projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: b1f8aeb) | patch
Hoist the contents of Loops in depth first order in the dominator tree,
authorChris Lattner <sabre@nondot.org>
Sun, 29 Sep 2002 21:46:09 +0000 (21:46 +0000)
committerChris Lattner <sabre@nondot.org>
Sun, 29 Sep 2002 21:46:09 +0000 (21:46 +0000)
commit952eaee239d6c38f5121ed68277555344c2d6bf0
treed11a491bf1b482682eee75831bfb66de3ed2b8bdtree | snapshot
parentb1f8aebc670769a19d54c9c4a8f3085f2bf52f1ccommit | diff
Hoist the contents of Loops in depth first order in the dominator tree,
rather than in random order.  This causes LICM to be DRAMATICALLY more
effective. For example, on bzip2.c, it is able to hoist 302 loads and
2380 total instructions, as opposed to 44/338 before.  This  obviously
makes other transformations much more powerful as well!

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@3978 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Transforms/Scalar/LICM.cpp diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors