Run an instcombine pass after inlining but before scalarrepl. This
allows instcombine to clean up the code, which makes more code
suitable for SRoA.This helps C++ code in particular, e.g. speeding up
tramp3d by 31%.
Index: llvm-backend.cpp
===================================================================
--- llvm-backend.cpp (revision 122440)
+++ llvm-backend.cpp (working copy)
@@ -245,6 +245,7 @@
}
PM->add(createTailDuplicationPass()); // Simplify cfg by
copying code
+ PM->add(createInstructionCombiningPass()); // Cleanup for
scalarrepl.
PM->add(createCFGSimplificationPass()); // Merge & remove BBs
PM->add(createScalarReplAggregatesPass()); // Break up
aggregate allocas
PM->add(createInstructionCombiningPass()); // Combine silly seq's
_______________________________________________
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
