[Mlir-commits] [mlir] [mlir][amdgpu] implement amdgpu.sparse_mfma wrapper for smfmac instructions (PR #171968)

[Jakub Kuderski]

================
@@ -1291,6 +1413,49 @@ struct ScaledMFMAOpLowering : public ConvertOpToLLVMPattern<ScaledMFMAOp> {
   }
 };
 
+struct SparseMFMAOpLowering : public ConvertOpToLLVMPattern<SparseMFMAOp> {
+  SparseMFMAOpLowering(const LLVMTypeConverter &converter, Chipset chipset)
+      : ConvertOpToLLVMPattern<SparseMFMAOp>(converter), chipset(chipset) {}
+
+  Chipset chipset;
+
+  LogicalResult
+  matchAndRewrite(SparseMFMAOp op, SparseMFMAOpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    Location loc = op.getLoc();
+    auto outType =
+        typeConverter->convertType<VectorType>(op.getDestC().getType());
+    if (!outType)
+      return rewriter.notifyMatchFailure(op, "type conversion failed");
+
+    // smfmac is supported on gfx942 and gfx950
----------------
kuhar wrote:

```suggestion
    // smfmac is supported on gfx942 and gfx950.
```

https://github.com/llvm/llvm-project/pull/171968