[Mlir-commits] [mlir] [mlir][gpu] Add 'cluster_size' attribute to gpu.subgroup_reduce (PR #104851)

[Jakub Kuderski]

================
@@ -1208,11 +1208,21 @@ def GPU_SubgroupReduceOp : GPU_Op<"subgroup_reduce", [SameOperandsAndResultType]
     ```mlir
     %1 = gpu.subgroup_reduce add %a : (f32) -> (f32)
     %2 = gpu.subgroup_reduce add %b : (vector<4xf16>) -> (vector<4xf16>)
+    %3 = gpu.subgroup_reduce add %c cluster_size(4) : (f32) -> (f32)
     ```
 
     If `uniform` flag is set either none or all lanes of a subgroup need to execute
-    this op in convergence. The reduction operation must be one
-    of:
+    this op in convergence.
+
+    If a `cluster_size` is not provided, the reduction covers all lanes of the
+    subgroup and the result is equal for all lanes.
+
+    If a `cluster_size` is provided, the subgroup is divided into clusters of
+    `cluster_size` contiguous lanes each, a reduction is done for all lanes of
----------------
kuhar wrote:

It would be nice to clarify that these clusters are contagious and adjacent (instead of, say, strided / interleaved).

https://github.com/llvm/llvm-project/pull/104851