[Mlir-commits] [mlir] [MLIR][XeGPU] Add lowering from transfer_read/transfer_write to load_gather/store_scatter (PR #152429)

[Adam Siemieniuk]

================
@@ -404,3 +406,18 @@ void xegpu::doSCFStructuralTypeConversionWithTensorType(
     (void)mlir::applyPartialConversion(op, target, std::move(patterns));
   }
 }
+
+std::optional<std::string> xegpu::getXeGPUChipStr(Operation *op) {
+  auto gpuModuleOp = op->getParentOfType<mlir::gpu::GPUModuleOp>();
+  if (gpuModuleOp) {
+    auto targetAttrs = gpuModuleOp.getTargets();
+    if (targetAttrs) {
----------------
adam-smnk wrote:

nit: I'd invert these check to be a guard clause: `if (!gpuModuleOp) return std::nullopt;` etc.

https://github.com/llvm/llvm-project/pull/152429