public inbox for glibc-bugs@sourceware.org
help / color / mirror / Atom feed
From: "jahns at dkrz dot de" <sourceware-bugzilla@sourceware.org>
To: glibc-bugs@sources.redhat.com
Subject: [Bug math/13185] New: fegetexceptflag does not report SSE exceptions on i386
Date: Tue, 13 Sep 2011 16:38:00 -0000	[thread overview]
Message-ID: <bug-13185-131@http.sourceware.org/bugzilla/> (raw)

http://sourceware.org/bugzilla/show_bug.cgi?id=13185

             Bug #: 13185
           Summary: fegetexceptflag does not report SSE exceptions on i386
           Product: glibc
           Version: unspecified
            Status: NEW
          Severity: minor
          Priority: P2
         Component: math
        AssignedTo: aj@suse.de
        ReportedBy: jahns@dkrz.de
    Classification: Unclassified


While SSE is not part of the ABI in i386, I'd like it to report SSE exceptions
the same way as in x86_64 in case the CPU does have SSE registers.

The difference is obvious if one looks at sysdeps/i386/fpu/fgetexcptflg.c vs.
sysdeps/x86_64/fpu/fgetexcptflg.c, i.e. while the latter queries the FPU status
word and the SSE status register, the former only inspects the FPU status word.

Specifically, if one compiles code with -mfpmath=sse and -msse or -msse2 on
i386, parts of fenv.h no longer work as advertised. One could argue this is use
of an unsupported feature from libc point-of-view. But since it took me some
time to figure out why code relying on proper exception handling no longer
worked when switching on sse, I assume other people would prefer not to go down
this rabbit-hole and have fenv.h functions "just work".

I assume this requires some tricks with CPUID in assembly and a one-time-check
if the kernel handles SSE on context-switch. I'd volunteer to read up on this
and provide a patch in case nobody cares. But first I'd like to hear if my idea
is consistent with how the maintainers feel the issue should be addressed. On
the other hand, gcc and any other optimizing compiler could provide an
implementation in its runtime, but fixing this once in glibc seems like the
saner solution to me.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.


             reply	other threads:[~2011-09-13 16:38 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-09-13 16:38 jahns at dkrz dot de [this message]
2011-12-22 21:31 ` [Bug math/13185] " drepper.fsp at gmail dot com
2014-06-27 12:04 ` fweimer at redhat dot com

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bug-13185-131@http.sourceware.org/bugzilla/ \
    --to=sourceware-bugzilla@sourceware.org \
    --cc=glibc-bugs@sources.redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).