlock
+        64,                   // BlockSize
+        16,                   // MPerBlock
+        16,                   // NPerBlock
+        32,                   // K0PerBlock
         8,                    // K1
-        32,                   // MPerXdl
-        32,                   // NPerXdl
-        2,                    // MXdlPerWave
-        2,                    // NXdlPerWave
-        S<1, 4, 16, 4>,       // ABlockTransferThreadClusterLengths_K0_M_K1
-        S<0, 3, 1, 2>,        // ABlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // ABlockTransferSrcAccessOrder
-        2,                    // ABlockTransferSrcVectorDim
+        16,                   // MPerXdl
+        16,                   // NPerXdl
+        1,                    // MXdlPerWave
+        1,                    // NXdlPerWave
+        S<4, 16, 1>,          // ABlockTransferThreadClusterLengths_K0_M_K1
+        S<2, 0, 1>,           // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,           // ABlockTransferSrcAccessOrder
+        1,                    // ABlockTransferSrcVectorDim
         1,                    // ABlockTransferSrcScalarPerVector
-        1,                    // ABlockTransferDstScalarPerVector_K1
-        true,                 // ABlockLdsAddExtraM
-        S<1, 4, 16, 4>,       // BBlockTransferThreadClusterLengths_K0_N_K1
-        S<0, 3, 1, 2>,        // BBlockTransferThreadClusterArrangeOrder
-        S<0, 2, 1, 3>,        // BBlockTransferSrcAccessOrder
-        2,                    // BBlockTransferSrcVectorDim
+        4,                    // ABlockTransferDstScalarPerVector_K1
+        false,                // ABlockLdsAddExtraM
+        S<4, 16, 1>,          // BBlockTransferThreadClusterLengths_K0_N_K1
+        S<2, 0, 1>,           // BBlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,           // BBlockTransferSrcAccessOrder
+        1,                    // BBlockTransferSrcVectorDim
         1,                    // BBlockTransferSrcScalarPerVector
-        1,                    // BBlockTransferDstScalarPerVector_K1
-        true,                 // BBlockLdsAddExtraN
+        4,                    // BBlockTransferDstScalarPerVector_K1
+        false,                // BBlockLdsAddExtraN
         1,                    // CShuffleMXdlPerWavePerShuffle
         1,                    // CShuffleNXdlPerWavePerShuffle
-        S<1, 32, 1, 4>,       // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
+        S<1, 8, 1, 8>,        // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
         2,                    // CBlockTransferScalarPerVector_NWaveNPerXdl
-        ComputeTypeA,         // ComputeTypeA
-        ComputeTypeB>;        // ComputeTypeB
+        ck::BlockGemmPipelineScheduler::Intrawave, // BlkGemmPipeSched
+        ck::BlockGemmPipelineVersion::v1,          // BlkGemmPipelineVer
+        1,                                         // NumGroupsToMerge
+        ComputeTypeA,                              // ComputeTypeA
+        ComputeTypeB>;                             // ComputeTypeB
 
 template <ck::index_t NDimSpatial>
 using HostConvBwdWeightInstance = ck::tensor_operation::host::ReferenceConvBwdWeight<NDimSpatial,
diff --git a/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp b/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp
index 996107343d..56dc01fcc4 100644
--- a/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp
@@ -8,93 +8,143 @@
 #include <sstream>
 
 #include "ck/utility/common_header.hpp"
+
 #include "ck/tensor_description/tensor_descriptor.hpp"
 #include "ck/tensor_description/tensor_descriptor_helper.hpp"
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/device/device_grouped_conv_bwd_weight.hpp"
 #include "ck/tensor_operation/operator_transform/transform_conv_bwd_weight_to_gemm.hpp"
+#include "ck/tensor_operation/operator_transform/transform_conv_bwd_weight_to_gemm_v2.hpp"
 #include "ck/tensor_operation/gpu/device/convolution_backward_weight_specialization.hpp"
-#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_elementwise_2d.hpp"
+#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
+#include "ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_cshuffle_bwd_weight_v3.hpp"
+#include <ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp>
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_utils.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
+
 #include "ck/host_utility/device_prop.hpp"
 #include "ck/host_utility/kernel_launch.hpp"
+#include "ck/host_utility/flush_cache.hpp"
 
 namespace ck {
 namespace tensor_operation {
 namespace device {
 
 template <typename GridwiseGemm,
-          typename FloatA,
-          typename FloatB,
-          typename FloatC,
-          typename AElementwiseOperation,
-          typename BElementwiseOperation,
-          typename CElementwiseOperation,
-          typename AGridDesc_B_K0_M_K1,
-          typename BGridDesc_B_K0_N_K1,
+          typename AGridDesc_AK0_M_K1,
+          typename BGridDesc_BK0_N_K1,
           typename CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
-          typename Block2CTileMap,
           typename ComputePtrOffsetOfBatch,
-          bool HasMainKBlockLoop>
+          index_t NumGroupsToMerge,
+          bool HasMainKBlockLoop,
+          InMemoryDataOperationEnum CGlobalMemoryDataOperation,
+          index_t MinimumOccupancy = 1,
+          TailNumber TailNum       = TailNumber::Full>
 __global__ void
 #if CK_USE_LAUNCH_BOUNDS
-    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, MinimumOccupancy)
 #endif
-        kernel_batched_gemm_xdlops_bwd_weight(
-            const FloatA* __restrict__ p_a_grid,
-            const FloatB* __restrict__ p_b_grid,
-            FloatC* __restrict__ p_c_grid,
-            const AElementwiseOperation a_element_op,
-            const BElementwiseOperation b_element_op,
-            const CElementwiseOperation c_element_op,
-            const index_t batch_count,
-            const AGridDesc_B_K0_M_K1 a_b_k0_m_k1_grid_desc,
-            const BGridDesc_B_K0_N_K1 b_b_k0_n_k1_grid_desc,
+        kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3(
+            typename GridwiseGemm::Argument karg,
+            const AGridDesc_AK0_M_K1 a_grid_desc_ak0_m_ak1,
+            const BGridDesc_BK0_N_K1 b_grid_desc_bk0_n_bk1,
             const CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
                 c_grid_desc_mblock_mperblock_nblock_nperblock,
-            const Block2CTileMap block_2_ctile_map,
-            const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch)
+            const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch,
+            const index_t num_k_per_block)
 {
 #if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
     defined(__gfx94__))
-    const index_t num_blocks_per_batch =
-        __builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
-    const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
-
-    const long_index_t a_batch_offset = compute_ptr_offset_of_batch.GetAPtrOffset(g_idx);
-    const long_index_t b_batch_offset = compute_ptr_offset_of_batch.GetBPtrOffset(g_idx);
-    const long_index_t c_batch_offset = compute_ptr_offset_of_batch.GetCPtrOffset(g_idx);
-
-    __shared__ FloatA p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte() / sizeof(FloatA)];
-
-    GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
-                                                  p_b_grid + b_batch_offset,
-                                                  p_c_grid + c_batch_offset,
-                                                  p_shared,
-                                                  a_b_k0_m_k1_grid_desc,
-                                                  b_b_k0_n_k1_grid_desc,
-                                                  c_grid_desc_mblock_mperblock_nblock_nperblock,
-                                                  a_element_op,
-                                                  b_element_op,
-                                                  c_element_op,
-                                                  block_2_ctile_map);
+    const index_t g_idx = __builtin_amdgcn_readfirstlane(blockIdx.z * NumGroupsToMerge);
+    const index_t k_idx = __builtin_amdgcn_readfirstlane(blockIdx.y * num_k_per_block);
+
+    const long_index_t a_batch_offset =
+        amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetAPtrOffset(g_idx));
+    const long_index_t b_batch_offset =
+        amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetBPtrOffset(g_idx));
+    const long_index_t e_batch_offset =
+        amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetEPtrOffset(g_idx));
+
+    __shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
+    GridwiseGemm::template Run<AGridDesc_AK0_M_K1,
+                               BGridDesc_BK0_N_K1,
+                               CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
+                               HasMainKBlockLoop,
+                               CGlobalMemoryDataOperation,
+                               TailNum>(karg.p_a_grid + a_batch_offset,
+                                        karg.p_b_grid + b_batch_offset,
+                                        karg.p_c_grid + e_batch_offset,
+                                        p_shared,
+                                        karg,
+                                        a_grid_desc_ak0_m_ak1,
+                                        b_grid_desc_bk0_n_bk1,
+                                        c_grid_desc_mblock_mperblock_nblock_nperblock,
+                                        k_idx);
 #else
-    ignore = p_a_grid;
-    ignore = p_b_grid;
-    ignore = p_c_grid;
-    ignore = a_b_k0_m_k1_grid_desc;
-    ignore = b_b_k0_n_k1_grid_desc;
-    ignore = c_grid_desc_mblock_mperblock_nblock_nperblock;
-    ignore = a_element_op;
-    ignore = b_element_op;
-    ignore = c_element_op;
-    ignore = batch_count;
-    ignore = block_2_ctile_map;
-    ignore = compute_ptr_offset_of_batch;
-
-    compute_ptr_offset_of_batch.GetAPtrOffset(0);
-    compute_ptr_offset_of_batch.GetBPtrOffset(0);
-    compute_ptr_offset_of_batch.GetCPtrOffset(0);
+    ignore = karg;
+#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
+}
+
+template <typename GridwiseGemm,
+          typename AGridDesc_AK0_M_K1,
+          typename BGridDesc_BK0_N_K1,
+          typename CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
+          typename ComputePtrOffsetOfBatch,
+          index_t NumGroupsToMerge,
+          bool HasMainKBlockLoop,
+          InMemoryDataOperationEnum CGlobalMemoryDataOperation,
+          index_t MinimumOccupancy = 1,
+          TailNumber TailNum       = TailNumber::Full>
+__global__ void
+#if CK_USE_LAUNCH_BOUNDS
+    __launch_bounds__(CK_MAX_THREAD_PER_BLOCK, MinimumOccupancy)
+#endif
+        kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3_2lds(
+            typename GridwiseGemm::Argument karg,
+            const AGridDesc_AK0_M_K1 a_grid_desc_ak0_m_ak1,
+            const BGridDesc_BK0_N_K1 b_grid_desc_bk0_n_bk1,
+            const CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
+                c_grid_desc_mblock_mperblock_nblock_nperblock,
+            const ComputePtrOffsetOfBatch compute_ptr_offset_of_batch,
+            const index_t num_k_per_block)
+{
+#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
+    defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__))
+    // offset base pointer for each work-group
+    const index_t g_idx = __builtin_amdgcn_readfirstlane(blockIdx.z * NumGroupsToMerge);
+    const index_t k_idx = __builtin_amdgcn_readfirstlane(blockIdx.y * num_k_per_block);
+
+    const long_index_t a_batch_offset =
+        amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetAPtrOffset(g_idx));
+    const long_index_t b_batch_offset =
+        amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetBPtrOffset(g_idx));
+    const long_index_t e_batch_offset =
+        amd_wave_read_first_lane(compute_ptr_offset_of_batch.GetEPtrOffset(g_idx));
+
+    // Pass two lds pointer is the key to tell compiler that ds_read/write
+    // operate on different lds chunk at same time without order dependecy
+    __shared__ char p_shared_0[GridwiseGemm::GetSharedMemoryNumberOfByte()];
+    __shared__ char p_shared_1[GridwiseGemm::GetSharedMemoryNumberOfByte()];
+
+    GridwiseGemm::template Run_2Lds<AGridDesc_AK0_M_K1,
+                                    BGridDesc_BK0_N_K1,
+                                    CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
+                                    HasMainKBlockLoop,
+                                    CGlobalMemoryDataOperation,
+                                    TailNum>(karg.p_a_grid + a_batch_offset,
+                                             karg.p_b_grid + b_batch_offset,
+                                             karg.p_c_grid + e_batch_offset,
+                                             p_shared_0,
+                                             p_shared_1,
+                                             karg,
+                                             a_grid_desc_ak0_m_ak1,
+                                             b_grid_desc_bk0_n_bk1,
+                                             c_grid_desc_mblock_mperblock_nblock_nperblock,
+                                             k_idx);
+#else
+    ignore = karg;
 #endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
 }
 
@@ -138,8 +188,11 @@ template <ck::index_t NDimSpatial,
           index_t CShuffleNXdlPerWavePerShuffle,
           typename CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
           index_t CBlockTransferScalarPerVector_NWaveNPerXdl,
-          typename ComputeTypeA = InDataType,
-          typename ComputeTypeB = ComputeTypeA>
+          BlockGemmPipelineScheduler BlkGemmPipeSched = BlockGemmPipelineScheduler::Intrawave,
+          BlockGemmPipelineVersion BlkGemmPipelineVer = BlockGemmPipelineVersion::v1,
+          index_t NumGroupsToMerge                    = 1,
+          typename ComputeTypeA                       = InDataType,
+          typename ComputeTypeB                       = ComputeTypeA>
 struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
     : public DeviceGroupedConvBwdWeight<NDimSpatial,
                                         InLayout,
@@ -154,6 +207,10 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
                                         ComputeTypeA,
                                         ComputeTypeB>
 {
+    static_assert(is_same_v<InElementwiseOperation, element_wise::PassThrough>);
+    static_assert(is_same_v<WeiElementwiseOperation, element_wise::PassThrough>);
+    static_assert(is_same_v<OutElementwiseOperation, element_wise::PassThrough>);
+
     using DeviceOp = DeviceGroupedConvBwdWeight_Xdl_CShuffle;
 
     using ADataType = OutDataType;
@@ -174,29 +231,17 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
     static constexpr auto I4 = Number<4>{};
     static constexpr auto I5 = Number<5>{};
 
-    static constexpr auto K1Number = Number<K1>{};
+    static constexpr GemmSpecialization GemmSpec = GemmSpecialization::Default;
+    static constexpr auto K1Number               = Number<K1>{};
 
     static constexpr auto conv_to_gemm_transformer =
-        TransformConvBwdWeightToGemm<NDimSpatial,
-                                     MPerBlock,
-                                     NPerBlock,
-                                     K1Number,
-                                     K0PerBlock,
-                                     ConvBackwardWeightSpecialization>{};
-
-    // Bytes per 32 lds bank: 32 * 4 bytes
-    static constexpr auto BankLength = 128;
-    static constexpr auto ElePerBank = BankLength / sizeof(ADataType);
-
-    // M1 & M0
-    static constexpr auto ABlockLdsM1PerBlock = ElePerBank / K1;
-    static constexpr auto ABlockLdsM0PerBlock = MPerBlock / ABlockLdsM1PerBlock;
-    static constexpr auto ABlockLdsM1Padding  = 4;
-
-    // N1 & N0
-    static constexpr auto BBlockLdsN1PerBlock = ElePerBank / K1;
-    static constexpr auto BBlockLdsN0PerBlock = NPerBlock / BBlockLdsN1PerBlock;
-    static constexpr auto BBlockLdsN1Padding  = 4;
+        TransformConvBwdWeightToGemmV2<NDimSpatial,
+                                       MPerBlock,
+                                       NPerBlock,
+                                       K1Number,
+                                       K0PerBlock / K1Number,
+                                       NumGroupsToMerge,
+                                       ConvBackwardWeightSpecialization>{};
 
     template <ck::index_t NDim, typename ck::enable_if<NDim == 1, bool>::type = false>
     static auto GetABCGridDesc()
@@ -279,66 +324,58 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
     using BGridDesc_K0_N_K1 = remove_cvref_t<decltype(ABCGridDescs{}[I1])>;
     using CGridDesc_M_N     = remove_cvref_t<decltype(ABCGridDescs{}[I2])>;
 
