projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 48afd9f) | patch
1. Drop default inline threshold back down to 200.
authorEvan Cheng <evan.cheng@apple.com>
Tue, 1 Apr 2008 23:59:29 +0000 (23:59 +0000)
committerEvan Cheng <evan.cheng@apple.com>
Tue, 1 Apr 2008 23:59:29 +0000 (23:59 +0000)
commit7c3becd9d729b976d79b18e6165e74390e613bb5
treef4c37ab938498cfa72b1a483b5161b49e30a888ftree | snapshot
parent48afd9f8a94e8db7b9894d97a7e85fff97cff1bdcommit | diff
1. Drop default inline threshold back down to 200.
2. Do not use # of basic blocks as part of the cost computation since it doesn't really figure into function size.
3. More aggressively inline function with vector code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@49061 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Transforms/IPO/Inliner.cpp diff | blob | history
lib/Transforms/Utils/InlineCost.cpp diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors