You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@arrow.apache.org by "Antoine Pitrou (Jira)" <ji...@apache.org> on 2021/01/11 23:58:00 UTC
[jira] [Commented] (ARROW-11217) [C++] Runtime SIMD check on Apple
hardware missing
[ https://issues.apache.org/jira/browse/ARROW-11217?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17262989#comment-17262989 ]
Antoine Pitrou commented on ARROW-11217:
----------------------------------------
Example code:
[https://github.com/numpy/numpy/pull/13339/files]
Also a list of possible features to check for:
[https://www.mersenneforum.org/showthread.php?p=561100#post561100]
> [C++] Runtime SIMD check on Apple hardware missing
> --------------------------------------------------
>
> Key: ARROW-11217
> URL: https://issues.apache.org/jira/browse/ARROW-11217
> Project: Apache Arrow
> Issue Type: Bug
> Components: C++
> Reporter: Neal Richardson
> Priority: Critical
> Fix For: 3.0.0
>
>
> [~jeroenooms] hit a crash in the "sum" compute kernel using the R package on a new M1 machine running the rosetta emulator: https://gist.github.com/jeroen/c60548b29ff7f6807a6554799bd01cb7
> According to https://developer.apple.com/documentation/apple_silicon/about_the_rosetta_translation_environment, we should be checking sysctlbyname for AVX* capabilities, but we are not. We only use that function in https://github.com/apache/arrow/blob/master/cpp/src/arrow/util/cpu_info.cc#L350-L359 to check cpu cache size.
> This may also explain a crash we observed previously on a very old macOS CRAN machine.
> I think we should to resolve this before the 3.0 release if possible, in order to avoid bug reports as more people get M1s.
> cc [~apitrou] [~uwe] [~kou] [~frankdu] [~yibo]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)