[Mlir-commits] [mlir] [mlir][linalg] Implement Winograd Conv2D. (PR #94470)

[Oleksandr Alex Zinenko]

================
@@ -1312,6 +1312,20 @@ FailureOr<Operation *> transposeBatchMatmul(RewriterBase &rewriter,
                                             linalg::BatchMatmulOp op,
                                             bool transposeLHS = true);
 
+/// Convert linalg.conv_2d_nhwc_fhwc to Winograd Conv2D algorithm.
+FailureOr<Operation *> winogradConv2D(RewriterBase &rewriter,
+                                      linalg::Conv2DNhwcFhwcOp op, int64_t m,
+                                      int64_t r);
+FailureOr<Operation *>
+winogradConv2DRewriteFilterTransform(RewriterBase &rewriter,
----------------
ftynse wrote:

Please document all top-level entities.

https://github.com/llvm/llvm-project/pull/94470