[Mlir-commits] [mlir] 59f0e46 - [mlir][Inliner] Don't optimize callees in async mode if there is only one to optimize

[River Riddle]

Author: River Riddle
Date: 2021-02-23T18:44:09-08:00
New Revision: 59f0e4627a5e8789c9ab997cd13f96600848fea8

URL: https://github.com/llvm/llvm-project/commit/59f0e4627a5e8789c9ab997cd13f96600848fea8
DIFF: https://github.com/llvm/llvm-project/commit/59f0e4627a5e8789c9ab997cd13f96600848fea8.diff

LOG: [mlir][Inliner] Don't optimize callees in async mode if there is only one to optimize

This avoids unnecessary async overhead in situations that won't benefit from it.

Added: 
    

Modified: 
    mlir/lib/Transforms/Inliner.cpp

Removed: 
    


################################################################################
diff  --git a/mlir/lib/Transforms/Inliner.cpp b/mlir/lib/Transforms/Inliner.cpp
index 0a2cc7eb4c98..bbcb1b2e0bc6 100644
--- a/mlir/lib/Transforms/Inliner.cpp
+++ b/mlir/lib/Transforms/Inliner.cpp
@@ -665,7 +665,7 @@ LogicalResult InlinerPass::optimizeSCC(CallGraph &cg, CGUseList &useList,
   // NOTE: This is simple now, because we don't enable optimizing nodes within
   // children. When we remove this restriction, this logic will need to be
   // reworked.
-  if (context->isMultithreadingEnabled()) {
+  if (context->isMultithreadingEnabled() && nodesToVisit.size() > 1) {
     if (failed(optimizeSCCAsync(nodesToVisit, context)))
       return failure();