[Mlir-commits] [mlir] [MLIR][ROCDL] Add conversion for gpu.subgroup_id to ROCDL (PR #136405)

[Krzysztof Drewniak]

================
@@ -406,15 +504,11 @@ void mlir::populateGpuToROCDLConversionPatterns(
   patterns.add<GPUDynamicSharedMemoryOpLowering>(converter);
 
   patterns.add<GPUShuffleOpLowering, GPULaneIdOpToROCDL>(converter);
-
+  patterns.add<GPUSubgroupIdOpToROCDL>(converter, chipset, subgroupSize);
   populateMathToROCDLConversionPatterns(converter, patterns);
 }
 
 std::unique_ptr<OperationPass<gpu::GPUModuleOp>>
-mlir::createLowerGpuOpsToROCDLOpsPass(const std::string &chipset,
-                                      unsigned indexBitwidth,
-                                      bool useBarePtrCallConv,
-                                      gpu::amd::Runtime runtime) {
----------------
krzysz00 wrote:

Getting rid of this constructor seems unrelated?

https://github.com/llvm/llvm-project/pull/136405