-    using GridwiseGemm = GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_bwd_weight<
-        BlockSize,
-        ADataType,
-        BDataType,
-        AccDataType,
-        CDataType,
-        InMemoryDataOperationEnum::AtomicAdd,
-        AGridDesc_K0_M_K1,
-        BGridDesc_K0_N_K1,
-        CGridDesc_M_N,
-        AElementwiseOperation,
-        BElementwiseOperation,
-        CElementwiseOperation,
-        MPerBlock,
-        NPerBlock,
-        K0PerBlock,
-        MPerXdl,
-        NPerXdl,
-        K1,
-        MXdlPerWave,
-        NXdlPerWave,
-        ABlockTransferThreadClusterLengths_K0_M_K1,
-        ABlockTransferThreadClusterArrangeOrder,
-        ABlockTransferSrcAccessOrder,
-        ABlockTransferSrcVectorDim,
-        ABlockTransferSrcScalarPerVector,
-        ABlockTransferDstScalarPerVector_K1,
-        false, // AThreadTransferSrcResetCoordinateAfterRun,
-        ABlockLdsAddExtraM,
-        ABlockLdsM1PerBlock,
-        ABlockLdsM0PerBlock,
-        ABlockLdsM1Padding,
-        BBlockTransferThreadClusterLengths_K0_N_K1,
-        BBlockTransferThreadClusterArrangeOrder,
-        BBlockTransferSrcAccessOrder,
-        BBlockTransferSrcVectorDim,
-        BBlockTransferSrcScalarPerVector,
-        BBlockTransferDstScalarPerVector_K1,
-        false, // BThreadTransferSrcResetCoordinateAfterRun,
-        BBlockLdsAddExtraN,
-        BBlockLdsN1PerBlock,
-        BBlockLdsN0PerBlock,
-        BBlockLdsN1Padding,
-        CShuffleMXdlPerWavePerShuffle,
-        CShuffleNXdlPerWavePerShuffle,
-        CBlockTransferScalarPerVector_NWaveNPerXdl,
-        CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
-        true,
-        true,
-        1,
-        PipelineVersion::v1,
-        ComputeTypeA,
-        ComputeTypeB>;
+    using GridwiseGemm =
+        GridwiseGemm_xdl_cshuffle_v3<tensor_layout::gemm::RowMajor,
+                                     tensor_layout::gemm::ColumnMajor,
+                                     tensor_layout::gemm::RowMajor,
+                                     ADataType,
+                                     BDataType,
+                                     AccDataType,
+                                     CDataType,
+                                     CDataType,
+                                     AElementwiseOperation,
+                                     BElementwiseOperation,
+                                     CElementwiseOperation,
+                                     GemmSpec,
+                                     BlockSize,
+                                     MPerBlock,
+                                     NPerBlock,
+                                     K0PerBlock,
+                                     K1,
+                                     K1,
+                                     MPerXdl,
+                                     NPerXdl,
+                                     MXdlPerWave,
+                                     NXdlPerWave,
+                                     ABlockTransferThreadClusterLengths_K0_M_K1,
+                                     ABlockTransferThreadClusterArrangeOrder,
+                                     ABlockTransferSrcAccessOrder,
+                                     ABlockTransferSrcVectorDim,
+                                     ABlockTransferSrcScalarPerVector,
+                                     ABlockTransferDstScalarPerVector_K1,
+                                     false,
+                                     ABlockLdsAddExtraM,
+                                     BBlockTransferThreadClusterLengths_K0_N_K1,
+                                     BBlockTransferThreadClusterArrangeOrder,
+                                     BBlockTransferSrcAccessOrder,
+                                     BBlockTransferSrcVectorDim,
+                                     BBlockTransferSrcScalarPerVector,
+                                     BBlockTransferDstScalarPerVector_K1,
+                                     false,
+                                     BBlockLdsAddExtraN,
+                                     CShuffleMXdlPerWavePerShuffle,
+                                     CShuffleNXdlPerWavePerShuffle,
+                                     CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
+                                     CBlockTransferScalarPerVector_NWaveNPerXdl,
+                                     BlkGemmPipeSched,
+                                     BlkGemmPipelineVer,
+                                     ComputeTypeA,
+                                     ComputeTypeB>;
 
     // Argument
     using CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
-        decltype(GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(CGridDesc_M_N{}));
-
-    using Block2CTileMap =
-        decltype(GridwiseGemm::MakeCBlockClusterAdaptor(CGridDesc_M_N{}, 1, 1, 1));
+        decltype(GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
+            CGridDesc_M_N{}, 1, 1));
 
     struct Argument : public BaseArgument
     {
@@ -368,7 +405,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
               b_grid_desc_kbatch_k0_n_k1_{},
               c_grid_desc_m_n_{},
               c_grid_desc_mblock_mperblock_nblock_nperblock_{},
-              block_2_ctile_map_{},
               compute_ptr_offset_of_batch_{},
               M01_{M01},
               N01_{N01},
@@ -420,9 +456,6 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
             b_grid_desc_kbatch_k0_n_k1_ = descs[I1];
             c_grid_desc_m_n_            = descs[I2];
 
-            block_2_ctile_map_ =
-                GridwiseGemm::MakeCBlockClusterAdaptor(c_grid_desc_m_n_, M01, N01, k_batch_);
-
             // A/B/C Batch Stride
             compute_ptr_offset_of_batch_.BatchStrideA_ = a_g_n_k_wos_strides[0];
             compute_ptr_offset_of_batch_.BatchStrideB_ = b_g_n_c_wis_strides[0];
@@ -432,15 +465,14 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
                                 end(filter_spatial_lengths_),
                                 index_t{1},
                                 std::multiplies<>{});
-
-            if(GridwiseGemm::CheckValidity(a_grid_desc_kbatch_k0_m_k1_,
-                                           b_grid_desc_kbatch_k0_n_k1_,
-                                           c_grid_desc_m_n_,
-                                           block_2_ctile_map_))
-            {
-                c_grid_desc_mblock_mperblock_nblock_nperblock_ =
-                    GridwiseGemm::MakeCGridDesc_MBlock_MPerBlock_NBlock_NPerBlock(c_grid_desc_m_n_);
-            }
+            const index_t GemmM = a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
+            const index_t GemmN = b_grid_desc_kbatch_k0_n_k1_.GetLength(I1);
+
+            c_grid_desc_mblock_mperblock_nblock_nperblock_ =
+                GridwiseGemm::MakeCGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
+                    c_grid_desc_m_n_,
+                    GridwiseGemm::CalculateMBlock(GemmM),
+                    GridwiseGemm::CalculateNBlock(GemmN));
         }
 
         const ADataType* p_a_grid_;
@@ -451,10 +483,8 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
         CGridDesc_M_N c_grid_desc_m_n_;
         CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock c_grid_desc_mblock_mperblock_nblock_nperblock_;
 
-        Block2CTileMap block_2_ctile_map_;
-
         // for computing batch offset
-        ComputePtrOffsetOfStridedBatch<> compute_ptr_offset_of_batch_;
+        ComputePtrOffsetOfStridedBatch<I1, I1, I0> compute_ptr_offset_of_batch_;
 
         index_t M01_;
         index_t N01_;
@@ -502,67 +532,628 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
 
         float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
         {
-            if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_kbatch_k0_m_k1_,
-                                            arg.b_grid_desc_kbatch_k0_n_k1_,
-                                            arg.c_grid_desc_m_n_,
-                                            arg.block_2_ctile_map_))
-            {
-                throw std::runtime_error(
-                    "wrong! GridwiseGemm_km_kn_m0m1n0n1_xdlops_v3r1 has invalid setting");
-            }
+            const index_t GemmM = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
+            const index_t GemmN = arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I1);
+            const index_t GemmK = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I0) *
+                                  arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I2);
+
+            const ADataType* p_a_grid = arg.p_a_grid_;
+            const BDataType* p_b_grid = arg.p_b_grid_;
+            typename GridwiseGemm::Argument gemm_arg{
+                p_a_grid, p_b_grid, arg.p_c_grid_, GemmM, GemmN, GemmK, I0, I0, I0, arg.k_batch_};
 
-            const index_t grid_size =
-                arg.block_2_ctile_map_.CalculateGridSize(arg.c_grid_desc_m_n_) * arg.Conv_G_;
-
-            const auto K0 = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
-
-            const bool has_main_k0_block_loop = GridwiseGemm::CalculateHasMainK0BlockLoop(K0);
-
-            auto launch_kernel = [&](auto has_main_k_block_loop) {
-                constexpr bool has_main_loop = has_main_k_block_loop.value;
-
-                const auto kernel = kernel_batched_gemm_xdlops_bwd_weight<
-                    GridwiseGemm,
-                    ADataType,
-                    BDataType,
-                    CDataType,
-                    OutElementwiseOperation,
-                    InElementwiseOperation,
-                    WeiElementwiseOperation,
-                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
-                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
-                    remove_reference_t<DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
-                    remove_reference_t<DeviceOp::Block2CTileMap>,
-                    ComputePtrOffsetOfStridedBatch<>,
-                    has_main_loop>;
-
-                return launch_and_time_kernel(stream_config,
-                                              kernel,
-                                              dim3(grid_size),
-                                              dim3(BlockSize),
-                                              0,
-                                              arg.p_a_grid_,
-                                              arg.p_b_grid_,
-                                              arg.p_c_grid_,
-                                              arg.a_element_op_,
-                                              arg.b_element_op_,
-                                              arg.c_element_op_,
-                                              arg.Conv_G_,
-                                              arg.a_grid_desc_kbatch_k0_m_k1_,
-                                              arg.b_grid_desc_kbatch_k0_n_k1_,
-                                              arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
-                                              arg.block_2_ctile_map_,
-                                              arg.compute_ptr_offset_of_batch_);
+            index_t gdx, gdy, gdz;
+            std::tie(gdx, gdy, gdz) = GridwiseGemm::CalculateGridSize(
+                gemm_arg.M, gemm_arg.N, gemm_arg.KBatch, arg.Conv_G_ / NumGroupsToMerge);
+
+            float ave_time = 0;
+
+            index_t k_grain                  = gemm_arg.KBatch * K0PerBlock;
+            index_t K_split                  = (gemm_arg.K + k_grain - 1) / k_grain * K0PerBlock;
+            const bool has_main_k_block_loop = GridwiseGemm::CalculateHasMainKBlockLoop(K_split);
+
+            const auto num_k_per_block =
+                arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(Number<0>{}) / gemm_arg.KBatch;
+
+            const auto Run = [&](const auto& kernel) {
+                if(stream_config.flush_cache)
+                {
+                    typename GridwiseGemm::Argument gemm_arg_ = gemm_arg;
+                    ck::utility::RotatingMemWrapper<typename GridwiseGemm::Argument> rotating_mem(
+                        gemm_arg_,
+                        stream_config.rotating_count,
+                        gemm_arg_.M * gemm_arg_.K * sizeof(ADataType),
+                        gemm_arg_.K * gemm_arg_.N * sizeof(BDataType));
+                    rotating_mem.Print();
+
+                    auto run_flush_cache = [&]() {
+                        // flush icache
+                        ck::utility::flush_icache();
+                        // rotating mem
+                        rotating_mem.Next();
+                    };
+                    ave_time += ck::utility::launch_and_time_kernel_with_preprocess<false>(
+                        stream_config,
+                        run_flush_cache,
+                        kernel,
+                        dim3(gdx, gdy, gdz),
+                        dim3(BlockSize),
+                        0,
+                        gemm_arg_,
+                        arg.a_grid_desc_kbatch_k0_m_k1_,
+                        arg.b_grid_desc_kbatch_k0_n_k1_,
+                        arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
+                        arg.compute_ptr_offset_of_batch_,
+                        num_k_per_block);
+                }
+                else
+                {
+                    ave_time +=
+                        launch_and_time_kernel(stream_config,
+                                               kernel,
+                                               dim3(gdx, gdy, gdz),
+                                               dim3(BlockSize),
+                                               0,
+                                               gemm_arg,
+                                               arg.a_grid_desc_kbatch_k0_m_k1_,
+                                               arg.b_grid_desc_kbatch_k0_n_k1_,
+                                               arg.c_grid_desc_mblock_mperblock_nblock_nperblock_,
+                                               arg.compute_ptr_offset_of_batch_,
+                                               num_k_per_block);
+                }
             };
 
