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

[Adam Siemieniuk]

================
@@ -123,6 +123,8 @@ Value createVectorWithShapeFromValues(OpBuilder &builder, Location loc,
 void doSCFStructuralTypeConversionWithTensorType(Operation *op,
                                                  TypeConverter converter);
 
+std::optional<std::string> getXeGPUChipStr(Operation *op);
----------------
adam-smnk wrote:

Shouldn't this live in XeVM?
Also, the dialect name can be removed from the API itself as it's covered by the namespace `xegpu::getChipStr`

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