You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Sep 21, 2019. It is now read-only.
Facing compilation errors from CUB while exercising Thrust API's on HIP/CUDA path. The branch currently under use is hip_port_1.7.3
Below is the list of issues and fix which works for Thrust API's, these need to be checked-in to the relevant cub-hip branch.
Issue 1: ../thrust/system/cuda/detail/cub/device/dispatch/dispatch_reduce.cuh(479): error: expected a ">>>"
../thrust/system/cuda/detail/cub/device/dispatch/dispatch_reduce.cuh(479): error: expected a "("
There exists a syntax error in the above mentioned file.
Solution: In file thrust/system/cuda/detail/cub-hip/cub/device/dispatch/dispatch_reduce.cuh ,line no:479 - hipLaunchKernelGGL(single_tile_kernel, 1, 256 0, stream, + hipLaunchKernelGGL(single_tile_kernel, 1, 256, 0, stream,
This resolves the issue.
Facing compilation errors from CUB while exercising Thrust API's on HIP/CUDA path. The branch currently under use is hip_port_1.7.3
Below is the list of issues and fix which works for Thrust API's, these need to be checked-in to the relevant cub-hip branch.
Issue 1: ../thrust/system/cuda/detail/cub/device/dispatch/dispatch_reduce.cuh(479): error: expected a ">>>"
../thrust/system/cuda/detail/cub/device/dispatch/dispatch_reduce.cuh(479): error: expected a "("
There exists a syntax error in the above mentioned file.
Solution: In file thrust/system/cuda/detail/cub-hip/cub/device/dispatch/dispatch_reduce.cuh ,line no:479
- hipLaunchKernelGGL(single_tile_kernel, 1, 256 0, stream,
+ hipLaunchKernelGGL(single_tile_kernel, 1, 256, 0, stream,
This resolves the issue.
Issue 2 : error: invalid conversion from ‘void ()(const int, int*, int, int, int, thrust::system::cuda::detail::cub_::GridEvenShare)’ to ‘const void*’ [-========= fpermissive]
return hipOccupancyMaxActiveBlocksPerMultiprocessor (
Solution: In file thrust/system/cuda/detail/cub/util_device.cuh, Line no:276
- kernel_ptr,
+ (const void *)kernel_ptr,
This resolves the issue.
Steps to reproduce:
$ git clone https://github.com/ROCmSoftwarePlatform/Thrust.git
$ cd Thrust
$ export HIP_PLATFORM=nvcc (For NVCC Platform )
$ cd examples
$ cp histogram.cu histogram.cpp
$ /opt/rocm/bin/hipcc histogram.cpp -I../ -o histogram.cpp.out
The text was updated successfully, but these errors were encountered: