[PATCH] D69825: [Clang][Driver] Re-use the calling process instead of creating a new process for the cc1 invocation

Russell Gallop via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Nov 27 02:31:25 PST 2019


russell.gallop added a comment.

In D69825#1760400 <https://reviews.llvm.org/D69825#1760400>, @aganea wrote:

> In D69825#1760373 <https://reviews.llvm.org/D69825#1760373>, @russell.gallop wrote:
>
> > It looks like the git apply didn't work, but the script continued so this was a duff experiment, please ignore. I'll try to fix this and run it again.
>


Note for other reviewers, the abba_test.ps1 script uses a git format patch, not the format you can download from Phabricator.

> Thanks Russell! Can you try running the script overnight? It'll give a better result and we'll see the error bar.

Okay. With the patch sorted, I ran for 10 hours overnight.

  Total iterations:  21
  
       |         Min       |        Mean       |       Median      |         Max       |
     A |  00:14:11.0340636 |  00:14:33.2494597 |  00:14:25.3644475 |  00:15:25.4082274 |
     B |  00:13:57.9445995 |  00:14:20.3073502 |  00:14:09.0937819 |  00:16:06.2885457 |
  Diff | -00:00:13.0894641 | -00:00:12.9421096 | -00:00:16.2706656 |  00:00:40.8803183 |

>From looking at the individual times this was stable for a couple of hours then there were some particularly long runs (on A and B), possibly some scheduled maintenance tasks, so I'd recommend using the median value to ignore these outliers. Based on the median this saves about 1.9% for me.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69825/new/

https://reviews.llvm.org/D69825





More information about the llvm-commits mailing list