projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 70a3511) | patch
[X86] Improved X86::CMOV to Branch heuristic.
authorAmjad Aboud <amjad.aboud@intel.com>
Tue, 8 Aug 2017 12:17:56 +0000 (12:17 +0000)
committerAmjad Aboud <amjad.aboud@intel.com>
Tue, 8 Aug 2017 12:17:56 +0000 (12:17 +0000)
commit6fa6813aeccadb8f0780aa0ee4468ccc443122d8
tree33fe6c66d7447c0288ad611bf712d5187ae23db6tree | snapshot
parent70a3511bd5a83d85ef277bb7b088e15917ba7915commit | diff
[X86] Improved X86::CMOV to Branch heuristic.
Resolved PR33954.
This patch contains two more constraints that aim to reduce the noise cases where we convert CMOV into branch for small gain, and end up spending more cycles due to overhead.

Differential Revision: https://reviews.llvm.org/D36081

llvm-svn: 310352
llvm/lib/Target/X86/X86CmovConversion.cpp diff | blob | history
llvm/test/CodeGen/X86/pr33954.ll [new file with mode: 0644] blob
llvm/test/CodeGen/X86/x86-cmov-converter.ll diff | blob | history
Domain: System / Toolchain;