[Mlir-commits] [mlir] [mlir][spirv] Simplify gpu reduction to spirv logic (PR #73546)

[Stanley Winata]

================
@@ -647,8 +647,108 @@ gpu.module @kernels {
     // CHECK: %{{.*}} = spirv.GroupNonUniformUMax "Subgroup" "Reduce" %[[ARG]] : i32
     %r0 = gpu.subgroup_reduce maxsi %arg : (i32) -> (i32)
     %r1 = gpu.subgroup_reduce maxui %arg : (i32) -> (i32)
+    %r3 = gpu.subgroup_reduce maxui %arg : (i32) -> (i32)
----------------
raikonenfnu wrote:

Not sure what this new line adds to.

https://github.com/llvm/llvm-project/pull/73546