-            if(has_main_k0_block_loop)
+            constexpr index_t minimum_occupancy =
+                BlkGemmPipeSched == BlockGemmPipelineScheduler::Intrawave ? 1 : 2;
+
+            if(has_main_k_block_loop)
             {
-                return launch_kernel(integral_constant<bool, true>{});
+                // Tail number always full
+                if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1 ||
+                             BlkGemmPipelineVer == BlockGemmPipelineVersion::v3)
+                {
+                    if(gemm_arg.KBatch > 1)
+                    {
+                        const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                            GridwiseGemm,
+                            remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                            remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                            remove_reference_t<
+                                DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                            ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                            NumGroupsToMerge,
+                            true,
+                            InMemoryDataOperationEnum::AtomicAdd,
+                            minimum_occupancy>;
+                        Run(kernel);
+                    }
+                    else
+                    {
+                        const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                            GridwiseGemm,
+                            remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                            remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                            remove_reference_t<
+                                DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                            ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                            NumGroupsToMerge,
+                            true,
+                            InMemoryDataOperationEnum::Set,
+                            minimum_occupancy>;
+                        Run(kernel);
+                    }
+                }
+                // Tail number could be One to Seven
+                else if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v2)
+                {
+                    if(gemm_arg.KBatch > 1)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::One)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::AtomicAdd,
+                                minimum_occupancy,
+                                TailNumber::One>;
+                            Run(kernel);
+                        }
+                        else if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                                TailNumber::Full)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::AtomicAdd,
+                                minimum_occupancy,
+                                TailNumber::Full>;
+                            Run(kernel);
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 2)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Two)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::AtomicAdd,
+                                    minimum_occupancy,
+                                    TailNumber::Two>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 3)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Three)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::AtomicAdd,
+                                    minimum_occupancy,
+                                    TailNumber::Three>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 4)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Four)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::AtomicAdd,
+                                    minimum_occupancy,
+                                    TailNumber::Four>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 5)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Five)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::AtomicAdd,
+                                    minimum_occupancy,
+                                    TailNumber::Five>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 6)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Six)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::AtomicAdd,
+                                    minimum_occupancy,
+                                    TailNumber::Six>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 7)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Seven)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::AtomicAdd,
+                                    minimum_occupancy,
+                                    TailNumber::Seven>;
+                                Run(kernel);
+                            }
+                        }
+                    }
+                    else
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::One)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::Set,
+                                minimum_occupancy,
+                                TailNumber::One>;
+                            Run(kernel);
+                        }
+                        else if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                                TailNumber::Full)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::Set,
+                                minimum_occupancy,
+                                TailNumber::Full>;
+                            Run(kernel);
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 2)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Two)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::Set,
+                                    minimum_occupancy,
+                                    TailNumber::Two>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 3)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Three)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::Set,
+                                    minimum_occupancy,
+                                    TailNumber::Three>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 4)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Four)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::Set,
+                                    minimum_occupancy,
+                                    TailNumber::Four>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 5)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Five)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::Set,
+                                    minimum_occupancy,
+                                    TailNumber::Five>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 6)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Six)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::Set,
+                                    minimum_occupancy,
+                                    TailNumber::Six>;
+                                Run(kernel);
+                            }
+                        }
+
+                        if constexpr(GridwiseGemm::BlockwiseGemmPipe::PrefetchStages > 7)
+                        {
+                            if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) ==
+                               TailNumber::Seven)
+                            {
+                                const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                    GridwiseGemm,
+                                    remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                    remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                    remove_reference_t<
+                                        DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                    ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                    NumGroupsToMerge,
+                                    true,
+                                    InMemoryDataOperationEnum::Set,
+                                    minimum_occupancy,
+                                    TailNumber::Seven>;
+                                Run(kernel);
+                            }
+                        }
+                    }
+                }
+
+                // Tail number could be Odd or Even
+                else if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v4)
+                {
+                    if(gemm_arg.KBatch > 1)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3_2lds<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::AtomicAdd,
+                                minimum_occupancy,
+                                TailNumber::Odd>;
+                            Run(kernel);
+                        }
+                        else
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3_2lds<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::AtomicAdd,
+                                minimum_occupancy,
+                                TailNumber::Even>;
+                            Run(kernel);
+                        }
+                    }
+                    else
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3_2lds<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::Set,
+                                minimum_occupancy,
+                                TailNumber::Odd>;
+                            Run(kernel);
+                        }
+                        else
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3_2lds<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::Set,
+                                minimum_occupancy,
+                                TailNumber::Even>;
+                            Run(kernel);
+                        }
+                    }
+                }
+                else
+                {
+                    if(gemm_arg.KBatch > 1)
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::AtomicAdd,
+                                minimum_occupancy,
+                                TailNumber::Odd>;
+                            Run(kernel);
+                        }
+                        else
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::AtomicAdd,
+                                minimum_occupancy,
+                                TailNumber::Even>;
+                            Run(kernel);
+                        }
+                    }
+                    else
+                    {
+                        if(GridwiseGemm::CalculateKBlockLoopTailNum(K_split) == TailNumber::Odd)
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::Set,
+                                minimum_occupancy,
+                                TailNumber::Odd>;
+                            Run(kernel);
+                        }
+                        else
+                        {
+                            const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                                GridwiseGemm,
+                                remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                                remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                                remove_reference_t<
+                                    DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                                ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                                NumGroupsToMerge,
+                                true,
+                                InMemoryDataOperationEnum::Set,
+                                minimum_occupancy,
+                                TailNumber::Even>;
+                            Run(kernel);
+                        }
+                    }
+                }
             }
             else
             {
-                return launch_kernel(integral_constant<bool, false>{});
+                // Tail number always 1
+                if constexpr(BlkGemmPipelineVer == BlockGemmPipelineVersion::v1)
+                {
+                    if(gemm_arg.KBatch > 1)
+                    {
+                        const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                            GridwiseGemm,
+                            remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                            remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                            remove_reference_t<
+                                DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                            ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                            NumGroupsToMerge,
+                            false,
+                            InMemoryDataOperationEnum::AtomicAdd,
+                            minimum_occupancy>;
+                        Run(kernel);
+                    }
+                    else
+                    {
+                        const auto kernel = kernel_grouped_conv_bwd_weight_xdl_cshuffle_v3<
+                            GridwiseGemm,
+                            remove_reference_t<DeviceOp::AGridDesc_K0_M_K1>,
+                            remove_reference_t<DeviceOp::BGridDesc_K0_N_K1>,
+                            remove_reference_t<
+                                DeviceOp::CGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>,
+                            ComputePtrOffsetOfStridedBatch<I1, I1, I0>,
+                            NumGroupsToMerge,
+                            false,
+                            InMemoryDataOperationEnum::Set,
+                            minimum_occupancy>;
+                        Run(kernel);
+                    }
+                }
             }
+            return ave_time;
         }
 
         float Run(const BaseArgument* p_arg,
@@ -580,10 +1171,28 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
 
     static bool IsSupportedArgument(const Argument& arg)
     {
+        const index_t GemmM = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I1);
+        const index_t GemmN = arg.b_grid_desc_kbatch_k0_n_k1_.GetLength(I1);
+        const index_t GemmK = arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I0) *
+                              arg.a_grid_desc_kbatch_k0_m_k1_.GetLength(I2);
+
+        typename GridwiseGemm::Argument gemm_arg{
+            nullptr, nullptr, nullptr, GemmM, GemmN, GemmK, I0, I0, I0, arg.k_batch_};
+
+        const auto num_k_loop = gemm_arg.AK0 / (K0PerBlock / K1);
+        if constexpr(BlkGemmPipelineVer != BlockGemmPipelineVersion::v1)
+        {
+            if(num_k_loop <= GridwiseGemm::BlockwiseGemmPipe::PrefetchStages)
+            {
+                return false;
+            }
+        }
+
         if(!ck::is_xdl_supported())
         {
             return false;
         }
+
         if constexpr(NDimSpatial == 1)
         {
             if constexpr(!is_GNWC_GKXC_GNWK<InLayout, WeiLayout, OutLayout>())
@@ -626,10 +1235,38 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
             }
         }
 
+        if constexpr(NumGroupsToMerge > 1)
+        {
+            // support only if whole M and N can be proccessed on one block
+            if(!(GemmM <= MPerBlock && GemmN <= NPerBlock))
+            {
+                return false;
+            }
+            if(!(arg.Conv_C_ == 1 && arg.Conv_K_ == 1))
+            {
+                return false;
+            }
+            if(arg.Conv_G_ % NumGroupsToMerge != 0)
+            {
+                return false;
+            }
+        }
+
+        if(!(arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0 &&
+             arg.Conv_K_ % ABlockTransferSrcScalarPerVector == 0))
+        {
+            if(!(arg.Conv_K_ == 1 && arg.compute_ptr_offset_of_batch_.BatchStrideA_ == 1))
+            {
+                return false;
+            }
+            if(!(arg.Conv_C_ == 1 && arg.compute_ptr_offset_of_batch_.BatchStrideB_ == 1))
+            {
+                return false;
+            }
+        }
+
         // vector load A/B matrix from global memory
-        if(!(ABlockTransferSrcVectorDim == 2 && BBlockTransferSrcVectorDim == 2 &&
-             arg.Conv_K_ % ABlockTransferSrcScalarPerVector == 0 &&
-             arg.Conv_C_ % BBlockTransferSrcScalarPerVector == 0))
+        if(!(ABlockTransferSrcVectorDim == 1 && BBlockTransferSrcVectorDim == 1))
         {
             return false;
         }
@@ -641,10 +1278,7 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
         }
 
         // Gridwise GEMM size
-        return GridwiseGemm::CheckValidity(arg.a_grid_desc_kbatch_k0_m_k1_,
-                                           arg.b_grid_desc_kbatch_k0_n_k1_,
-                                           arg.c_grid_desc_m_n_,
-                                           arg.block_2_ctile_map_);
+        return true;
     }
 
     bool IsSupportedArgument(const BaseArgument* p_arg) override
@@ -766,6 +1400,21 @@ struct DeviceGroupedConvBwdWeight_Xdl_CShuffle
 
         return str.str();
     }
+
+    void SetWorkSpacePointer(BaseArgument* p_arg,
+                             void* p_workspace,
+                             const StreamConfig& = StreamConfig{}) const override
+    {
+        auto p_arg_ = dynamic_cast<Argument*>(p_arg);
+        if(p_arg_)
+        {
+            p_arg_->p_workspace_ = p_workspace;
+        }
+        else
+            throw std::runtime_error(
+                "The argument pointer is not an object of "
+                "DeviceGroupedConvBwdWeightTwoStage_Xdl_CShuffle::Argument structure!");
+    }
 };
 
 } // namespace device
diff --git a/include/ck/tensor_operation/operator_transform/transform_conv_bwd_weight_to_gemm_v2.hpp b/include/ck/tensor_operation/operator_transform/transform_conv_bwd_weight_to_gemm_v2.hpp
index bc290d5641..19341906bc 100644
--- a/include/ck/tensor_operation/operator_transform/transform_conv_bwd_weight_to_gemm_v2.hpp
+++ b/include/ck/tensor_operation/operator_transform/transform_conv_bwd_weight_to_gemm_v2.hpp
@@ -34,6 +34,94 @@ struct TransformConvBwdWeightToGemmV2
     static constexpr auto I0 = Number<0>{};
     static constexpr auto I1 = Number<1>{};
 
+    template <index_t NDim, typename enable_if<NDim == 1, bool>::type = false>
+    constexpr static auto
+    make_out_grid_desc(const index_t N,
+                       const index_t Wo,
+                       const index_t K,
+                       const std::array<index_t, NDimSpatial + 3>& output_strides)
+    {
+        const index_t BatchStride = output_strides[0];
+        const index_t WoStride    = output_strides[3];
+        const auto KStride        = Number<1>{};
+        return make_naive_tensor_descriptor(make_tuple(N * Wo, NumGroupsToMerge, K),
+                                            make_tuple(WoStride, BatchStride, KStride));
+    }
+
+    template <index_t NDim, typename enable_if<NDim == 1, bool>::type = false>
+    constexpr static auto
+    make_in_grid_desc(const index_t N,
+                      const index_t Wi,
+                      const index_t C,
+                      const std::array<index_t, NDimSpatial + 3>& input_strides)
+    {
+        const index_t BatchStride = input_strides[0];
+        const index_t NStride     = input_strides[1];
+        const index_t WiStride    = input_strides[3];
+        const auto CStride        = input_strides[2];
+        if constexpr(ConvBackwardWeightSpecialization ==
+                     device::ConvolutionBackwardWeightSpecialization::Filter1x1Stride1Pad0)
+        {
+            return make_naive_tensor_descriptor(make_tuple(N * Wi, NumGroupsToMerge, C),
+                                                make_tuple(WiStride, BatchStride, CStride));
+        }
+        else
+        {
+            return make_naive_tensor_descriptor(
+                make_tuple(N, Wi, NumGroupsToMerge, C),
+                make_tuple(NStride, WiStride, BatchStride, CStride));
+        }
+    }
+
+    template <index_t NDim, typename enable_if<NDim == 1, bool>::type = false>
+    constexpr static auto
+    make_wei_grid_desc(const index_t K,
+                       const index_t X,
+                       const index_t C,
+                       const std::array<index_t, NDimSpatial + 3>& weights_strides)
+    {
+        const auto CStride     = Number<1>{};
+        const auto KStride     = weights_strides[1];
+        const auto XStride     = weights_strides[3];
+        const auto BatchStride = weights_strides[0];
+        // Add NumGroupsToMerge for Batch+M dimension and, 1 as a placehorder
+        // for Batch+N dimension
+        const auto desc = make_naive_tensor_descriptor(
+            make_tuple(NumGroupsToMerge, K, X, 1, C),
+            make_tuple(BatchStride, KStride, XStride, BatchStride, CStride));
+        // Padd 1 to NumGroupsToMerge
+        const auto padded_desc = transform_tensor_descriptor(
+            desc,
+            make_tuple(make_pass_through_transform(NumGroupsToMerge),
+                       make_pass_through_transform(K),
+                       make_pass_through_transform(X),
+                       make_pad_transform(1, 0, NumGroupsToMerge - 1),
+                       make_pass_through_transform(C)),
+            make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}, Sequence<4>{}),
+            make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}, Sequence<4>{}));
+        // We need only matrices from diagonal. Xor returns 0 for the same
+        // values. So if matrices is not on diagonal then it will be stored in padding.
+        // To avoid use of modulo after xor we assume that NumBatch to merge is power of 2.
+        static_assert(NumGroupsToMerge == 1 || NumGroupsToMerge == 2 || NumGroupsToMerge == 4 ||
+                      NumGroupsToMerge == 8 || NumGroupsToMerge == 16 || NumGroupsToMerge == 32 ||
+                      NumGroupsToMerge == 64);
+        const auto unmerged_padded_desc = transform_tensor_descriptor(
+            padded_desc,
+            make_tuple(make_xor_transform(make_tuple(NumGroupsToMerge, NumGroupsToMerge)),
+                       make_pass_through_transform(K),
+                       make_pass_through_transform(X),
+                       make_pass_through_transform(C)),
+            make_tuple(Sequence<0, 3>{}, Sequence<1>{}, Sequence<2>{}, Sequence<4>{}),
+            make_tuple(Sequence<0, 3>{}, Sequence<1>{}, Sequence<2>{}, Sequence<4>{}));
+        // Merge To M, N
+        return transform_tensor_descriptor(
+            unmerged_padded_desc,
+            make_tuple(make_merge_transform(make_tuple(NumGroupsToMerge, K)),
+                       make_merge_transform(make_tuple(X, NumGroupsToMerge, C))),
+            make_tuple(Sequence<0, 1>{}, Sequence<2, 3, 4>{}),
+            make_tuple(Sequence<0>{}, Sequence<1>{}));
+    }
+
     template <index_t NDim, typename enable_if<NDim == 2, bool>::type = false>
     constexpr static auto
     make_out_grid_desc(const index_t N,
@@ -221,6 +309,187 @@ struct TransformConvBwdWeightToGemmV2
             make_tuple(Sequence<0>{}, Sequence<1>{}));
     }
 
