[PATCH] D38154: [PassManager] Run global opts after the inliner

Davide Italiano via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Oct 4 14:47:52 PDT 2017


davide added a comment.

In https://reviews.llvm.org/D38154#888787, @mzolotukhin wrote:

> Hm, from the files you uploaded I don't see a slowdown on 7zip, but see some others. Are we talking about the same numbers?:)
>
>   Metric: compile_time
>   Program                                         unpatched  patched diff
>  
>   lencod/lencod                                  12.42      12.71    2.3%
>   kimwitu++/kc                                   13.47      13.76    2.1%
>   Bullet/bullet                                  28.41      28.99    2.0%
>   sqlite3/sqlite3                                 7.75       7.88    1.7%
>   consumer-typeset/consumer-typeset              11.19      11.37    1.6%
>   mafft/pairlocalalign                            7.16       7.25    1.3%
>   SPASS/SPASS                                    14.27      14.43    1.1%
>   ClamAV/clamscan                                12.96      13.06    0.8%
>   tramp3d-v4/tramp3d-v4                          18.24      18.19   -0.3%
>   7zip/7zip-benchmark                            34.97      34.96   -0.0%
>  
>  
>   Metric: size
>   Program                                         unpatched  patched diff
>  
>   mafft/pairlocalalign                           355304     353552  -0.5%
>   tramp3d-v4/tramp3d-v4                          954304     950064  -0.4%
>   lencod/lencod                                  646000     644472  -0.2%
>   Bullet/bullet                                  1752656    1751688 -0.1%
>   7zip/7zip-benchmark                            1224296    1224336  0.0%
>   kimwitu++/kc                                   1157536    1157504 -0.0%
>   ClamAV/clamscan                                658536     658536   0.0%
>   SPASS/SPASS                                    566744     566744   0.0%
>   consumer-typeset/consumer-typeset              544184     544184   0.0%
>   sqlite3/sqlite3                                402200     402200   0.0%
>
>
> BTW, is it possible to pull a testcase from rust codebase for CTMark? I think it's good to add more real world examples there.
>
> Michael


Sorry, I meant `sqlite3/sqlite3`.
How do you get the diff out of curiosity? I eyeballed the differences (and I missed something, of course :))
BTW, yes, we can probably  work out to add some rust test to ctmark, I'll ask Alex Crichton


https://reviews.llvm.org/D38154





More information about the llvm-commits mailing list