[mlir][AMDGPU] Implement gpu.subgroup_reduce with DPP intrinsics on AMD GPUs#133204

Merged

pashu123 merged 28 commits intollvm:mainfrom

Muzammiluddin-Syed-ECE:muzasyed/sub

Apr 24, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[mlir][AMDGPU] Implement gpu.subgroup_reduce with DPP intrinsics on AMD GPUs#133204

[mlir][AMDGPU] Implement gpu.subgroup_reduce with DPP intrinsics on AMD GPUs#133204
pashu123 merged 28 commits intollvm:mainfrom
Muzammiluddin-Syed-ECE:muzasyed/sub

Commits on Apr 16, 2025

Creates AMDToGPUPass to house a subgroup reduce lowering pattern to DPP

Fix for numerical issues in MatVec tests

Rewrites pattern to be closer to device lib impl.

Removes AMDToGPUPass, moving pattern into existing pass

Adding permlanex16 and other dpp related ops to mlir dialect

Fixing permlanex16 intrinsic failure

simplify verbose typing

testing numerics

fixing

fixing

fixing

trying again

Fixing implementation

Adding DPP test

Addressing PR comments

removing unnecessary header

Addressing PR comments

moving permlanex16 changes to another commit

fixing test

fixing code formatting

Updating implementation to support gfx 10+

Small formatting change

Removing ReadlaneOps from test

Improve dpp implementation

Commits on Apr 17, 2025

fixing formatting

Commits on Apr 22, 2025

Fixing implementation of DPP subgroup reduce

Commits on Apr 23, 2025

Addressing PR comments

Fixing Typo in RUN command

[mlir][AMDGPU] Implement gpu.subgroup_reduce with DPP intrinsics on AMD GPUs#133204pashu123 merged 28 commits intollvm:mainllvm/llvm-project:mainfrom Muzammiluddin-Syed-ECE:muzasyed/subMuzammiluddin-Syed-ECE/llvm-project:muzasyed/subCopy head branch name to clipboard

Commits

Commits on Apr 16, 2025

Commits on Apr 17, 2025

Commits on Apr 22, 2025

Commits on Apr 23, 2025

[mlir][AMDGPU] Implement gpu.subgroup_reduce with DPP intrinsics on AMD GPUs#133204
pashu123 merged 28 commits intollvm:mainfrom
Muzammiluddin-Syed-ECE:muzasyed/sub