IMEX v0.1

Preview release of Intel® Extension for MLIR (IMEX)

This release focuses on functionally enabling Deep Learning workloads with static shaped input tensors written in upstream MLIR Entry dialects (Linalg and other lower level dialects) on select Intel GPUs. The development is test case driven and does not cover all ops in the MLIR entry dialects.

Supported configurations

• Ubuntu 20.04 LTS
• x86 CPU
• Intel® Data Center GPU Flex Series or Intel® Data Center GPU Max Series

Supported data types for GPU

• Fp32

Dependencies for GPU execution

GPU execution supports two different wrapper libraries for interacting with GPU. Level Zero and Sycl wrapper libraries.

• oneAPI Level Zero: https://github.com/oneapi-src/level-zero (Required for both Level Zero and Sycl wrapper)
• Intel® oneAPI DPC++/C++ Compiler: https://www.intel.com/content/www/us/en/developer/articles/tool/oneapi-standalone-components.html#dpcpp-cpp (Required for Sycl wrapper)

Tested DL workloads

• Resnet-50 inference
• Mobilenet-v3 inference
• Mnist mlp

Tested DL Operations

• Elementwise
• Convolution
• GEMM
• Reduction
• Data movement
• Padding