+    template <index_t NDim, typename enable_if<NDim == 1, bool>::type = false>
+    static auto MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
+        const index_t N,
+        const index_t K,
+        const index_t C,
+        const std::array<index_t, NDimSpatial>& input_spatial_lengths,
+        const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
+        const std::array<index_t, NDimSpatial>& output_spatial_lengths,
+        const std::array<index_t, NDimSpatial + 3>& input_strides,
+        const std::array<index_t, NDimSpatial + 3>& weights_strides,
+        const std::array<index_t, NDimSpatial + 3>& output_strides,
+        const std::array<index_t, NDimSpatial>& conv_filter_strides,
+        const std::array<index_t, NDimSpatial>& conv_filter_dilations,
+        const std::array<index_t, NDimSpatial>& input_left_pads,
+        const std::array<index_t, NDimSpatial>& input_right_pads,
+        const index_t batch_k)
+    {
+        using namespace ck;
+
+        const index_t Wi = input_spatial_lengths[0];
+
+        const index_t Wo = output_spatial_lengths[0];
+
+        const index_t X = filter_spatial_lengths[0];
+
+        const index_t ConvStrideW = conv_filter_strides[0];
+
+        const index_t ConvDilationW = conv_filter_dilations[0];
+
+        const index_t InLeftPadW = input_left_pads[0];
+
+        const index_t InRightPadW = input_right_pads[0];
+
+        const index_t GemmKTotal = N * Wo;
+        const index_t GemmM      = K * NumGroupsToMerge;
+        const index_t GemmN      = C * X * NumGroupsToMerge;
+
+        const auto PadGemmM = MPerBlock - GemmM % MPerBlock;
+        const auto PadGemmN = NPerBlock - GemmN % NPerBlock;
+
+        const index_t GemmKBatch = batch_k;
+        const index_t GemmK0 =
+            math::integer_divide_ceil(GemmKTotal, GemmK1Number * K0PerBlock * GemmKBatch) *
+            K0PerBlock;
+        const index_t GemmKPad = GemmKBatch * GemmK0 * GemmK1Number;
+
+        const auto out_grid_desc = make_out_grid_desc<NDim>(N, Wo, K, output_strides);
+        const auto in_grid_desc  = make_in_grid_desc<NDim>(N, Wi, C, input_strides);
+        const auto wei_grid_desc = make_wei_grid_desc<NDim>(K, X, C, weights_strides);
+
+        if constexpr(ConvBackwardWeightSpecialization ==
+                     device::ConvolutionBackwardWeightSpecialization::Filter1x1Stride1Pad0)
+        {
+            // A: output tensor
+            const auto out_gemmkpad_gemmm_grid_desc = transform_tensor_descriptor(
+                out_grid_desc,
+                make_tuple(
+                    make_right_pad_transform(GemmKTotal, GemmKPad - GemmKTotal),
+                    make_merge_transform(make_tuple(NumGroupsToMerge, GemmM / NumGroupsToMerge))),
+                make_tuple(Sequence<0>{}, Sequence<1, 2>{}),
+                make_tuple(Sequence<0>{}, Sequence<1>{}));
+
+            const auto out_gemmkbatch_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
+                out_gemmkpad_gemmm_grid_desc,
+                make_tuple(make_unmerge_transform(make_tuple(GemmKBatch * GemmK0, GemmK1Number)),
+                           make_pass_through_transform(GemmM)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}),
+                make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
+
+            // B: input tensor
+            const auto in_gemmkpad_gemmn_grid_desc = transform_tensor_descriptor(
+                in_grid_desc,
+                make_tuple(
+                    make_right_pad_transform(GemmKTotal, GemmKPad - GemmKTotal),
+                    make_merge_transform(make_tuple(NumGroupsToMerge, GemmN / NumGroupsToMerge))),
+                make_tuple(Sequence<0>{}, Sequence<1, 2>{}),
+                make_tuple(Sequence<0>{}, Sequence<1>{}));
+
+            const auto in_gemmkbatch_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
+                in_gemmkpad_gemmn_grid_desc,
+                make_tuple(make_unmerge_transform(make_tuple(GemmKBatch * GemmK0, GemmK1Number)),
+                           make_pass_through_transform(GemmN)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}),
+                make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
+
+            return make_tuple(out_gemmkbatch_gemmk0_gemmm_gemmk1_grid_desc,
+                              in_gemmkbatch_gemmk0_gemmn_gemmk1_grid_desc,
+                              wei_grid_desc);
+        }
+        else
+        {
+            // A: output tensor
+            const auto out_gemmkpad_gemmm_grid_desc = transform_tensor_descriptor(
+                out_grid_desc,
+                make_tuple(
+                    make_right_pad_transform(GemmKTotal, GemmKPad - GemmKTotal),
+                    make_merge_transform(make_tuple(NumGroupsToMerge, GemmM / NumGroupsToMerge))),
+                make_tuple(Sequence<0>{}, Sequence<1, 2>{}),
+                make_tuple(Sequence<0>{}, Sequence<1>{}));
+
+            const auto out_gemmkbatch_gemmk0_gemmm_gemmk1_grid_desc = transform_tensor_descriptor(
+                out_gemmkpad_gemmm_grid_desc,
+                make_tuple(make_unmerge_transform(make_tuple(GemmKBatch * GemmK0, GemmK1Number)),
+                           make_pass_through_transform(GemmM)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}),
+                make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
+
+            // B: input tensor
+            const auto in_n_hip_wip_c_grid_desc = transform_tensor_descriptor(
+                in_grid_desc,
+                make_tuple(make_pass_through_transform(N),
+                           make_pad_transform(Wi, InLeftPadW, InRightPadW),
+                           make_pass_through_transform(NumGroupsToMerge),
+                           make_pass_through_transform(C)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
+                make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}));
+
+            const auto in_n_y_ho_x_wo_c_grid_desc = transform_tensor_descriptor(
+                in_n_hip_wip_c_grid_desc,
+                make_tuple(
+                    make_pass_through_transform(N),
+                    make_embed_transform(make_tuple(X, Wo), make_tuple(ConvDilationW, ConvStrideW)),
+                    make_pass_through_transform(NumGroupsToMerge),
+                    make_pass_through_transform(C)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
+                make_tuple(Sequence<0>{}, Sequence<1, 2>{}, Sequence<3>{}, Sequence<4>{}));
+
+            const auto in_gemmktotal_gemmn_grid_desc = transform_tensor_descriptor(
+                in_n_y_ho_x_wo_c_grid_desc,
+                make_tuple(make_merge_transform(make_tuple(X, NumGroupsToMerge, C)),
+                           make_merge_transform(make_tuple(N, Wo))),
+                make_tuple(Sequence<1, 3, 4>{}, Sequence<0, 2>{}),
+                make_tuple(Sequence<1>{}, Sequence<0>{}));
+
+            const auto in_gemmkpad_gemmn_grid_desc = transform_tensor_descriptor(
+                in_gemmktotal_gemmn_grid_desc,
+                make_tuple(make_right_pad_transform(GemmKTotal, GemmKPad - GemmKTotal),
+                           make_pass_through_transform(GemmN)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}),
+                make_tuple(Sequence<0>{}, Sequence<1>{}));
+
+            const auto in_gemmkbatch_gemmk0_gemmn_gemmk1_grid_desc = transform_tensor_descriptor(
+                in_gemmkpad_gemmn_grid_desc,
+                make_tuple(make_unmerge_transform(make_tuple(GemmKBatch * GemmK0, GemmK1Number)),
+                           make_pass_through_transform(GemmN)),
+                make_tuple(Sequence<0>{}, Sequence<1>{}),
+                make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
+
+            // Padd
+            const auto out_gemmkbatch_gemmk0_gemmm_gemmk1_pad_grid_desc =
+                transform_tensor_descriptor(
+                    out_gemmkbatch_gemmk0_gemmm_gemmk1_grid_desc,
+                    make_tuple(make_pass_through_transform(GemmKBatch * GemmK0),
+                               make_right_pad_transform(GemmM, PadGemmM),
+                               make_pass_through_transform(GemmK1Number)),
+                    make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
+                    make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
+
+            const auto in_gemmkbatch_gemmk0_gemmn_gemmk1_pad_grid_desc =
+                transform_tensor_descriptor(
+                    in_gemmkbatch_gemmk0_gemmn_gemmk1_grid_desc,
+                    make_tuple(make_pass_through_transform(GemmKBatch * GemmK0),
+                               make_right_pad_transform(GemmN, PadGemmN),
+                               make_pass_through_transform(GemmK1Number)),
+                    make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}),
+                    make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
+
+            const auto wei_gemmm_gemmn_pad_grid_desc =
+                transform_tensor_descriptor(wei_grid_desc,
+                                            make_tuple(make_right_pad_transform(GemmM, PadGemmM),
+                                                       make_right_pad_transform(GemmN, PadGemmN)),
+                                            make_tuple(Sequence<0>{}, Sequence<1>{}),
+                                            make_tuple(Sequence<0>{}, Sequence<1>{}));
+
+            return make_tuple(out_gemmkbatch_gemmk0_gemmm_gemmk1_pad_grid_desc,
+                              in_gemmkbatch_gemmk0_gemmn_gemmk1_pad_grid_desc,
+                              wei_gemmm_gemmn_pad_grid_desc);
+        }
+
+    } // function end
+
     template <index_t NDim, typename enable_if<NDim == 2, bool>::type = false>
     static auto MakeABCGridDescriptor_A_K0_M_K1_B_K0_N_K1_C_M_N(
         const index_t N,
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp
index a08d73546d..c0daefee98 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp
@@ -44,32 +44,24 @@ template <ck::index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
           typename ELayout,
-          ConvolutionBackwardWeightSpecialization ConvSpec>
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances = std::tuple<
     // clang-format off
-        //#########################################|     Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer|
-        //#########################################|     Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|
-        //#########################################| Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl|
-        //#########################################|        |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |
-        // generic instance
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  4,   32,   32,    2,    2,  S<1, 4, 16, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              1,              4,      true,  S<1, 4, 16, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              1,              4,      true,           1,           1,   S<1, 16, 1, 4>,               1>,
-        // instances for small conv.K and conv.C
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  4,   32,   32,    2,    1,  S<1, 4, 32, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              1,              1,      true,           1,           1,   S<1, 32, 1, 4>,               1>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  4,   32,   32,    1,    2,  S<1, 4, 8,  2>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              1,              2,      true,  S<1, 4, 16, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 16, 1, 4>,               4>,
-
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   256,   128,     4,  4,   32,   32,    4,    2,  S<1, 4, 64, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              2,      true,           1,           1,   S<1, 32, 1, 8>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   256,     4,  4,   32,   32,    2,    4,  S<1, 4, 32, 2>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              2,      true,  S<1, 4, 64, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 32, 1, 8>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,   128,     4,  4,   32,   32,    4,    2,  S<1, 4, 32, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 32, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   128,     4,  4,   32,   32,    2,    2,  S<1, 4, 32, 2>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              2,      true,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              2,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    64,     4,  4,   32,   32,    2,    2,  S<1, 4, 32, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              2,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    64,   128,     4,  4,   32,   32,    2,    2,  S<1, 4, 16, 2>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              2,      true,  S<1, 4, 32, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  4,   32,   32,    2,    2,  S<1, 4, 16, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 16, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 16, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,    64,     4,  4,   32,   32,    2,    1,  S<1, 4, 32, 2>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              2,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              1,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,    64,   128,     4,  4,   32,   32,    1,    2,  S<1, 4, 16, 4>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              1,      true,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              2,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  4,   32,   32,    2,    1,  S<1, 4, 32, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              1,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    32,   128,     4,  4,   32,   32,    1,    2,  S<1, 4, 8,  4>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              1,      true,  S<1, 4, 32, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 32, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     4,  4,   32,   32,    2,    1,  S<1, 4, 16, 1>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              2,      true,           1,           1,   S<1, 16, 1, 4>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  4,   32,   32,    1,    2,  S<1, 4, 8,  2>,   S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              4,              2,      true,  S<1, 4, 16, 1>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              4,              4,      true,           1,           1,   S<1, 16, 1, 4>,               4>
+        //################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,   32,   8,   16,   16,    1,    1,    S<4, 8,  1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              1,              4,      false,      S<4, 8,  1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,   32,   8,   32,   32,    1,    1,    S<4, 8,  1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              2,              2,      false,      S<4, 16, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 2>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,   32,   8,   32,   32,    1,    2,    S<4, 8,  1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              4,              4,      false,      S<4, 16, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,   32,   8,   32,   32,    1,    4,    S<4, 4,  1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              8,              8,      false,      S<4, 16, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              8,              8,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 8>,
+
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,   32,   8,   16,   16,    1,    1,    S<4, 8,  1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              1,              4,      false,       S<4, 8, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,   32,   8,   32,   32,    1,    1,    S<4, 16, 1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              2,              2,      false,       S<4, 8, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 2>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,   32,   8,   32,   32,    2,    1,    S<4, 16, 1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              4,              4,      false,       S<4, 8, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F32,     F32,     F32,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,   32,   8,   32,   32,    4,    1,    S<4, 16, 1>,        S<2, 0, 1>,     S<1, 0, 2>,            1,              8,              8,      false,       S<4, 4, 1>,    S<2, 0, 1>,     S<1, 0, 2>,             1,              8,              8,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 8>
     // clang-format on
     >;
 
@@ -77,34 +69,32 @@ template <ck::index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
           typename ELayout,
-          ConvolutionBackwardWeightSpecialization ConvSpec>
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances = std::tuple<
     // clang-format off
-        //#########################################|     Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer|
-        //#########################################|     Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|
-        //#########################################| Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl|
-        //#########################################|        |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |
-        // generic instance
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              2,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              2,              4,      true,           1,           1,   S<1, 16, 1, 4>,               2>,   
-        // instance for small conv.K
-        // for fp16 conv.K and conv.C must be divisible by 2
-        // since half_t atomic_add require scalar_per_x_vector % 2 == 0
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              2,              1,      true,           1,           1,   S<1, 32, 1, 4>,               2>,  
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              2,              2,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8>,
-
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 8>,               8>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 8>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 8,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 8,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 16, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8>
+        //################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,     32,   8,   16,   16,    1,    1,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              1,              4,      false,  S<4, 8,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 1, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              2,              2,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     32,   8,   32,   32,    1,    2,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              4,              4,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,     32,   8,   32,   32,    1,    4,  S<4, 4,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              8,              8,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 4, 1, 8>,               2, Scheduler, PipelineVersion, 8, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              2,              2,      false,  S<4, 8,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     32,   8,   32,   32,    2,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              4,              4,      false,  S<4, 8,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,     32,   8,   32,   32,    4,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              8,              8,      false,  S<4, 4,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 8, 1, 4>,               8, Scheduler, PipelineVersion, 8, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              2,              2,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     32,   8,   32,   32,    1,    2,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              4,              4,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,     32,   8,   32,   32,    1,    4,  S<4, 4,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              8,              8,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 4, 1, 8>,               2, Scheduler, PipelineVersion, 8, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     32,   8,   32,   32,    2,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              4,              4,      false,  S<4, 8,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,     32,   8,   32,   32,    4,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              8,              8,      false,  S<4, 4,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 8, 1, 4>,               2, Scheduler, PipelineVersion, 8, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,     32,   8,   32,   32,    1,    1,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              2,              2,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     32,   8,   32,   32,    1,    2,  S<4, 8,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              4,              4,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,     32,   8,   32,   32,    1,    4,  S<4, 4,  1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              8,              8,      false,  S<4, 16, 1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 4, 1, 8>,               2, Scheduler, PipelineVersion, 8, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     32,   8,   32,   32,    2,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              4,              4,      false,  S<4, 8,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, F16, F16>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,     32,   8,   32,   32,    4,    1,  S<4, 16, 1>,        S<2, 0, 1>,  S<1, 0, 2>,                 1,              8,              8,      false,  S<4, 4,  1>,       S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 8, 1, 4>,               2, Scheduler, PipelineVersion, 8, F16, F16>
     // clang-format on
     >;
 
