14 Oct
2009
14 Oct
'09
11:06 a.m.
Am Mittwoch 14 Oktober 2009 14:15:27 schrieb Jonathan E. Brickman:
Thanks for that hint - got it running now without those flags. :)
Nah, while
experimenting i tried to alter the 32bit flags with ccmake
(using -march=i686, -O2 or removing the -msse2 because i wasn't sure if
my AthlonXP CPU supports it) which led to even more problems. So i think
they are in effect, and I stayed with the flags ccmake determined by
itself, "-O3 - march=native -msse -msse2 -mfpmath=sse -ffast-math
-fomit-frame-pointer"
Probably is it. Take out -msse and -msse2 and -mfpmath=sse, and see
what it does.
J.E.B.
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
Edgar
17 Oct
17 Oct
10:28 p.m.
New subject: yet another yoshimi bug fix, -> 0.022
On Wed, Oct 14, 2009 at 03:06:32PM +0200, Edgar Aichinger wrote:
Am Mittwoch 14 Oktober 2009 14:15:27 schrieb Jonathan
E. Brickman:
Thanks for that hint - got it running now without those flags. :)
Hmm. Why?
What is wrong with the code that SSE breaks?
-ken
Nah,
while experimenting i tried to alter the 32bit flags with ccmake
(using -march=i686, -O2 or removing the -msse2 because i wasn't sure if
my AthlonXP CPU supports it) which led to even more problems. So i think
they are in effect, and I stayed with the flags ccmake determined by
itself, "-O3 - march=native -msse -msse2 -mfpmath=sse -ffast-math
-fomit-frame-pointer"
Probably is it. Take out -msse and -msse2 and -mfpmath=sse, and see
what it does. 
21 Oct
21 Oct
8:20 a.m.
New subject: yet another yoshimi bug fix, -> 0.022
Am Sonntag 18 Oktober 2009 03:31:49 schrieb Jonathan E. Brickman:
Probably is it. Take out -msse and -msse2 and
-mfpmath=sse, and see
what it does.
Thanks for that hint - got it running now without those flags. :)
Hmm. Why?
What is wrong with the code that SSE breaks?
-ken
Nothing wrong with the code. The hardware being used included an AthlonXP
CPU, some models of which do not have SSE or SSE2, but instead have an
older technology, 3Dnow, only. It is possible that using -m3dnow instead
of all of those might help some. There is no -mfpmath=3dnow, though, I
checked the gcc docs.
J.E.B.
Well mine says it supports SSE only, but removing either of -msse2 or -
mfpmath=sse resulted in some compiler error about a headerfile IIRC, i had to
remove all the three options ccmake had chosen by default...
No time these days to investigate further. Will stay like that for at least
1-2 weeks as we are doing a bit of house renovation.
Edgar
8:55 a.m.
New subject: yet another yoshimi bug fix, -> 0.022
Edgar Aichinger wrote:
Am Sonntag 18 Oktober 2009 03:31:49 schrieb Jonathan
E. Brickman:
(fwiw) I think the potential benefits you might or might not gain over a basic
'-march=native -O3' really wouldn't justify the effort. Go the renovations
I'd
say!
cheers, Cal
Probably is it. Take out -msse and -msse2 and
-mfpmath=sse, and see
what it does.
Thanks for that hint - got it running now without those flags. :)
[ ... ]
No time these days to investigate further. Will stay like that for at least
1-2 weeks as we are doing a bit of house renovation.
5558
days inactive
5565
days old
linux-audio-user@lists.linuxaudio.org
4 comments
4 participants
participants (4)
-
cal -
Edgar Aichinger -
Jonathan E. Brickman -
Ken Restivo