[jira] [Commented] (KUDU-830) Ensure thirdparty builds with fastest compiler for releases

["Todd Lipcon (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KUDU-830?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16420894#comment-16420894 ] 

Todd Lipcon commented on KUDU-830:
----------------------------------

We should also look at enabling PGO on our compression codecs:
http://blosc.org/posts/codecs-pgo/

I did a few tests with lz4 here comparing clang4, clang6, gcc4, gcc7 and found that gcc is much better than clang: https://gist.github.com/ad2e68cfcdf9e7ca1957dfe3e981fbf9 and PGO only helps a bit. The dataset in these tests is a column of fairly-compressible (4x ratio) strings from a user-provided dataset.

> Ensure thirdparty builds with fastest compiler for releases
> -----------------------------------------------------------
>
>                 Key: KUDU-830
>                 URL: https://issues.apache.org/jira/browse/KUDU-830
>             Project: Kudu
>          Issue Type: Improvement
>          Components: build, perf
>    Affects Versions: Public beta
>            Reporter: Todd Lipcon
>            Priority: Major
>
> On el6 slaves, it appears we currently build all of thirdparty with clang 3.3 from toolchain. This is a pretty old compiler which generates slower code than gcc or newer clangs. It's not a big deal for most stuff, but probably measurable difference on compression codecs like zlib/lz4. We should make sure we use the best compiler for these perf hotspots.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)