[PATCH] D17288: [CodeGenPrepare] Do select to branch transform when cmp's operand is expensive.
Junmo Park via llvm-commits
llvm-commits at lists.llvm.org
Thu Apr 7 17:12:22 PDT 2016
flyingforyou updated this revision to Diff 52981.
flyingforyou added a comment.
Addressed Gerolf's comments.
Gerolf, I still don't understand what you said "In the case both paths consume the long latency select is still the better choice.".
Why is choosing csel better than branch when both patchs consume the long latency?
I think we can hide long latency by using branch.
http://reviews.llvm.org/D17288
Files:
include/llvm/MC/MCSchedule.h
include/llvm/Target/TargetSchedule.td
lib/CodeGen/CodeGenPrepare.cpp
lib/MC/MCSchedule.cpp
lib/Target/AArch64/AArch64SchedA57.td
test/CodeGen/X86/machine-combiner.ll
test/Transforms/CodeGenPrepare/X86/select.ll
utils/TableGen/SubtargetEmitter.cpp
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D17288.52981.patch
Type: text/x-patch
Size: 11704 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160408/8b088b0d/attachment.bin>
More information about the llvm-commits
mailing list