[llvm-commits] Patch: Math Lib call optimization

[Weiming Zhao]

Hi Eric,

 

This is before the patch. 

After the patch, the failure number is the same, no new failures.

 

The regression tests results (before our patch) were included in the fail
list. After the patch, there is no changes on  failure numbers. (Expected
pass number is increased by one because we added a unit test file)

Regressin test results: (before patch):

  Expected Passes    : 5422

  Expected Failures  : 61

  Unsupported Tests  : 13

  Unexpected Passes  : 1

  Unexpected Failures: 40

 

Weiming Zhao

 

From: Eric Christopher [mailto:echristo at apple.com] 
Sent: Wednesday, November 02, 2011 1:11 PM
To: Weiming Zhao
Cc: llvm-commits at cs.uiuc.edu; rajav at codeaurora.org
Subject: Re: [llvm-commits] Patch: Math Lib call optimization

 

 

On Nov 2, 2011, at 12:55 PM, Weiming Zhao wrote:





Hi Eric,

 

We took LLVM r143352, Clang r142951 and run against the tests on ARM
Scorpion devices. The test-suite reported 61 fails and the test reported  1
Unexpected Passes  and 40  Unexpected Failures.

 

The test-suites were compiled with options: -O3, SMALL_PROBLEM_SIZE=1,
USE_REFERENCE_OUTPUT=1, TEST=SIMPLE

Among the 61 test-suites, 1 is compiling failure
(MultiSource/Application/ClamAV)  and the rests are running failures.

 

Is this with or without your patch? Were there regressions? Have you checked
performance? Were there regressions?

 

-eric

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20111102/caed70df/attachment.html>