[Mlir-commits] [mlir] [mlir][gpu] Add patterns to break down subgroup reduce (PR #76271)

[Kunwar Grover]

================
@@ -62,6 +63,12 @@ void populateGpuShufflePatterns(RewritePatternSet &patterns);
 /// Collect a set of patterns to rewrite all-reduce ops within the GPU dialect.
 void populateGpuAllReducePatterns(RewritePatternSet &patterns);
 
+/// Collect a set of patterns to break down subgroup_reduce ops into smaller
+/// ones supported by the target of size <= `maxShuffleBitwidth`.
----------------
Groverkss wrote:

Can you mention what "size" here is? It looks like it's the number of bits used by the vector?

https://github.com/llvm/llvm-project/pull/76271