Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 11 Feb 2003 02:00:13 -0800
From:      Alfred Perlstein <bright@mu.org>
To:        Stijn Hoop <stijn@win.tue.nl>
Cc:        "Thomas E. Zander" <riggs@rrr.de>, Alexandr Kovalenko <never@nevermind.kiev.ua>, Oliver Braun <obraun@freebsd.org>, ports-committers@freebsd.org, cvs-ports@freebsd.org, cvs-all@freebsd.org
Subject:   Re: cvs commit: ports/multimedia/mplayer Makefile distinfo pkg-plist ports/multimedia/mplayer/files patch-ad
Message-ID:  <20030211100013.GO88781@elvis.mu.org>
In-Reply-To: <20030211095745.GA17571@pcwin002.win.tue.nl>
References:  <200302101928.h1AJS6Gs088748@repoman.freebsd.org> <20030210200115.GA832@nevermind.kiev.ua> <20030211013436.GB576@trillian.mugiri.au> <20030211095745.GA17571@pcwin002.win.tue.nl>

next in thread | previous in thread | raw e-mail | index | archive | help
* Stijn Hoop <stijn@win.tue.nl> [030211 01:58] wrote:
> On Tue, Feb 11, 2003 at 09:34:36AM +0800, Thomas E. Zander wrote:
> > Am Mon, dem 10. Feb 2003, um 22:01 +0200 Uhr schrubte Alexandr Kovalenko:
> > > We should also note that if you are planning to play QT movies, you
> > > should add CPU_ENABLE_SSE to your kernel so that mplayer will not crash
> > > attempting to play it (looks like either mplayer or QT win32 libs can't
> > > detect if SSE is enabled).
> > 
> > No, that is not a qt-specific problem.
> > Mplayer checks which cpu it is running on and assumes (as I would do if
> > I were a program :-) ) that it can use SSE if it is a SSE-capable CPU.
> 
> I think this is a bad assumption. There should be a mechanism for mplayer
> to detect whether it can use SSE instructions without 'guessing' by
> checking the CPU type. As another poster asked, how do I know I can
> use the option? It's not exactly common knowledge that the kernel option
> might be needed.
> 
> But I also think it is strange that the fact that SSE instructions are
> available cannot be detected (and used) at runtime by the kernel. Is
> this unwise to do for performance reasons? Anyone care to elaborate?
> 
> A quick grep through /usr/src shows that there are not too many places
> that this option is used although it seems to affect process
> register storage, so I imagine it could be a performance hit if
> another check is needed per process context switch.
> 
> However I'm absolutely no kernel developer, so this is mostly guessing.
> Anyone with more clue reading this?

I would seem trivial to add a sysctl under machdep to tell userland
if sse is available and configured.

-Alfred

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-ports" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030211100013.GO88781>