[MPlayer-DOCS] [PATCH] sse disabled on win32 into docs

[Rich Felker]

On Wed, Nov 30, 2005 at 04:10:20PM +0100, Guillaume Poirier wrote:
> compn wrote:
> > On Tue, 29 Nov 2005 12:57:11 +0100
> > Sascha Sommer <saschasommer at freenet.de> wrote:
> > 
> > 
> >>On Tuesday 29 November 2005 12:37, John Brown wrote:
> >>
> >>>>... the SSE routins get disabled by configure completely under win32, even
> >>>>when mencoder is not build.
> >>>
> >>>Are you sure?
> >>>
> >>>config.h
> >>>-----------
> >>>/* Extension defines */
> >>>...
> >>>#define HAVE_SSE 1	// only define if you have SSE (Intel Pentium III/4 or
> >>>Celeron II)
> >>>#define HAVE_SSE2 1	// only define if you have SSE2 (Intel Pentium 4)
> >>>
> >>>config.mak
> >>>--------------
> >>>TARGET_SSE = yes
> >>>
> >>>
> >>
> >>It gets disabled in the runtime cpudetection code.
> >>
> >>Sascha
> > 
> > 
> > patch to add 'sse support when building on win32 is disabled' to docs.
> > 
> > second patch to add note for 'p4 users to turn off hyperthreading if the are
> > experiencing a crash using realplayer codecs'. based upon bugreport in
> > mplayer-advusers.
> 
> Why on earth is sse disabled on win32? Could it be because currently
> ./configure would be using only /proc/cpuinfo to fetch informations
> about the CPU?
> In that case, I suggest we use TOOLS/cpuinfo.c to get a "/proc/cpuinfo
> output" which configure could parse and use...

No, it's because mingw (msvc) malloc is broken and does not provide
malign/aligned memory. sse instructions will cause segfault if the
alignment is incorrect.

Rich