[Mlir-commits] [mlir] [MLIR][XeGPU] Validate DPAS operand types against uArch in XeGPUToXeVM conversion (PR #185081)

[Arjun Bhamra]

================
@@ -0,0 +1,14 @@
+// RUN: mlir-opt --convert-xegpu-to-xevm %s -split-input-file -verify-diagnostics
+
+// Verify that xegpu.dpas with unsupported element types (i16) is rejected
+// during XeGPUToXeVM conversion rather than crashing.
+
+gpu.module @test_kernel [#xevm.target<chip = "pvc">] {
+  func.func @main() {
+    %0 = spirv.Constant dense<0> : vector<4xi16>
+    %1 = spirv.Constant dense<0> : vector<4xi32>
----------------
abhamra wrote:

This is fixed in the latest commits, thank you!

https://github.com/llvm/llvm-project/pull/185081