[MLIR] [AMDGPU] refactor how we handle target features #129206

Open

Open

[MLIR] [AMDGPU] refactor how we handle target features#129206

Labels

Currently, we hardcode hardware features in the code (see

llvm-project/mlir/lib/Dialect/AMDGPU/Transforms/EmulateAtomics.cpp

Line 173 in 15c49b9

if (chipset.majorVersion == 11) {

). We want to implement a better way of handling subtarget features. See this PR for discussion about possible implementations: #129029

Metadata

Assignees

No one assigned

Labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests