[mlir] Set top-down traversal for LinalgElementwiseOpFusion

The primary pattern for this pass clones many operations from producers
to consumers. Doing this top down prevents duplicated work when a
producer has multiple consumers, if it also is consuming another
linalg.generic.

As an example, a chain of ~2600 generics that are fused into ~70
generics was resulting in 16255 pattern invocations. This took 14
seconds on one machine but takes only 0.3 seconds with top-down
traversal.

Differential Revision: https://reviews.llvm.org/D107818

diff --git a/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp b/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
index fdca523..43a4105 100644 (file)
--- a/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
@@ -1294,7 +1294,12 @@ struct LinalgElementwiseOpFusionPass
         patterns,
         LinalgElementwiseFusionOptions().setControlFoldingReshapes(
             allowFoldingUnitDimReshapes ? allowFoldingFn : skipUnitDimReshape));
-    (void)applyPatternsAndFoldGreedily(op->getRegions(), std::move(patterns));
+
+    // Use TopDownTraversal for compile time reasons
+    GreedyRewriteConfig grc;
+    grc.useTopDownTraversal = true;
+    (void)applyPatternsAndFoldGreedily(op->getRegions(), std::move(patterns),
+                                       grc);
   }
 };