[SWP] attempt to remove a workaround for a bug

Summary: Triton LLVM codegen has a bug where local_loads from #shared to
than the mma tile. Remove the workaround.

See #3561.

Test Plan:

Reviewers:

Subscribers:

Tasks:

Tags:

lib/Dialect/TritonGPU/Transforms/Pipeliner/MatmulLoopPipeline.cpp

@@ -451,6 +451,7 @@ assignMemoryLayouts(llvm::SmallVector<std::tuple<Operation *, int, Operation *>>
         // If we can't agree on a shared encoding skip pipelinig the load.
         if (incompatible)
           continue;
+#if 0
         // HACK: Triton LLVM codegen has a bug where local_loads from #shared to
         // #mma layout can lead to invalid code if the loaded shape is smaller
         // than the mma tile (e.g. loading a 128x1 tensor for an MMAv2 dot with
@@ -481,6 +482,7 @@ assignMemoryLayouts(llvm::SmallVector<std::tuple<Operation *, int, Operation *>>
               continue;
           }
         }
+#endif
       }
     } else if (auto loadOp = dyn_cast<tt::LoadOp>(use)) {
       // The use of this loadOp is another loadOp. If the use is not in the

test/TritonGPU/loop-pipeline.mlir

@@ -1448,7 +1448,8 @@ module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 2 :
 // -----
 
 // COMMON-LABEL: @dont_pipeline_128x1
-// COMMON-NOT: local_load{{.*}}128x1
+// AMD-NOT: local_load{{.*}}128x1
+// CHECK: local_load{{.*}}128x1
 #blocked = #triton_gpu.blocked<{sizePerThread = [1, 1], threadsPerWarp = [32, 1], warpsPerCTA = [4, 1], order = [0, 1]}>
 #mma = #triton_gpu.nvidia_mma<{versionMajor = 2, versionMinor = 0, warpsPerCTA = [4, 1], instrShape = [16, 8]}>
 module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 : i32} {