@@ -112,32 +102,31 @@ template <ck::index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
           typename ELayout,
-          ConvolutionBackwardWeightSpecialization ConvSpec>
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances = std::tuple<
     // clang-format off
-        //#########################################|     Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|               ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer|
-        //#########################################|     Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                     Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|
-        //#########################################| Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|             Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl|
-        //#########################################|        |         |          |          |       |        |        |        |            |            |            |                           |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |
+        //################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
         // generic instance
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              1,              4,      true,   S<1, 4, 8,  2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              1,              4,      true,           1,           1,   S<1, 16, 1, 4>,               1>,   
-        // instance for small conv.K
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 4,  8>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              1,              1,      true,           1,           1,   S<1, 32, 1, 4>,               1>, 
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              1,              2,      true,   S<1, 4, 8,  2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               4>,
-
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 16, 4>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 8>,               4>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 16, 4>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              2,      true,   S<1, 4, 32, 2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 8>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 16, 2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 16, 2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 4>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              2,      true,   S<1, 4, 16, 4>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 8,  4>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  4>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              2,      true,   S<1, 4, 16, 2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 8,  2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 4>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              2,      true,   S<1, 4, 8,  8>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 8,  8>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              1,      true,   S<1, 4, 16, 4>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 4,  8>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  8>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              1,      true,   S<1, 4, 16, 2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,    64,    64,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 8,  2>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              4,      true,   S<1, 4, 4,  4>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 16, 1, 4>,               4>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                   ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>, S<0, 3, 1, 2>,   S<0, 2, 1, 3>,              2,              8,              2,      true,   S<1, 4, 8,  2>,   S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               4>
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,    32,  8,   32,   32,    2,    2,     S<4, 8,  2>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              1,              4,      false,     S<4, 8,  2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              1,              4,     false,           1,           1,   S<1, 16, 1, 4>,               2, Scheduler, PipelineVersion, 1>,
+        // instance for small conv.K                                                                                                                                                                                                                                                                                                                                                                                                                                                          
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,    32,  8,   32,   32,    1,    2,     S<4, 4,  4>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              1,              2,      false,     S<4, 8,  2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 16, 1, 4>,               2, Scheduler, PipelineVersion, 4>,                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   256,   128,    32,  8,   32,   32,    4,    2,     S<4, 32, 2>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              4,      false,     S<4, 16, 4>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              2,     false,           1,           1,   S<1, 32, 1, 8>,               4, Scheduler, PipelineVersion, 4>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   256,    32,  8,   32,   32,    2,    4,     S<4, 16, 4>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              2,      false,     S<4, 32, 2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 32, 1, 8>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,   128,    32,  8,   32,   32,    4,    2,     S<4, 16, 2>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              4,      false,     S<4, 16, 2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   128,    32,  8,   32,   32,    2,    2,     S<4, 16, 4>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              2,      false,     S<4, 16, 4>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              2,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    64,    32,  8,   32,   32,    2,    2,     S<4, 16, 2>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              4,      false,     S<4, 8,  4>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              2,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    64,   128,    32,  8,   32,   32,    2,    2,     S<4, 8,  4>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              2,      false,     S<4, 16, 2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,    32,  8,   32,   32,    2,    2,     S<4, 8,  2>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              4,      false,     S<4, 8,  2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 16, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,    64,    32,  8,   32,   32,    2,    1,     S<4, 16, 4>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              2,      false,     S<4, 8,  8>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              1,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,    64,   128,    32,  8,   32,   32,    1,    2,     S<4, 8,  8>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              1,      false,     S<4, 16, 4>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              2,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    32,   128,    32,  8,   32,   32,    1,    2,     S<4, 4,  8>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              1,      false,     S<4, 16, 2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 32, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,    32,  8,   32,   32,    2,    1,     S<4, 8,  2>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              4,      false,     S<4, 4,  4>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              2,     false,           1,           1,   S<1, 16, 1, 4>,               2, Scheduler, PipelineVersion, 4>,   
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,     F32,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,    32,  8,   32,   32,    1,    2,     S<4, 4,  4>,       S<2, 0, 1>,    S<1, 0, 2>,              1,              8,              2,      false,     S<4, 8,  2>,   S<2, 0, 1>,       S<1, 0, 2>,             1,              8,              4,     false,           1,           1,   S<1, 16, 1, 4>,               2, Scheduler, PipelineVersion, 4>
     // clang-format on
     >;
 
@@ -145,34 +134,22 @@ template <ck::index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
           typename ELayout,
-          ConvolutionBackwardWeightSpecialization ConvSpec>
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances = std::tuple<
     // clang-format off
-        //#########################################|     Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer|
-        //#########################################|     Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|
-        //#########################################| Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl|
-        //#########################################|        |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |
-        // generic instance
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              2,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              2,              4,      true,           1,           1,   S<1, 16, 1, 4>,               2>,   
-        // instance for small conv.K
-        // for bf16 conv.K and conv.C must be divisible by 2
-        // since half_t atomic_add require scalar_per_x_vector % 2 == 0
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              2,              1,      true,           1,           1,   S<1, 32, 1, 4>,               2>,  
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              2,              2,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8>,
-
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 8>,               8>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 8>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 8,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 8,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 16, 1, 4>,               8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8>
+        //################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,    32,  8,   16,   16,    1,    1,    S<4, 8,  1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              1,              4,     false,    S<4, 8, 1>,      S<2, 0, 1>,    S<1, 0, 2>,               1,              1,              4,    false,           1,           1,   S<1, 8, 1, 8>,                 2, Scheduler, PipelineVersion, 1>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,    32,  8,   32,   32,    1,    1,    S<4, 8,  1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              2,              2,     false,    S<4, 16, 1>,     S<2, 0, 1>,    S<1, 0, 2>,               1,              2,              2,    false,           1,           1,   S<1, 8, 1, 8>,                 4, Scheduler, PipelineVersion, 2>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,    32,  8,   32,   32,    1,    2,    S<4, 8,  1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              4,              4,     false,    S<4, 16, 1>,     S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,    false,           1,           1,   S<1, 8, 1, 8>,                 2, Scheduler, PipelineVersion, 4>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,    32,  8,   32,   32,    1,    4,    S<4, 4,  1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              8,              8,     false,    S<4, 16, 1>,     S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,    false,           1,           1,   S<1, 8, 1, 8>,                 2, Scheduler, PipelineVersion, 8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,    32,  8,   32,   32,    1,    1,    S<4, 16, 1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              2,              2,     false,    S<4, 8, 1>,      S<2, 0, 1>,    S<1, 0, 2>,               1,              2,              2,    false,           1,           1,   S<1, 8, 1, 8>,                 4, Scheduler, PipelineVersion, 2>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,    32,  8,   32,   32,    2,    1,    S<4, 16, 1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              4,              4,     false,    S<4, 8, 1>,      S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,    false,           1,           1,   S<1, 8, 1, 8>,                 2, Scheduler, PipelineVersion, 4>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,   BF16,    BF16,    BF16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,    32,  8,   32,   32,    4,    1,    S<4, 16, 1>,       S<2, 0, 1>,     S<1, 0, 2>,               1,              8,              8,     false,    S<4, 4, 1>,      S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,    false,           1,           1,   S<1, 8, 1, 8>,                 2, Scheduler, PipelineVersion, 8>
     // clang-format on
     >;
 
@@ -180,35 +157,35 @@ template <ck::index_t NDimSpatial,
           typename ALayout,
           typename BLayout,
           typename ELayout,
-          ConvolutionBackwardWeightSpecialization ConvSpec>
+          ConvolutionBackwardWeightSpecialization ConvSpec,
+          BlockGemmPipelineScheduler Scheduler,
+          BlockGemmPipelineVersion PipelineVersion>
 using device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances = std::tuple<
 // clang-format off
-        //#########################################|     Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| Compute| Compute|
-        //#########################################|     Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|   TypeA|   TypeB|
-        //#########################################| Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl|        |        |
-        //#########################################|        |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |        |        |
+        //################################|              Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|              ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer| BlockGemm| BlockGemm| NumGroups|
+        //################################|              Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                    Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|  Pipeline|  Pipeline|   ToMerge|
+        //################################|          Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|            Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl| Scheduler|   Version|          |
+        //################################|                 |         |          |          |       |        |        |        |            |            |            |                          |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |          |          |          |
 #if defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-        // generic instance
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              2,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              2,              4,      true,           1,           1,   S<1, 16, 1, 4>,               2,       BF8,     F8>,   
-        // instance for small conv.K
-        // for fp16 conv.K and conv.C must be divisible by 2
-        // since half_t atomic_add require scalar_per_x_vector % 2 == 0
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              2,              1,      true,           1,           1,   S<1, 32, 1, 4>,               2,       BF8,     F8>,  
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              2,              2,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8,       BF8,     F8>,
-
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   256,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 8>,               8,       BF8,     F8>,
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   256,     4,  8,   32,   32,    2,    4,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 32, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 8>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,   128,     4,  8,   32,   32,    4,    2,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    64,   128,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    64,     4,  8,   32,   32,    2,    2,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,   128,    64,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 8,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   256,    64,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 8,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,   128,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              1,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,     4,  8,   32,   32,    2,    1,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              4,      true,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              2,      true,           1,           1,   S<1, 16, 1, 4>,               8,       BF8,     F8>,   
-        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  4>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              2,      true,  S<1, 4, 8,  2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 16, 1, 4>,               8,       BF8,     F8>
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    16,    16,    32,   8,  16,   16,    1,    1,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              1,              4,      false,    S<4, 8,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              1,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 1, BF8, F8>,
+                           
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,    32,   8,  32,   32,    1,    1,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              2,              2,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 2, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,    32,   8,  32,   32,    1,    2,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,    32,   8,  32,   32,    1,    4,     S<4, 4,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 4, 1, 8>,               2, Scheduler, PipelineVersion, 8, BF8, F8>,
+
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,    32,   8,  32,   32,    1,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              2,              2,      false,    S<4, 8,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,    32,   8,  32,   32,    2,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,      false,    S<4, 8,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,    32,   8,  32,   32,    4,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,      false,    S<4, 4,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 8, 1, 4>,               2, Scheduler, PipelineVersion, 8, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,    32,   8,  32,   32,    1,    1,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              2,              2,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,    32,   8,  32,   32,    1,    2,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,    32,   8,  32,   32,    1,    4,     S<4, 4,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 4, 1, 8>,               2, Scheduler, PipelineVersion, 8, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,    32,   8,  32,   32,    2,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,      false,    S<4, 8,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,    32,   8,  32,   32,    4,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,      false,    S<4, 4,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 8, 1, 4>,               2, Scheduler, PipelineVersion, 8, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    32,    32,   8,  32,   32,    1,    1,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              2,              2,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              2,              2,      false,           1,           1,   S<1, 8, 1, 8>,               4, Scheduler, PipelineVersion, 2, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,    64,    32,   8,  32,   32,    1,    2,     S<4, 8,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    32,   128,    32,   8,  32,   32,    1,    4,     S<4, 4,  1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,      false,    S<4, 16, 1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 4, 1, 8>,               2, Scheduler, PipelineVersion, 8, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,    64,    32,    32,   8,  32,   32,    2,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              4,              4,      false,    S<4, 8,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              4,              4,      false,           1,           1,   S<1, 8, 1, 8>,               2, Scheduler, PipelineVersion, 4, BF8, F8>,
+        DeviceGroupedConvBwdWeight_Xdl_CShuffle< NDimSpatial,  ALayout,   BLayout,   ELayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,    64,   128,    32,    32,   8,  32,   32,    4,    1,     S<4, 16, 1>,       S<2, 0, 1>,    S<1, 0, 2>,               1,              8,              8,      false,    S<4, 4,  1>,     S<2, 0, 1>,  S<1, 0, 2>,                1,              8,              8,      false,           1,           1,   S<1, 8, 1, 4>,               2, Scheduler, PipelineVersion, 8, BF8, F8>
 #endif
     // clang-format on
     >;
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight.hpp b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight.hpp
index db17f0f381..6e0d318e6d 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight.hpp
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight.hpp
@@ -276,7 +276,14 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instances(
+                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_FP16
@@ -284,7 +291,14 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instances(
+                        op_ptrs);
                 }
 #endif
 #ifdef CK_ENABLE_BF16
@@ -293,7 +307,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -309,7 +329,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -318,7 +344,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instances(
                         op_ptrs);
                 }
 #endif
@@ -328,7 +360,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -341,7 +379,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -350,7 +394,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instances(
                         op_ptrs);
                     add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev2_instances(
                         op_ptrs);
@@ -364,7 +414,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instances(
                         op_ptrs);
                 }
                 if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
@@ -373,7 +429,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instances(
                         op_ptrs);
                     add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_bf16_pipev2_instances(
                         op_ptrs);
@@ -421,7 +483,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -430,7 +498,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -440,7 +514,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
@@ -453,7 +533,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, float> && is_same_v<ComputeTypeA, float> &&
                              is_same_v<ComputeTypeB, float>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instances(
                         op_ptrs);
                 }
 #endif
@@ -462,7 +548,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, half_t> &&
                              is_same_v<ComputeTypeB, half_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instances(
                         op_ptrs);
                     add_device_grouped_conv3d_bwd_weight_two_stage_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pipev2_instances(
                         op_ptrs);
@@ -476,7 +568,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instances(
                         op_ptrs);
                 }
                 if constexpr(is_same_v<InDataType, ck::bhalf_t> &&
@@ -485,7 +583,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<ComputeTypeA, ck::bhalf_t> &&
                              is_same_v<ComputeTypeB, ck::bhalf_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instances(
                         op_ptrs);
                     add_device_grouped_conv3d_bwd_weight_two_stage_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pipev2_instances(
                         op_ptrs);
@@ -498,7 +602,13 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
                              is_same_v<OutDataType, half_t> && is_same_v<ComputeTypeA, bf8_t> &&
                              is_same_v<ComputeTypeB, f8_t>)
                 {
-                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_default_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_default_pipev5_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_pad0_pipev2_instances(
+                        op_ptrs);
+                    add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_pad0_pipev5_instances(
                         op_ptrs);
                 }
 #endif
diff --git a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_xdl.inc b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_xdl.inc
index 132dde81ae..cedf82c717 100644
--- a/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_xdl.inc
+++ b/library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight_xdl.inc
@@ -10,7 +10,40 @@ namespace instance {
 
