[flang-commits] [flang] f239922 - [flang] Enable hlfir.sum inlining by default. (#119937)

[via flang-commits]

Author: Slava Zakharin
Date: 2024-12-16T07:59:15-08:00
New Revision: f239922cdc15c68266abb94b4b58ed46aa2572b4

URL: https://github.com/llvm/llvm-project/commit/f239922cdc15c68266abb94b4b58ed46aa2572b4
DIFF: https://github.com/llvm/llvm-project/commit/f239922cdc15c68266abb94b4b58ed46aa2572b4.diff

LOG: [flang] Enable hlfir.sum inlining by default. (#119937)

There is already a LIT test for hlfir.sum inlining that uses
the engineering option. I would like to keep the option
for short period of time to be able to revert
in case of performance regressions that I was not able to see.

Added: 
    

Modified: 
    flang/lib/Optimizer/HLFIR/Transforms/SimplifyHLFIRIntrinsics.cpp

Removed: 
    


################################################################################
diff  --git a/flang/lib/Optimizer/HLFIR/Transforms/SimplifyHLFIRIntrinsics.cpp b/flang/lib/Optimizer/HLFIR/Transforms/SimplifyHLFIRIntrinsics.cpp
index 3e9d956b6e56dd..a90c5a9ef5a4e5 100644
--- a/flang/lib/Optimizer/HLFIR/Transforms/SimplifyHLFIRIntrinsics.cpp
+++ b/flang/lib/Optimizer/HLFIR/Transforms/SimplifyHLFIRIntrinsics.cpp
@@ -33,7 +33,8 @@ namespace hlfir {
 static llvm::cl::opt<bool>
     simplifySum("flang-simplify-hlfir-sum",
                 llvm::cl::desc("Expand hlfir.sum into an inline sequence"),
-                llvm::cl::init(false));
+                llvm::cl::init(true));
+
 namespace {
 
 class TransposeAsElementalConversion