[ROCm] Some fixes of ROCm codegen#16404
Merged
junrushao merged 1 commit intoapache:mainfrom Jan 16, 2024
spectrometerHBH:rocm-fix
Merged
[ROCm] Some fixes of ROCm codegen#16404junrushao merged 1 commit intoapache:mainfrom spectrometerHBH:rocm-fix
junrushao merged 1 commit intoapache:mainfrom
spectrometerHBH:rocm-fix
Conversation
tqchen
approved these changes
Jan 15, 2024
junrushao
approved these changes
Jan 15, 2024
- Handle tvm_thread_invariant as no op. - `llvm.amdgcn.ds.bpermute` requires i32 as its input, but it can handle all 32 bit types - ocml intrinsics lead to incorrect codegen when used with vectorization, remove it and use llvm intrinsics instead
junrushao
pushed a commit
to junrushao/tvm
that referenced
this pull request
Jan 17, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (mlc-ai/relax#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (mlc-ai/relax#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (mlc-ai/relax#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (apache#16404, apache#16405)
junrushao
added a commit
to junrushao/tvm
that referenced
this pull request
Jan 17, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (mlc-ai/relax#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (mlc-ai/relax#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (mlc-ai/relax#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (apache#16404, apache#16405) Co-authored-by: Bohan Hou <bohanhou@andrew.cmu.edu> Co-authored-by: Lesheng Jin <leshenj15@gmail.com>
junrushao
added a commit
to junrushao/tvm
that referenced
this pull request
Jan 17, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (mlc-ai/relax#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (mlc-ai/relax#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (mlc-ai/relax#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (apache#16404, apache#16405) Co-authored-by: Bohan Hou <bohanhou@andrew.cmu.edu> Co-authored-by: Lesheng Jin <leshenj15@gmail.com>
junrushao
added a commit
to mlc-ai/relax
that referenced
this pull request
Jan 17, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (apache/tvm#16404, apache/tvm#16405) Co-authored-by: Bohan Hou <bohanhou@andrew.cmu.edu> Co-authored-by: Lesheng Jin <leshenj15@gmail.com>
junrushao
added a commit
to mlc-ai/relax
that referenced
this pull request
Jan 17, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (apache/tvm#16404, apache/tvm#16405) Co-authored-by: Bohan Hou <bohanhou@andrew.cmu.edu> Co-authored-by: Lesheng Jin <leshenj15@gmail.com>
tqchen
pushed a commit
that referenced
this pull request
Jan 17, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (mlc-ai/relax#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (mlc-ai/relax#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (mlc-ai/relax#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (#16404, #16405) Co-authored-by: Bohan Hou <bohanhou@andrew.cmu.edu> Co-authored-by: Lesheng Jin <leshenj15@gmail.com>
junrushao
added a commit
to mlc-ai/relax
that referenced
this pull request
Jan 21, 2024
This PR upstreams a few commits that recovers the unity branch from broken wheel packages. It includes the following changes: - Fix MSVC build in `pipe.h` where `DWORD` is not cast to proper return type (#306); - Fix MSVC build warnings on not recognizing "#pragma GCC" (#307); - Fix NVCC build warnings where it fails to infer if "[[noreturn]]" actually does not return (#308); - Fix ROCM/Vulkan backend which fails compilation for operators like group GEMM, paged attention, etc. (apache/tvm#16404, apache/tvm#16405) Co-authored-by: Bohan Hou <bohanhou@andrew.cmu.edu> Co-authored-by: Lesheng Jin <leshenj15@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
llvm.amdgcn.ds.bpermuterequires i32 as its input, but it can handle all 32 bit types