[llvm-commits] PATCH: Add pass printing and a basic sanity test for BlockFrequencyInfo analysis

[Nick Lewycky]

diff --git a/lib/Analysis/BlockFrequencyInfo.cpp
b/lib/Analysis/BlockFrequencyInfo.cpp
index ee89915..9e05065 100644
--- a/lib/Analysis/BlockFrequencyInfo.cpp
+++ b/lib/Analysis/BlockFrequencyInfo.cpp
@@ -49,6 +49,10 @@ bool BlockFrequencyInfo::runOnFunction(Function &F) {
   return false;
 }

+void BlockFrequencyInfo::print(raw_ostream &O, const Module * /*M*/) const
{

Just const const Module* will do.

+  if (BFI) BFI->print(O);
+}
+

+; Loop backedges are weighted, and their bodies represented as having a
greater
+; frequency.

Please re-english. :)

Please commit once you've fixed those.

Nick

On 16 October 2011 23:30, Chandler Carruth <chandlerc at google.com> wrote:

> (Sent to the wrong address for Jakub it seems...)
>
>
> On Sun, Oct 16, 2011 at 11:25 PM, Chandler Carruth <chandlerc at google.com>wrote:
>
>> Hello,
>>
>> So it turns out that BranchProbabilityInfo and BlockFrequencyInfo aren't
>> rigged up to __builtin_expect at all. I think no one realized because we
>> have essentially no test coverage for these analysis passes. See
>> http://llvm.org/PR2577
>>
>> This patch adds test coverage for BlockFrequencyInfo which is the easiest
>> to test, and essentially a good way to strictly cover BranchProbabilityInfo.
>> Subsequent patches to add the remaining features to actually use
>> __builtin_expect will follow with tests based on this.
>> -Chandler
>>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20111017/eeabb18f/attachment.html>