[Triton-MLIR][OPTIMIZER] Cleaned up swizzling (#869)

Swizzling is no longer implemented as a separate pass. It is instead
done in a specialized constructor of SharedEncodingAttr, and tested via
google tests instead of triton-opt + filecheck.

In the future we may want to implement it as a pass again once we have
an additional dialect between TritonGPU and LLVM.
This commit is contained in:
Philippe Tillet
2022-11-10 12:05:46 -08:00
committed by GitHub
parent 2aa538ec2e
commit f40c63fb03
14 changed files with 170 additions and 318 deletions

View File

@@ -1264,10 +1264,6 @@ void init_triton_ir(py::module &&m) {
[](mlir::PassManager &self) {
self.addPass(mlir::createTritonGPUCombineOpsPass());
})
.def("add_triton_gpu_swizzle_pass",
[](mlir::PassManager &self) {
self.addPass(mlir::createTritonGPUSwizzlePass());
})
.def("add_triton_gpu_to_llvm",
[](mlir::PassManager &self) {
self.addPass(mlir::triton::createConvertTritonGPUToLLVMPass());