 // conv1d backward weight
 #ifdef CK_ENABLE_BF16
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
                                                            GNWC,
                                                            GKXC,
@@ -23,7 +56,40 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_insta
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP16
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
                                                            GNWC,
                                                            GKXC,
@@ -36,7 +102,40 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP32
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
                                                            GNWC,
                                                            GKXC,
@@ -50,7 +149,40 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(
 #endif
 // conv2d backward weight
 #ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            GNHWC,
                                                            GKYXC,
@@ -63,7 +195,40 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_in
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            GNHWC,
                                                            GKYXC,
@@ -76,7 +241,40 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            GNHWC,
                                                            GKYXC,
@@ -89,7 +287,7 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_BF16
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
                                                            GKYXC,
@@ -100,8 +298,73 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
                                                            GKYXC,
@@ -112,7 +375,6 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_in
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
 void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_bf16_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
@@ -136,83 +398,254 @@ void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_bf16_p
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_bf16_pipev2_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NGCHW,
-                                                           GKYXC,
-                                                           NGKHW,
-                                                           BF16,
+void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_bf16_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NGCHW,
+                                                           GKYXC,
+                                                           NGKHW,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_bf16_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NGCHW,
+                                                           GKYXC,
+                                                           NGKHW,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP16
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_f16_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NGCHW,
+                                                           GKYXC,
+                                                           NGKHW,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+
+void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_f16_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NGCHW,
+                                                           GKYXC,
+                                                           NGKHW,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+#ifdef CK_ENABLE_FP32
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+#endif
+// conv3d backward weight
+#ifdef CK_ENABLE_BF16
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            BF16,
+                                                           F32,
                                                            BF16,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_bf16_pipev5_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NGCHW,
-                                                           GKYXC,
-                                                           NGKHW,
-                                                           BF16,
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            BF16,
+                                                           F32,
                                                            BF16,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP16
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
-                                                           PassThrough,
-                                                           PassThrough,
-                                                           PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev2_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            F16,
                                                            F16,
                                                            F16,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev5_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            F16,
                                                            F16,
                                                            F16,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_f16_pipev2_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NGCHW,
-                                                           GKYXC,
-                                                           NGKHW,
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            F16,
                                                            F16,
                                                            F16,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
-void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_f16_pipev5_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NGCHW,
-                                                           GKYXC,
-                                                           NGKHW,
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            F16,
                                                            F16,
                                                            F16,
@@ -221,47 +654,40 @@ void add_device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_f16_pi
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP32
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
-    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
-                                                           NHWGC,
-                                                           GKYXC,
-                                                           NHWGK,
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
                                                            F32,
                                                            F32,
                                                            F32,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-#endif
-// conv3d backward weight
-#ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            GNDHWC,
                                                            GKZYXC,
                                                            GNDHWK,
-                                                           BF16,
                                                            F32,
-                                                           BF16,
+                                                           F32,
+                                                           F32,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            GNDHWC,
                                                            GKZYXC,
                                                            GNDHWK,
-                                                           F16,
-                                                           F16,
-                                                           F16,
+                                                           F32,
+                                                           F32,
+                                                           F32,
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-#endif
-#ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            GNDHWC,
                                                            GKZYXC,
@@ -274,7 +700,7 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_BF16
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -285,8 +711,73 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -297,7 +788,6 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
 void add_device_grouped_conv3d_bwd_weight_two_stage_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
@@ -346,7 +836,40 @@ void add_device_grouped_conv3d_bwd_weight_two_stage_xdl_ngcdhw_gkzyxc_ngkdhw_bf1
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP16
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -357,7 +880,6 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances
                                                            PassThrough,
                                                            PassThrough,
                                                            PassThrough>>>& instances);
-
 void add_device_grouped_conv3d_bwd_weight_two_stage_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
@@ -406,7 +928,40 @@ void add_device_grouped_conv3d_bwd_weight_two_stage_xdl_ngcdhw_gkzyxc_ngkdhw_f16
                                                            PassThrough>>>& instances);
 #endif
 #ifdef CK_ENABLE_FP32
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -419,7 +974,46 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances
                                                            PassThrough>>>& instances);
 #endif
 #if defined CK_ENABLE_FP16 && defined CK_ENABLE_FP8 && defined CK_ENABLE_BF8
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances);
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_pad0_pipev5_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/CMakeLists.txt
index b057e0c8d2..923d3c15eb 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/CMakeLists.txt
@@ -1,8 +1,17 @@
 # ONLY XDL_AND_DL_KERNELS
 set(GROUPED_CONV1D_BWD_WEIGHT
-    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instance.cpp
-    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instance.cpp
-    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instance.cpp)
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instance.cpp)
 
 if(DL_KERNELS)
     list(APPEND GROUPED_CONV1D_BWD_WEIGHT
@@ -15,3 +24,4 @@ if(DL_KERNELS)
 endif()
 
 add_instance_library(device_grouped_conv1d_bwd_weight_instance ${GROUPED_CONV1D_BWD_WEIGHT})
+
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instance.cpp
similarity index 75%
rename from library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instance.cpp
index 9c97d80c82..a23690d9ad 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_instances(
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
                                                            GNWC,
                                                            GKXC,
@@ -21,7 +21,6 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_insta
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
     add_device_operation_instances(
         instances,
         device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
@@ -29,16 +28,9 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_insta
             GNWC,
             GKXC,
             GNWK,
-            ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
-            1,
-            GNWC,
-            GKXC,
-            GNWK,
-            ConvBwdWeightFilter1x1Stride1Pad0>{});
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..dd8b3fbf83
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            1,
+            GNWC,
+            GKXC,
+            GNWK,
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
new file mode 100644
index 0000000000..f62e5c0200
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            1,
+            GNWC,
+            GKXC,
+            GNWK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..bb98dd3de9
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_bf16_f32_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            1,
+            GNWC,
+            GKXC,
+            GNWK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..d7a6e24ff3
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       1,
+                                       GNWC,
+                                       GKXC,
+                                       GNWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..f70de63343
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       1,
+                                       GNWC,
+                                       GKXC,
+                                       GNWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instance.cpp
similarity index 72%
rename from library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instance.cpp
index edba11efba..ef13e05934 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
                                                            GNWC,
                                                            GKXC,
@@ -21,22 +21,15 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<1,
-                                                                   GNWC,
-                                                                   GKXC,
-                                                                   GNWK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
                                        1,
                                        GNWC,
                                        GKXC,
                                        GNWK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..768a7c0c5e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       1,
+                                       GNWC,
+                                       GKXC,
+                                       GNWK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..2d4f89f912
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       1,
+                                       GNWC,
+                                       GKXC,
+                                       GNWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..2093af67bd
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       1,
+                                       GNWC,
+                                       GKXC,
+                                       GNWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instance.cpp
similarity index 72%
rename from library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instance.cpp
index 5f28994f09..ecf06901b6 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
                                                            GNWC,
                                                            GKXC,
@@ -21,22 +21,15 @@ void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<1,
-                                                                   GNWC,
-                                                                   GKXC,
-                                                                   GNWK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
                                        1,
                                        GNWC,
                                        GKXC,
                                        GNWK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..fa85df8fab
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv1d_bwd_weight/xdl/device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv1d_bwd_weight_xdl_gnwc_gkxc_gnwk_f32_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<1,
+                                                           GNWC,
+                                                           GKXC,
+                                                           GNWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       1,
+                                       GNWC,
+                                       GKXC,
+                                       GNWK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/CMakeLists.txt
index 546a62a8af..d3ad680b41 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/CMakeLists.txt
@@ -1,12 +1,33 @@
 # ONLY XDL_AND_DL_KERNELS
 set(GROUPED_CONV2D_BWD_WEIGHT
-    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instance.cpp
-    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instance.cpp
-    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instance.cpp
-    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instance.cpp
-    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instance.cpp
-    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instance.cpp
-    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instance.cpp
     xdl/device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev2_instance.cpp
     xdl/device_grouped_conv2d_bwd_weight_two_stage_xdl_nhwgc_gkyxc_nhwgk_f16_pipev5_instance.cpp
     xdl/device_grouped_conv2d_bwd_weight_two_stage_xdl_ngchw_gkyxc_ngkhw_f16_pipev2_instance.cpp
@@ -28,3 +49,4 @@ if(DL_KERNELS)
 endif()
 
 add_instance_library(device_grouped_conv2d_bwd_weight_instance ${GROUPED_CONV2D_BWD_WEIGHT})
+
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instance.cpp
similarity index 76%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instance.cpp
index 69e22dee41..e0a067637c 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            GNHWC,
                                                            GKYXC,
@@ -22,7 +22,6 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_in
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
     add_device_operation_instances(
         instances,
         device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
@@ -30,16 +29,9 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_in
             GNHWC,
             GKYXC,
             GNHWK,
-            ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
-            2,
-            GNHWC,
-            GKYXC,
-            GNHWK,
-            ConvBwdWeightFilter1x1Stride1Pad0>{});
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..c4f14df51f
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            2,
+            GNHWC,
+            GKYXC,
+            GNHWK,
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
new file mode 100644
index 0000000000..5b9d1ae2fa
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            2,
+            GNHWC,
+            GKYXC,
+            GNHWK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..11647b20cc
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_bf16_f32_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            2,
+            GNHWC,
+            GKYXC,
+            GNHWK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..5ae84d658e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       2,
+                                       GNHWC,
+                                       GKYXC,
+                                       GNHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..67f2cc23f9
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       2,
+                                       GNHWC,
+                                       GKYXC,
+                                       GNHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instance.cpp
similarity index 73%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instance.cpp
index abb90b7c71..9968da0547 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            GNHWC,
                                                            GKYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<2,
-                                                                   GNHWC,
-                                                                   GKYXC,
-                                                                   GNHWK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
                                        2,
                                        GNHWC,
                                        GKYXC,
                                        GNHWK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..f78a064c5f
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       2,
+                                       GNHWC,
+                                       GKYXC,
+                                       GNHWK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..cc19b6239c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       2,
+                                       GNHWC,
+                                       GKYXC,
+                                       GNHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..7f9096e98b
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       2,
+                                       GNHWC,
+                                       GKYXC,
+                                       GNHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instance.cpp
similarity index 73%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instance.cpp
index 2e0fef9cf8..0b0664f9b2 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            GNHWC,
                                                            GKYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<2,
-                                                                   GNHWC,
-                                                                   GKYXC,
-                                                                   GNHWK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
                                        2,
                                        GNHWC,
                                        GKYXC,
                                        GNHWK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..667cd0a5f4
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_gnhwc_gkyxc_gnhwk_f32_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           GNHWC,
+                                                           GKYXC,
+                                                           GNHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       2,
+                                       GNHWC,
+                                       GKYXC,
+                                       GNHWK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..26d9a07de9
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..85be6b7c6e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
similarity index 80%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
index cac9353354..ee0415762b 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
                                                            GKYXC,
@@ -22,7 +22,6 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_in
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
     add_device_operation_instances(
         instances,
         device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
@@ -30,16 +29,9 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_in
             NHWGC,
             GKYXC,
             NHWGK,
-            ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
-            2,
-            NHWGC,
-            GKYXC,
-            NHWGK,
-            ConvBwdWeightFilter1x1Stride1Pad0>{});
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..3cefa52a3c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            2,
+            NHWGC,
+            GKYXC,
+            NHWGK,
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
new file mode 100644
index 0000000000..32c3320e5d
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            2,
+            NHWGC,
+            GKYXC,
+            NHWGK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..93833e8c7c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_f32_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            2,
+            NHWGC,
+            GKYXC,
+            NHWGK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instance.cpp
similarity index 76%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instance.cpp
index ee71e37e79..d9269180cb 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instance.cpp
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
                                                            GKYXC,
@@ -22,22 +22,16 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<2,
-                                                                    NHWGC,
-                                                                    GKYXC,
-                                                                    NHWGK,
-                                                                    ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
+
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
                                        2,
                                        NHWGC,
                                        GKYXC,
                                        NHWGK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..8694612d1c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..0fa04cc1d5
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+    ;
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..0a35297081
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+    ;
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instance.cpp
similarity index 73%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instance.cpp
index f6e1ada352..687e0f7f59 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
                                                            GKYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<2,
-                                                                   NHWGC,
-                                                                   GKYXC,
-                                                                   NHWGK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
                                        2,
                                        NHWGC,
                                        GKYXC,
                                        NHWGK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..e32a9d22bd
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..ae91e47bf6
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..bde4d5b1a2
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instance.cpp
similarity index 73%
rename from library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instance.cpp
index 384706414a..e616d944be 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
                                                            NHWGC,
                                                            GKYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_instances(
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<2,
-                                                                   NHWGC,
-                                                                   GKYXC,
-                                                                   NHWGK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
                                        2,
                                        NHWGC,
                                        GKYXC,
                                        NHWGK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..caaf847ca6
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv2d_bwd_weight/xdl/device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv2d_bwd_weight_xdl_nhwgc_gkyxc_nhwgk_f32_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<2,
+                                                           NHWGC,
+                                                           GKYXC,
+                                                           NHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       2,
+                                       NHWGC,
+                                       GKYXC,
+                                       NHWGK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/CMakeLists.txt b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/CMakeLists.txt
index c8c30897cf..21a2e849a7 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/CMakeLists.txt
@@ -1,12 +1,33 @@
  # XDL_DL_WMMA_KERNELS
 set(GROUPED_CONV3D_BWD_WEIGHT
-     xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instance.cpp
-     xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instance.cpp
-     xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instance.cpp
-     xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp
-     xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instance.cpp
-     xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instance.cpp
-     xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instance.cpp
+    xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev5_instance.cpp
     xdl/device_grouped_conv3d_bwd_weight_two_stage_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pipev2_instance.cpp
     xdl/device_grouped_conv3d_bwd_weight_two_stage_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pipev5_instance.cpp
     xdl/device_grouped_conv3d_bwd_weight_two_stage_xdl_ngcdhw_gkzyxc_ngkdhw_f16_pipev2_instance.cpp
@@ -39,7 +60,10 @@ list(APPEND GROUPED_CONV3D_BWD_WEIGHT
 
 if((DTYPES MATCHES "fp8" AND DTYPES MATCHES "bf8" AND DTYPES MATCHES "fp16") OR NOT DEFINED DTYPES)
     list(APPEND GROUPED_CONV3D_BWD_WEIGHT
-      xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_instance.cpp)
+        xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev2_instance.cpp
+        xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev5_instance.cpp
+        xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev2_instance.cpp
+        xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev5_instance.cpp)
 endif()
 
 add_instance_library(device_grouped_conv3d_bwd_weight_instance ${GROUPED_CONV3D_BWD_WEIGHT})
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instance.cpp
similarity index 75%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instance.cpp
index 81d64344f7..ad164add3a 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            GNDHWC,
                                                            GKZYXC,
@@ -21,7 +21,6 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
     add_device_operation_instances(
         instances,
         device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
@@ -29,16 +28,9 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16
             GNDHWC,
             GKZYXC,
             GNDHWK,
-            ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
-            3,
-            GNDHWC,
-            GKZYXC,
-            GNDHWK,
-            ConvBwdWeightFilter1x1Stride1Pad0>{});
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..7fb024d144
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            3,
+            GNDHWC,
+            GKZYXC,
+            GNDHWK,
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
new file mode 100644
index 0000000000..32a0f9d633
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            3,
+            GNDHWC,
+            GKZYXC,
+            GNDHWK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..4d47ad063b
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_bf16_f32_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            3,
+            GNDHWC,
+            GKZYXC,
+            GNDHWK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..24daa8aa09
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       3,
+                                       GNDHWC,
+                                       GKZYXC,
+                                       GNDHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..afcd613328
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       3,
+                                       GNDHWC,
+                                       GKZYXC,
+                                       GNDHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instance.cpp
similarity index 72%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instance.cpp
index d03f0a7bac..c77761b745 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            GNDHWC,
                                                            GKZYXC,
@@ -21,22 +21,15 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_instances
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<3,
-                                                                   GNDHWC,
-                                                                   GKZYXC,
-                                                                   GNDHWK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
                                        3,
                                        GNDHWC,
                                        GKZYXC,
                                        GNDHWK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..0e445ea5a8
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instance.cpp
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       3,
+                                       GNDHWC,
+                                       GKZYXC,
+                                       GNDHWK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..ca1376f9e7
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instance.cpp
@@ -0,0 +1,37 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       3,
+                                       GNDHWC,
+                                       GKZYXC,
+                                       GNDHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..c71940ebd3
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instance.cpp
@@ -0,0 +1,37 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       3,
+                                       GNDHWC,
+                                       GKZYXC,
+                                       GNDHWK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instance.cpp
similarity index 72%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instance.cpp
index 7c24cc8fd4..bef77308fb 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            GNDHWC,
                                                            GKZYXC,
@@ -21,24 +21,16 @@ void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_instances
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<3,
-                                                                   GNDHWC,
-                                                                   GKZYXC,
-                                                                   GNDHWK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
                                        3,
                                        GNDHWC,
                                        GKZYXC,
                                        GNDHWK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
-
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..12a6d35c38
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instance.cpp
@@ -0,0 +1,37 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_gndhwc_gkzyxc_gndhwk_f32_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           GNDHWC,
+                                                           GKZYXC,
+                                                           GNDHWK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       3,
+                                       GNDHWC,
+                                       GKZYXC,
+                                       GNDHWK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..830202f40c
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..cc3fc3674b
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
similarity index 80%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
index 679f30a3d9..7e4d7ef547 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instance.cpp
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -22,7 +22,6 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
     add_device_operation_instances(
         instances,
         device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
@@ -30,16 +29,9 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16
             NDHWGC,
             GKZYXC,
             NDHWGK,
-            ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
-            3,
-            NDHWGC,
-            GKZYXC,
-            NDHWGK,
-            ConvBwdWeightFilter1x1Stride1Pad0>{});
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..c332b26cd2
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            3,
+            NDHWGC,
+            GKZYXC,
+            NDHWGK,
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
new file mode 100644
index 0000000000..0e28889ea3
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            3,
+            NDHWGC,
+            GKZYXC,
+            NDHWGK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..77a8ed852e
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,40 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_f32_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           F32,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_f32_bf16_instances<
+            3,
+            NDHWGC,
+            GKZYXC,
+            NDHWGK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instance.cpp
similarity index 76%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instance.cpp
index f1ea371819..7f13e01b60 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instance.cpp
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_instance
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<3,
-                                                                    NDHWGC,
-                                                                    GKZYXC,
-                                                                    NDHWGK,
-                                                                    ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
                                        3,
                                        NDHWGC,
                                        GKZYXC,
                                        NDHWGK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..7843a32ddd
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_bf16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           BF16,
+                                                           BF16,
+                                                           BF16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_bf16_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev2_instance.cpp
similarity index 78%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev2_instance.cpp
index 6e7f22b7e5..48b72bfe23 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
@@ -9,7 +9,7 @@ namespace tensor_operation {
 namespace device {
 namespace instance {
 
