Closed nhat-nguyen closed 3 weeks ago
Thanks for working on this! I tried this branch with latest triton today. The only issue is:
.../triton/third_party/triton_shared/tools/triton-shared-opt/../RegisterTritonSharedDialects.h:39:3: error: no member named 'registerTritonGPUPasses' in namespace 'mlir'; did you mean 'mlir::triton::gpu::registerTritonGPUPasses'?
39 | mlir::registerTritonGPUPasses();
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| mlir::triton::gpu::registerTritonGPUPasses
.../triton/python/build/cmake.linux-x86_64-cpython-3.10/include/triton/Dialect/TritonGPU/Transforms/Passes.h.inc:1120:13: note: 'mlir::triton::gpu::registerTritonGPUPasses' declared here
1120 | inline void registerTritonGPUPasses() {
| ^
1 error generated.
The suggested fix works for me.
@zhaoshiz Thanks. The branch has been 3-week old, I will update this again with the latest triton and apply your suggested fix.
.dyn_cast
,.cast
,.isa
are all deprecatedopenai/triton
totriton-lang/triton
tensor.expand_shape
Fixes #134