hipamd

hipamd copied to clipboard

Inconsistent CMake module names

3

Documentation https://rocmdocs.amd.com/en/latest/Installation_Guide/Using-CMake-with-AMD-ROCm.html states to use `find_package(hip)` but HIP provides `FindHIP.cmake`, not `Findhip.cmake` as seen here: https://github.com/ROCm-Developer-Tools/HIP/blob/develop/cmake/FindHIP.cmake resulting in the following error: ```` [cmake] CMake Error at CMakeLists.txt:8 (find_package): [cmake] By...

pvelesko

There was a mismatch in capitalization between the definition and use by cmake of `ROCCLR_DIR`

ldrumm

merge release code into development

rocm-ci

This code leads to compilation failures with CUDA 12: https://github.com/ROCm-Developer-Tools/hipamd/blob/4209792929ddf54ba9530813b7879cfdee42df14/include/hip/nvidia_detail/nvidia_hip_runtime_api.h#L2800

garymm

Possible bug in occupancy calculation

1

I've been porting the algorithm in `ihipOccupancyMaxActiveBlocksPerMultiprocessor` to Julia to use for our own occupancy calculations, and noticed a potential typo in: https://github.com/ROCm-Developer-Tools/hipamd/blob/05e3016405bdaec0b8acfe8378df968dd02136d5/src/hip_platform.cpp#L351 I assume this should actually be `if...

jpsamaroo

Compile error in hip_cooperative_groups.h with ROCm 5.5

1

With ROCm 5.5, the inclusion of `hip/hip_cooperative_groups.h` results in a compile error on debug builds: ```c++ In file included from /opt/rocm-5.5.0/include/hip/hip_cooperative_groups.h:38: /opt/rocm-5.5.0/include/hip/amd_detail/amd_hip_cooperative_groups.h:455:7: error: invalid instruction, did you mean: s_trap? __hip_assert(false...

nolmoonen

Implement float/double atomicMin/Max in terms of integer atomics

12

Currently, HIP implements atomicMin/Max for single and double precision floating point values as CAS loops. However, in fast math scenarios, on architectures with hardware support for signed/unsigned integer atomicMin/Max a...

FreddieWitherden

There is a global `#define __local` in `device_library_decls.h`. This symbol is already used by some libcxx templates. Building (and depending on) libcxx with the (soon to be default) `_LIBCPP_REMOVE_TRANSITIVE_INCLUDES` macro...

aaronmondal

Fix operator<< after e3eb54c

1

aaronmondal

Fix inconsistent overrides

1

This caused `-Winconsistent-missing-override` to complain with Clang 17.

aaronmondal