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_default_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -25,16 +25,7 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_
 {
 #if CK_BUILD_DEPRECATED
 #pragma message "These instances are getting deprecated"
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances<
-            3,
-            NDHWGC,
-            GKZYXC,
-            NDHWGK,
-            ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
+
     add_device_operation_instances(
         instances,
         device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances<
@@ -42,7 +33,10 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_
             NDHWGC,
             GKZYXC,
             NDHWGK,
-            ConvBwdWeightFilter1x1Stride1Pad0>{});
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+
 #else
 #pragma message "These instances were deprecated"
     std::ignore = instances;
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..80aaf9371d
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_default_pipev5_instance.cpp
@@ -0,0 +1,49 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances)
+{
+#if CK_BUILD_DEPRECATED
+#pragma message "These instances are getting deprecated"
+
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances<
+            3,
+            NDHWGC,
+            GKZYXC,
+            NDHWGK,
+            ConvBwdWeightDefault,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+
+#else
+#pragma message "These instances were deprecated"
+    std::ignore = instances;
+#endif
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev2_instance.cpp
new file mode 100644
index 0000000000..5ba1d00fab
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev2_instance.cpp
@@ -0,0 +1,47 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_pad0_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances)
+{
+#if CK_BUILD_DEPRECATED
+#pragma message "These instances are getting deprecated"
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances<
+            3,
+            NDHWGC,
+            GKZYXC,
+            NDHWGK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v2>{});
+#else
+#pragma message "These instances were deprecated"
+    std::ignore = instances;
+#endif
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..e3b5f94717
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_fp8_pad0_pipev5_instance.cpp
@@ -0,0 +1,47 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_comp_bf8_f8_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           BF8,
+                                                           F8>>>& instances)
+{
+#if CK_BUILD_DEPRECATED
+#pragma message "These instances are getting deprecated"
+    add_device_operation_instances(
+        instances,
+        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_comp_bf8_f8_instances<
+            3,
+            NDHWGC,
+            GKZYXC,
+            NDHWGK,
+            ConvBwdWeightFilter1x1Stride1Pad0,
+            BlockGemmPipelineScheduler::Intrawave,
+            BlockGemmPipelineVersion::v5>{});
+#else
+#pragma message "These instances were deprecated"
+    std::ignore = instances;
+#endif
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..7489addecc
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..956e20ae49
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instance.cpp
similarity index 73%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instance.cpp
index eba721c7b8..a6c31b0108 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_instances
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<3,
-                                                                   NDHWGC,
-                                                                   GKZYXC,
-                                                                   NDHWGK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
                                        3,
                                        NDHWGC,
                                        GKZYXC,
                                        NDHWGK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..c5a2a6f908
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f16_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F16,
+                                                           F16,
+                                                           F16,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f16_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instance.cpp
new file mode 100644
index 0000000000..d061d033e9
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev2_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instance.cpp
new file mode 100644
index 0000000000..4145ccf376
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_default_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightDefault,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instance.cpp
similarity index 73%
rename from library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instance.cpp
rename to library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instance.cpp
index 7dd289139c..3e5b4bb6bc 100644
--- a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instance.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
@@ -10,7 +10,7 @@ namespace device {
 namespace instance {
 
 // Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
-void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances(
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev2_instances(
     std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
                                                            NDHWGC,
                                                            GKZYXC,
@@ -22,22 +22,15 @@ void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_instances
                                                            PassThrough,
                                                            PassThrough>>>& instances)
 {
-    // 1. Default
-    add_device_operation_instances(
-        instances,
-        device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<3,
-                                                                   NDHWGC,
-                                                                   GKZYXC,
-                                                                   NDHWGK,
-                                                                   ConvBwdWeightDefault>{});
-    // 2. Filter1x1Stride1Pad0
     add_device_operation_instances(instances,
                                    device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
                                        3,
                                        NDHWGC,
                                        GKZYXC,
                                        NDHWGK,
-                                       ConvBwdWeightFilter1x1Stride1Pad0>{});
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v2>{});
 }
 
 } // namespace instance
diff --git a/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev5_instance.cpp b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev5_instance.cpp
new file mode 100644
index 0000000000..58922a8647
--- /dev/null
+++ b/library/src/tensor_operation_instance/gpu/grouped_conv3d_bwd_weight/xdl/device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev5_instance.cpp
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_conv_bwd_weight/device_grouped_conv_bwd_weight_xdl_instance.hpp"
+
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+
+// Compilation parameters for in[n, hi, wi, g, c] * wei[g, k, y, x, c] = out[n, ho, wo, g, k]
+void add_device_grouped_conv3d_bwd_weight_xdl_ndhwgc_gkzyxc_ndhwgk_f32_pad0_pipev5_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvBwdWeight<3,
+                                                           NDHWGC,
+                                                           GKZYXC,
+                                                           NDHWGK,
+                                                           F32,
+                                                           F32,
+                                                           F32,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           PassThrough>>>& instances)
+{
+    add_device_operation_instances(instances,
+                                   device_grouped_conv_bwd_weight_xdl_c_shuffle_f32_instances<
+                                       3,
+                                       NDHWGC,
+                                       GKZYXC,
+                                       NDHWGK,
+                                       ConvBwdWeightFilter1x1Stride1Pad0,
+                                       BlockGemmPipelineScheduler::Intrawave,
+                                       BlockGemmPipelineVersion::v5>{});
+}
+
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
diff --git a/test/grouped_convnd_bwd_weight/CMakeLists.txt b/test/grouped_convnd_bwd_weight/CMakeLists.txt
index 313b5ba4ca..539f137336 100644
--- a/test/grouped_convnd_bwd_weight/CMakeLists.txt
+++ b/test/grouped_convnd_bwd_weight/CMakeLists.txt
@@ -7,7 +7,7 @@ if(GPU_TARGETS MATCHES "gfx9" OR DL_KERNELS)
 endif()
 add_gtest_executable(test_grouped_convnd_bwd_weight_interface_xdl test_grouped_convnd_bwd_weight_interface_xdl.cpp)
 if(result EQUAL 0)
-   target_link_libraries(test_grouped_convnd_bwd_weight_interface_xdl PRIVATE utility)
+   target_link_libraries(test_grouped_convnd_bwd_weight_interface_xdl PRIVATE utility device_grouped_conv1d_bwd_weight_instance device_grouped_conv2d_bwd_weight_instance device_grouped_conv3d_bwd_weight_instance)
 endif()
 add_gtest_executable(test_grouped_convnd_bwd_weight_interface_wmma test_grouped_convnd_bwd_weight_interface_wmma.cpp)
 if(result EQUAL 0)
diff --git a/test/grouped_convnd_bwd_weight/test_grouped_convnd_bwd_weight_interface_xdl.cpp b/test/grouped_convnd_bwd_weight/test_grouped_convnd_bwd_weight_interface_xdl.cpp
index cfbf13f00e..fb951413b1 100644
--- a/test/grouped_convnd_bwd_weight/test_grouped_convnd_bwd_weight_interface_xdl.cpp
+++ b/test/grouped_convnd_bwd_weight/test_grouped_convnd_bwd_weight_interface_xdl.cpp
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <cstdlib>
 #include <iostream>
@@ -12,69 +12,143 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_weight_xdl_cshuffle.hpp"
-
+#include "ck/library/reference_tensor_operation/cpu/reference_conv_bwd_weight.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_convolution_backward_weight.hpp"
+#include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/convolution_parameter.hpp"
 #include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
+#include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/device_memory.hpp"
 
 #include <gtest/gtest.h>
 
-using F16         = ck::half_t;
-using F32         = float;
+namespace ctl = ck::tensor_layout::convolution;
+
 using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
+using InDataType  = ck::bhalf_t;
+using WeiDataType = float;
+using OutDataType = ck::bhalf_t;
+using AccDataType = float;
 template <ck::index_t... Is>
+
 using S = ck::Sequence<Is...>;
-using ConvolutionBackwardWeightSpecialization =
-    ck::tensor_operation::device::ConvolutionBackwardWeightSpecialization;
 
-static constexpr auto ConvBwdWeightDefault = ConvolutionBackwardWeightSpecialization::Default;
-static constexpr auto Filter1x1Stride1Pad0 =
-    ConvolutionBackwardWeightSpecialization::Filter1x1Stride1Pad0;
+static constexpr auto ConvBwdWeightDefault =
+    ck::tensor_operation::device::ConvolutionBackwardWeightSpecialization::Default;
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
 
-template <typename Tuple, ConvolutionBackwardWeightSpecialization ConvSpec>
-class TestGroupedConvndBwdWeight : public ::testing::Test
+template <typename InputLay, typename WeightLay, typename OutputLay>
+struct CommonLayoutSetting
 {
-    protected:
-    static constexpr ck::index_t NDimSpatial = 2;
+    using InputLayout  = InputLay;
+    using WeightLayout = WeightLay;
+    using OutputLayout = OutputLay;
+};
+
+template <ck::index_t NDimSpatial>
+struct CommonLayoutSettingSelector
+    : CommonLayoutSetting<ck::tuple_element_t<NDimSpatial - 1,
+                                              ck::Tuple<ck::tensor_layout::convolution::GNWC,
+                                                        ck::tensor_layout::convolution::GNHWC,
+                                                        ck::tensor_layout::convolution::GNDHWC>>,
+                          ck::tuple_element_t<NDimSpatial - 1,
+                                              ck::Tuple<ck::tensor_layout::convolution::GKXC,
+                                                        ck::tensor_layout::convolution::GKYXC,
+                                                        ck::tensor_layout::convolution::GKZYXC>>,
+                          ck::tuple_element_t<NDimSpatial - 1,
+                                              ck::Tuple<ck::tensor_layout::convolution::GNWK,
+                                                        ck::tensor_layout::convolution::GNHWK,
+                                                        ck::tensor_layout::convolution::GNDHWK>>>
+{
+};
+
+template <ck::index_t NDimSpatial>
+using InputLayout = typename CommonLayoutSettingSelector<NDimSpatial>::InputLayout;
 
-    using InLayout  = std::tuple_element_t<2, Tuple>;
-    using WeiLayout = std::tuple_element_t<1, Tuple>;
-    using OutLayout = std::tuple_element_t<0, Tuple>;
+template <ck::index_t NDimSpatial>
+using WeightLayout = typename CommonLayoutSettingSelector<NDimSpatial>::WeightLayout;
 
-    // clang-format off
-    using GroupedConvBwdWeightDeviceInstance = ck::tensor_operation::device::DeviceGroupedConvBwdWeight_Xdl_CShuffle
-        //##########|     Num| InLayout| WeiLayout| OutLayout| InData| WeiData| OutData| AccData|          In|         Wei|         Out|                      ConvBackward| Block|  MPer|  NPer| K0Per| K1| MPer| NPer| MXdl| NXdl|  ABlockTransfer|   ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|  BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|    CShuffle|    CShuffle|   CBlockTransfer|  CBlockTransfer|
-        //##########|     Dim|         |          |          |   Type|    Type|    Type|    Type| Elementwise| Elementwise| Elementwise|                            Weight|  Size| Block| Block| Block|   |  XDL|  XDL|  Per|  Per|   ThreadCluster|    ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|   ThreadCluster|  ThreadCluster| SrcAccessOrder|  SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave|   ClusterLengths| ScalarPerVector|
-        //##########| Spatial|         |          |          |       |        |        |        |   Operation|   Operation|   Operation|                    Specialization|      |      |      |      |   |     |     | Wave| Wave| Lengths_K0_M_K1|     ArrangeOrder|               |               |      PerVector|   PerVector_K1|          | Lengths_K0_N_K1|   ArrangeOrder|               |              |      PerVector|   PerVector_K1|          |  PerShuffle|  PerShuffle| MBlock_MPerBlock|    NWaveNPerXdl|
-        //##########|        |         |          |          |       |        |        |        |            |            |            |                                  |      |      |      |      |   |     |     |     |     |                |                 |               |               |               |               |          |                |               |               |              |               |               |          |            |            | NBlock_NPerBlock|                |
-        <         NDimSpatial,  InLayout, WeiLayout,OutLayout,    F16,     F16,     F16,     F32, PassThrough, PassThrough, PassThrough,                  ConvSpec,   128,    32,   128,     4,  8,   32,   32,    1,    2,  S<1, 4, 4,  8>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,              2,              8,              1,      true,  S<1, 4, 16, 2>,  S<0, 3, 1, 2>,  S<0, 2, 1, 3>,             2,              8,              4,      true,           1,           1,   S<1, 32, 1, 4>,               8>;
-    // clang-format on
+template <ck::index_t NDimSpatial>
+using OutputLayout = typename CommonLayoutSettingSelector<NDimSpatial>::OutputLayout;
 
