[Mlir-commits] [mlir] [mlir][xegpu] Raise the reduction vector size to match maxReduceVectorSize (PR #225122)

[Jianhui Li]

================
@@ -2436,7 +2540,7 @@ xegpu::SliceAttr xegpu::setupMultiReductionResultLayout(
   auto context = srcVecTy.getContext();
 
   const int subgroupSize = uArch->getSubgroupSize();
-  int64_t maxReduceVectorSize = 1; // could extend to spirv vector Size
+  int64_t maxReduceVectorSize = 16; // extend to spirv vector Size
----------------
Jianhui-Li wrote:

Right, I see your point and had a discussion with @silee2.  It is not clear at this point which layer is best responsible to mapping this "virtual vector size" for reduction to the HW size. So I would keep it as is for now and defer the decision to future HW targets. @Garra1980

https://github.com/llvm/llvm-project/pull/225122