[Mlir-commits] [mlir] [mlir][GPU] Extend gpu.barrier with scope and named-barrier support (PR #195692)

[Guray Ozen]

================
@@ -370,6 +370,37 @@ struct AssertOpToAssertfailLowering
 /// Import the GPU Ops to NVVM Patterns.
 #include "GPUToNVVM.cpp.inc"
 
+struct GPUBarrierOpNVVMLowering final
+    : public ConvertOpToLLVMPattern<gpu::BarrierOp> {
+  using ConvertOpToLLVMPattern::ConvertOpToLLVMPattern;
+
+  LogicalResult
+  matchAndRewrite(gpu::BarrierOp op, gpu::BarrierOp::Adaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    if (op.getNamedBarrier())
+      return rewriter.notifyMatchFailure(
+          op, "named barriers are not supported on NVVM");
----------------
grypp wrote:

The named barriers are supported in the NVVM dialect. We have the following op. There are 16 named barrier IDs, and you can synchronize a specified number of threads:

```
nvvm.barrier` (`id` `=` $barrierId^)? (`number_of_threads` `=` $numberOfThreads^)?
```
more info:
https://mlir.llvm.org/docs/Dialects/NVVMDialect/#nvvmbarrier-nvvmbarrierop


https://github.com/llvm/llvm-project/pull/195692