[Mlir-commits] [mlir] [MLIR][XeGPU] Introduce `xegpu::uArch` usage in target-sensitive passes (PR #163801)

[Artem Kroviakov]

================
@@ -42,12 +40,61 @@ struct Xe2Plus : public uArch {
               &instrs = {})
       : uArch(archName, archDescription, regInfo, cacheInfo, instrs),
         xeCore(xeCore) {}
+  int getSubgroupSize() const override { return 16; }
+  unsigned getPackedFormatBitSizeGatherScatter() const override { return 32; }
+  unsigned getPackedFormatBitSize() const override { return 16; }
----------------
akroviakov wrote:

Yes, will be renamed. And I think it should be a member of dpas instruction per uarch instance. We might want to split this PR into two parts to have a substantial discussion in each: (1) uArch modification and (2) propagation option and uArch application in passes.

https://github.com/llvm/llvm-project/pull/163801