Jacob Lambert
Jacob Lambert
Currently Comgr includes mutex locks around compilation actions to avoid unexpected LLVM multi-threading side effects. https://github.com/RadeonOpenCompute/ROCm-CompilerSupport/blob/amd-stg-open/lib/comgr/src/comgr.cpp#L1252 It would be nice to fix the issues with LLVM preventing multiple threads from...
We should performance test both compilation time and generated binary runtime with -use-spirv-backend and -no-use-spirv-backend to verify there are no critical regressions with the new backend. We could use the...
- atomic vectors - visibility - float controls 2 (to support fine grained fast-math flags and contract): disabled by https://github.com/llvm/llvm-project/pull/169659 . @jmmartinez is implementing it in the translator.
https://github.com/ROCm/llvm-project/blob/75216c09375a8b262ecc3d7efc31abddf684b535/amd/comgr/src/comgr-compiler.cpp#L1021
Once optional support for the SPIRV backend is added for `--offload-arch=amdgcnspirv`, we need to identify the delta between that and the existing default SPIRV-translator-based implementation. Likely differences may include: -...
To test the new SPIRV-backend-based implementation (and AMDGPU + SPIRV in general), we should create a downstream buildbot that tests the following: - LLVM build and LIT spirv-related tests -...
Currently, we don't have runtime support for SPIRV on AMDGPUs for OpenCL (only HIP). We should extend the AMDGPU OpenCL Runtime to also support SPIRV, which should be a more...