[Bug 256152] graphics/qt5-3d: fails to work on processors with CPU < haswell (anything without AVX2)

From: <bugzilla-noreply_at_freebsd.org>
Date: Tue, 25 May 2021 15:52:23 +0000
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=256152

            Bug ID: 256152
           Summary: graphics/qt5-3d: fails to work on processors with CPU
                    < haswell (anything without AVX2)
           Product: Ports & Packages
           Version: Latest
          Hardware: amd64
                OS: Any
            Status: New
          Severity: Affects Many People
          Priority: ---
         Component: Individual Port(s)
          Assignee: kde_at_FreeBSD.org
          Reporter: crahman_at_gmail.com
             Flags: maintainer-feedback?(kde_at_FreeBSD.org)
          Assignee: kde_at_FreeBSD.org

The fix to bug#248155 forces the build of qt5-3d to generate AVX2 instructions
even on systems that do not support it.

On a Q9650 penryn (which does not support AVX2), with or without CPUTYPE set to
core2:

  Use SSE2 instructions .................. yes
  Use AVX2 instructions .................. yes

results during configuration.

As a result, use of qt5-3d after building it on a pre-AVX2 machine results in
failure via SIGILL.  This includes attempting to build the dependent port
pyside2, and is the source of bug#255753.

The problem is related to comment#6 in bug#248155:
 "Yeah, someone should invest some time in why the AVX2 detection fails."

However, until that problem is fixed it would be better to tolerate the
diminished performance resulting from underestimating CPU capabilities than the
cryptic failure of dependent systems to run at all.  Perhaps in the meanwhile
the use of AVX2 instructions should be made a port configuration option?

Thanks!

-- 
You are receiving this mail because:
You are the assignee for the bug.
Received on Tue May 25 2021 - 15:52:23 UTC

Original text of this message