projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e770787) | patch
Implement smarter algorithm for choosing which blocks to tail-merge.
authorDale Johannesen <dalej@apple.com>
Fri, 1 Jun 2007 23:02:45 +0000 (23:02 +0000)
committerDale Johannesen <dalej@apple.com>
Fri, 1 Jun 2007 23:02:45 +0000 (23:02 +0000)
commita5a2117a4659ecadbec9476f2e537bcb7501b509
treef2514b99ff19042b85cb4f1e2d8bd28c682e9f46tree | snapshot
parente770787be101e522425f658f76e4bb3091498f99commit | diff
Implement smarter algorithm for choosing which blocks to tail-merge.
See test/CodeGen/X86/test-pic-jtbl.ll for a case where it works well;
shaves another 10K off our favorite benchmark.  I was hesitant about
this because of compile speed, but seems to do OK on a bootstrap.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@37392 91177308-0d34-0410-b5e6-96231b3b80d8
lib/CodeGen/BranchFolding.cpp diff | blob | history
test/CodeGen/X86/test-pic-jtbl.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors