diff options
| author | Evan Cheng <evan.cheng@apple.com> | 2008-03-24 06:37:48 +0000 |
|---|---|---|
| committer | Evan Cheng <evan.cheng@apple.com> | 2008-03-24 06:37:48 +0000 |
| commit | 8d84d5b62cdf2a772d51338136c7022a6e1ff931 (patch) | |
| tree | 4ad44203af29e98cb7c89ddd60b731e582e9fd4a /include/llvm/Transforms/IPO | |
| parent | 16b412c6e278d84bcf71b1bf07a7d3c24e81a49e (diff) | |
| download | external_llvm-8d84d5b62cdf2a772d51338136c7022a6e1ff931.zip external_llvm-8d84d5b62cdf2a772d51338136c7022a6e1ff931.tar.gz external_llvm-8d84d5b62cdf2a772d51338136c7022a6e1ff931.tar.bz2 | |
Increasing the inline limit from (overly conservative) 200 to 300. Given each BB costs 20 and each instruction costs 5, 200 means a 4 BB function + 24 instructions (actually less because caller's size also contributes to it).
Furthermore, double the limit when more than 10% of the callee instructions are vector instructions. Multimedia kernels tend to love inlining.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48725 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'include/llvm/Transforms/IPO')
| -rw-r--r-- | include/llvm/Transforms/IPO/InlinerPass.h | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/include/llvm/Transforms/IPO/InlinerPass.h b/include/llvm/Transforms/IPO/InlinerPass.h index 26cf4e4..d34641d 100644 --- a/include/llvm/Transforms/IPO/InlinerPass.h +++ b/include/llvm/Transforms/IPO/InlinerPass.h @@ -55,6 +55,11 @@ struct Inliner : public CallGraphSCCPass { /// virtual int getInlineCost(CallSite CS) = 0; + // getInlineFudgeFactor - Return a > 1.0 factor if the inliner should use a + // higher threshold to determine if the function call should be inlined. + /// + virtual float getInlineFudgeFactor(CallSite CS) = 0; + private: // InlineThreshold - Cache the value here for easy access. unsigned InlineThreshold; |