+class TestGroupedConvndBwdWeight : public ::testing::Test
+{
+    protected:
     ck::utils::conv::ConvParam conv_param;
-    ck::index_t split_k{2};
 
     template <ck::index_t NDimSpatial>
-    bool Run()
+    void RunReference(Tensor<InDataType>& in,
+                      Tensor<WeiDataType>& wei_host_result,
+                      Tensor<OutDataType>& out)
+    {
+        auto ref_conv     = ck::tensor_operation::host::ReferenceConvBwdWeight<NDimSpatial,
+                                                                           InDataType,
+                                                                           WeiDataType,
+                                                                           OutDataType,
+                                                                           PassThrough,
+                                                                           PassThrough,
+                                                                           PassThrough>{};
+        auto ref_invoker  = ref_conv.MakeInvoker();
+        auto ref_argument = ref_conv.MakeArgument(in,
+                                                  wei_host_result,
+                                                  out,
+                                                  conv_param.conv_filter_strides_,
+                                                  conv_param.conv_filter_dilations_,
+                                                  conv_param.input_left_pads_,
+                                                  conv_param.input_right_pads_,
+                                                  PassThrough{},
+                                                  PassThrough{},
+                                                  PassThrough{},
+                                                  {},
+                                                  {},
+                                                  {});
+
+        ref_invoker.Run(ref_argument);
+    }
+
+    template <ck::index_t NDimSpatial>
+    bool PerformConvWeight(ck::index_t split_k)
     {
+        bool passed{true};
 
         const auto in_g_n_c_wis_desc =
-            ck::utils::conv::make_input_host_tensor_descriptor_g_n_c_wis_packed<InLayout>(
-                conv_param);
+            ck::utils::conv::make_input_host_tensor_descriptor_g_n_c_wis_packed<
+                InputLayout<NDimSpatial>>(conv_param);
 
         const auto wei_g_k_c_xs_desc =
-            ck::utils::conv::make_weight_host_tensor_descriptor_g_k_c_xs_packed<WeiLayout>(
-                conv_param);
+            ck::utils::conv::make_weight_host_tensor_descriptor_g_k_c_xs_packed<
+                WeightLayout<NDimSpatial>>(conv_param);
 
         const auto out_g_n_k_wos_desc =
-            ck::utils::conv::make_output_host_tensor_descriptor_g_n_k_wos_packed<OutLayout>(
-                conv_param);
+            ck::utils::conv::make_output_host_tensor_descriptor_g_n_k_wos_packed<
+                OutputLayout<NDimSpatial>>(conv_param);
+
+        Tensor<InDataType> in(in_g_n_c_wis_desc);
+        Tensor<WeiDataType> wei_host_result(wei_g_k_c_xs_desc);
+        Tensor<WeiDataType> wei_device_result(wei_g_k_c_xs_desc);
+        Tensor<OutDataType> out(out_g_n_k_wos_desc);
+
+        in.GenerateTensorValue(GeneratorTensor_2<InDataType>{-5, 5});
+        out.GenerateTensorValue(GeneratorTensor_2<OutDataType>{-5, 5});
+
+        DeviceMem in_device_buf(sizeof(InDataType) * in.mDesc.GetElementSpaceSize());
+        DeviceMem wei_device_buf(sizeof(WeiDataType) *
+                                 wei_device_result.mDesc.GetElementSpaceSize());
+        DeviceMem out_device_buf(sizeof(OutDataType) * out.mDesc.GetElementSpaceSize());
+
+        in_device_buf.ToDevice(in.mData.data());
+        out_device_buf.ToDevice(out.mData.data());
+
+        // init to 0
+        wei_device_buf.SetZero();
 
         std::array<ck::index_t, NDimSpatial + 3> input_lengths{};
-        std::array<ck::index_t, NDimSpatial + 3> filter_lengths{};
-        std::array<ck::index_t, NDimSpatial + 3> output_lengths{};
         std::array<ck::index_t, NDimSpatial + 3> input_strides{};
+        std::array<ck::index_t, NDimSpatial + 3> filter_lengths{};
         std::array<ck::index_t, NDimSpatial + 3> weights_strides{};
+        std::array<ck::index_t, NDimSpatial + 3> output_lengths{};
         std::array<ck::index_t, NDimSpatial + 3> output_strides{};
         std::array<ck::index_t, NDimSpatial> conv_filter_strides{};
         std::array<ck::index_t, NDimSpatial> conv_filter_dilations{};
@@ -94,86 +168,261 @@ class TestGroupedConvndBwdWeight : public ::testing::Test
         range_copy(conv_param.input_left_pads_, begin(input_left_pads));
         range_copy(conv_param.input_right_pads_, begin(input_right_pads));
 
-        auto conv = GroupedConvBwdWeightDeviceInstance{};
-
-        auto argument = conv.MakeArgument(nullptr,
-                                          nullptr,
-                                          nullptr,
-                                          input_lengths,
-                                          input_strides,
-                                          filter_lengths,
-                                          weights_strides,
-                                          output_lengths,
-                                          output_strides,
-                                          conv_filter_strides,
-                                          conv_filter_dilations,
-                                          input_left_pads,
-                                          input_right_pads,
-                                          PassThrough{},
-                                          PassThrough{},
-                                          PassThrough{},
-                                          split_k);
-        return conv.IsSupportedArgument(argument);
+        RunReference<NDimSpatial>(in, wei_host_result, out);
+
+        using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvBwdWeight_Xdl_CShuffle<
+            NDimSpatial,
+            ck::tuple_element_t<NDimSpatial - 1,
+                                ck::Tuple<ck::tensor_layout::convolution::GNWC,
+                                          ck::tensor_layout::convolution::GNHWC,
+                                          ck::tensor_layout::convolution::GNDHWC>>,
+            ck::tuple_element_t<NDimSpatial - 1,
+                                ck::Tuple<ck::tensor_layout::convolution::GKXC,
+                                          ck::tensor_layout::convolution::GKYXC,
+                                          ck::tensor_layout::convolution::GKZYXC>>,
+            ck::tuple_element_t<NDimSpatial - 1,
+                                ck::Tuple<ck::tensor_layout::convolution::GNWK,
+                                          ck::tensor_layout::convolution::GNHWK,
+                                          ck::tensor_layout::convolution::GNDHWK>>,
+            InDataType,           // InDataType
+            WeiDataType,          // WeiDataType
+            OutDataType,          // OutDataType
+            AccDataType,          // AccDataType
+            PassThrough,          // InElementwiseOperation
+            PassThrough,          // WeiElementwiseOperation
+            PassThrough,          // OutElementwiseOperation
+            ConvBwdWeightDefault, // ConvolutionBackwardWeightSpecialization
+            64,                   // BlockSize
+            16,                   // MPerBlock
+            16,                   // NPerBlock
+            32,                   // K0PerBlock
+            8,                    // K1
+            16,                   // MPerXdl
+            16,                   // NPerXdl
+            1,                    // MXdlPerWave
+            1,                    // NXdlPerWave
+            S<4, 16, 1>,          // ABlockTransferThreadClusterLengths_K0_M_K1
+            S<2, 0, 1>,           // ABlockTransferThreadClusterArrangeOrder
+            S<1, 0, 2>,           // ABlockTransferSrcAccessOrder
+            1,                    // ABlockTransferSrcVectorDim
+            1,                    // ABlockTransferSrcScalarPerVector
+            4,                    // ABlockTransferDstScalarPerVector_K1
+            false,                // ABlockLdsAddExtraM
+            S<4, 16, 1>,          // BBlockTransferThreadClusterLengths_K0_N_K1
+            S<2, 0, 1>,           // BBlockTransferThreadClusterArrangeOrder
+            S<1, 0, 2>,           // BBlockTransferSrcAccessOrder
+            1,                    // BBlockTransferSrcVectorDim
+            1,                    // BBlockTransferSrcScalarPerVector
+            4,                    // BBlockTransferDstScalarPerVector_K1
+            false,                // BBlockLdsAddExtraN
+            1,                    // CShuffleMXdlPerWavePerShuffle
+            1,                    // CShuffleNXdlPerWavePerShuffle
+            S<1, 8, 1, 8>,        // CBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
+            1>;                   // CBlockTransferScalarPerVector_NWaveNPerXdl
+
+        auto conv_ptr = DeviceOp{};
+        auto argument =
+            conv_ptr.MakeArgument(static_cast<InDataType*>(in_device_buf.GetDeviceBuffer()),
+                                  static_cast<WeiDataType*>(wei_device_buf.GetDeviceBuffer()),
+                                  static_cast<OutDataType*>(out_device_buf.GetDeviceBuffer()),
+                                  input_lengths,
+                                  input_strides,
+                                  filter_lengths,
+                                  weights_strides,
+                                  output_lengths,
+                                  output_strides,
+                                  conv_filter_strides,
+                                  conv_filter_dilations,
+                                  input_left_pads,
+                                  input_right_pads,
+                                  PassThrough{},
+                                  PassThrough{},
+                                  PassThrough{},
+                                  split_k);
+
+        auto invoker_ptr = conv_ptr.MakeInvoker();
+
+        if(conv_ptr.IsSupportedArgument(argument))
+        {
+            float avg_time = invoker_ptr.Run(argument, StreamConfig{nullptr, false});
+            wei_device_buf.FromDevice(wei_device_result.mData.data());
+            passed &= ck::utils::check_err(
+                wei_device_result.mData, wei_host_result.mData, "Error: incorrect results!");
+
+            std::size_t flop = conv_param.GetFlops() +
+                               3 * conv_param.GetOutputByte<WeiDataType>() / sizeof(WeiDataType);
+            std::size_t num_bytes = conv_param.GetByte<InDataType, WeiDataType, OutDataType>() +
+                                    conv_param.GetOutputByte<WeiDataType>();
+
+            float tflops     = static_cast<float>(flop) / 1.E9 / avg_time;
+            float gb_per_sec = num_bytes / 1.E6 / avg_time;
+
+            std::cout << "Perf: " << std::setw(10) << avg_time << " ms, " << tflops << " TFlops, "
+                      << gb_per_sec << " GB/s, "
+                      << "split_k " << split_k << std::endl;
+        }
+        return passed;
+    }
+
+    template <ck::index_t NDimSpatial>
+    void Run()
+    {
+        bool pass = true;
+
+        for(auto split_k : {1, 2})
+        {
+            pass = pass && PerformConvWeight<NDimSpatial>(split_k);
+            EXPECT_TRUE(pass);
+        }
     }
 };
 
-using GNHWC = ck::tensor_layout::convolution::GNHWC;
-using NHWGC = ck::tensor_layout::convolution::NHWGC;
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_Filter_1x1)
+{
+    this->conv_param = {
+        1, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<1>();
+}
 
-using GKYXC = ck::tensor_layout::convolution::GKYXC;
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_Filter_3x3)
+{
+    this->conv_param = {
+        1, 2, 4, 192, 192, {3, 3, 3}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<1>();
+}
 
-using GNHWK = ck::tensor_layout::convolution::GNHWK;
-using NHWGK = ck::tensor_layout::convolution::NHWGK;
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_Filter_1x1)
+{
+    this->conv_param = {
+        2, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<2>();
+}
 
-using KernelTypes =
-    ::testing::Types<std::tuple<GNHWK, GKYXC, GNHWC>, std::tuple<NHWGK, GKYXC, NHWGC>>;
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_Filter_3x3)
+{
+    this->conv_param = {
+        2, 2, 4, 192, 192, {3, 3, 3}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<2>();
+}
 
-template <typename Tuple>
-class TestGroupedConvndBwdWeightDefault
-    : public TestGroupedConvndBwdWeight<Tuple, ConvBwdWeightDefault>
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_3_Filter_1x1)
 {
-};
+    this->conv_param = {
+        3, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<3>();
+}
 
-template <typename Tuple>
-class TestGroupedConvndBwdWeightFilter1x1
-    : public TestGroupedConvndBwdWeight<Tuple, Filter1x1Stride1Pad0>
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_3_Filter_3x3)
 {
-};
+    this->conv_param = {
+        3, 2, 4, 192, 192, {3, 3, 3}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<3>();
+}
 
-TYPED_TEST_SUITE(TestGroupedConvndBwdWeightDefault, KernelTypes);
-TYPED_TEST_SUITE(TestGroupedConvndBwdWeightFilter1x1, KernelTypes);
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_Stride_1x1)
+{
+    this->conv_param = {
+        1, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<1>();
+}
 
-TYPED_TEST(TestGroupedConvndBwdWeightFilter1x1, SpecializationCheck)
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_Stride_2x2)
 {
-    // Check filter 3,3 instead of 1,1
-    this->conv_param  = {2, 2, 4, 192, 192, {3, 3}, {28, 28}, {1, 1}, {1, 1}, {0, 0}, {0, 0}};
-    bool is_supported = this->template Run<2>();
-    EXPECT_FALSE(is_supported);
+    this->conv_param = {
+        1, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<1>();
+}
 
-    // Check strides 2,2 instead of 1,1
-    this->conv_param = {2, 2, 4, 192, 192, {1, 1}, {28, 28}, {2, 2}, {1, 1}, {0, 0}, {0, 0}};
-    is_supported     = this->template Run<2>();
-    EXPECT_FALSE(is_supported);
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_Stride_1x1)
+{
+    this->conv_param = {
+        2, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<2>();
+}
 
-    // Check with pad
-    this->conv_param = {2, 2, 4, 192, 192, {1, 1}, {28, 28}, {1, 1}, {1, 1}, {1, 1}, {1, 1}};
-    is_supported     = this->template Run<2>();
-    EXPECT_FALSE(is_supported);
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_Stride_2x2)
+{
+    this->conv_param = {
+        2, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<2>();
+}
 
-    // Supported version
-    this->conv_param = {2, 2, 128, 128, 256, {1, 1}, {3, 3}, {1, 1}, {1, 1}, {0, 0}, {0, 0}};
-    is_supported     = this->template Run<2>();
-    EXPECT_TRUE(is_supported);
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_3_Stride_1x1)
+{
+    this->conv_param = {
+        3, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<3>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_3_Stride_2x2)
+{
+    this->conv_param = {
+        3, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}};
+    this->template Run<3>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_WithPadding)
+{
+    this->conv_param = {
+        1, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}};
+    this->template Run<1>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_WithPadding)
+{
+    this->conv_param = {
+        2, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}};
+    this->template Run<2>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_3_WithPadding)
+{
+    this->conv_param = {
+        3, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}};
+    this->template Run<3>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_SupportedVersion)
+{
+    this->conv_param = {
+        1, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}};
+    this->template Run<1>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_SupportedVersion)
+{
+    this->conv_param = {
+        2, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}};
+    this->template Run<2>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_3_SupportedVersion)
+{
+    this->conv_param = {
+        3, 2, 4, 192, 192, {1, 1, 1}, {28, 28, 28}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}};
+    this->template Run<3>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_VectorLoadForA)
+{
+    this->conv_param = {1, 2, 128, 129, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}};
+    this->template Run<1>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_VectorLoadForA)
+{
+    this->conv_param = {2, 2, 128, 129, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}};
+    this->template Run<2>();
+}
+
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_1_VectorLoadForB_E_DS)
+{
+    this->conv_param = {1, 2, 128, 128, 257, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}};
+    this->template Run<1>();
 }
 
-TYPED_TEST(TestGroupedConvndBwdWeightDefault, VectorLoadCheck)
+TEST_F(TestGroupedConvndBwdWeight, TestGroupedConvndBwdWeight_NDimSpatial_2_VectorLoadForB_E_DS)
 {
-    // vector load for A
-    this->conv_param  = {2, 2, 128, 129, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}};
-    bool is_supported = this->template Run<2>();
-    EXPECT_FALSE(is_supported);
-    // vector load for B, E, Ds
     this->conv_param = {2, 2, 128, 128, 257, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}};
-    is_supported     = this->template Run<2>();
-    EXPECT_FALSE(is_supported);
+    this->template Run<2>();
 }