hiya,
...so I'm at apple's wwdc, and I just had to test pd and gem against
the new x86 OSX boxes: they're running a 3.6GHz P4 with 2GB RAM, 2MB
L2 cache...just downloaded the pd-0.38-4 source from miller's site,
and only had to change the location of Wish Shell.app to get it
running: tiger has a Tk.framework (8.4.7, so my cg stuff is
shipping!), and the framework has Wish Shell.app embedded...
...only one "error":
consistency check failed: cos~: unexpected machine alignment
...and the audio ain't working, but that's probably an endian issue
in portaudio...
...then chris egged me on over chat to try GEM...at first we tried
his old G3 build of 0.90, but no luck...then I just grabbed cvs HEAD
and used the new configure/make build, and a few tweaks later, it was
working quite respectably! Not bad for such a big change...btw, yes,
I did include the "--with-MMX" configure flag :-P
woot, jamie
you go tigital!
On Jun 8, 2005, at 12:58 AM, james tittle wrote:
hiya,
...so I'm at apple's wwdc, and I just had to test pd and gem against the new x86 OSX boxes: they're running a 3.6GHz P4 with 2GB RAM, 2MB L2 cache...just downloaded the pd-0.38-4 source from miller's site, and only had to change the location of Wish Shell.app to get it running: tiger has a Tk.framework (8.4.7, so my cg stuff is shipping!), and the framework has Wish Shell.app embedded...
...only one "error":
consistency check failed: cos~: unexpected machine alignment
...and the audio ain't working, but that's probably an endian issue in portaudio...
...then chris egged me on over chat to try GEM...at first we tried his old G3 build of 0.90, but no luck...then I just grabbed cvs HEAD and used the new configure/make build, and a few tweaks later, it was working quite respectably! Not bad for such a big change...btw, yes, I did include the "--with-MMX" configure flag :-P
woot, jamie
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Patrick Pagano, B.S.,M.F.A Digital Media Specialist Digital Worlds Institute University Of Florida (352) 294-2082
Hi Jamie,
The "cos~" problem I should have anticipated... if we're in MacOSX, the code assumes big-endian... I think. Proably the __FreeBSD_ test could also include MAXOSX, assuming it defines the constants BYTE_ORDER and LITTLE_ENDIAN in a unix-ish way...
#ifdef __FreeBSD__
... would become...
#if defined (__FreeBSD__) || defined(MACOSX)
cheers Miller
PS There might be 1 or 2 other endianness tests in the code...
On Tue, Jun 07, 2005 at 09:58:05PM -0700, james tittle wrote:
hiya,
...so I'm at apple's wwdc, and I just had to test pd and gem against
the new x86 OSX boxes: they're running a 3.6GHz P4 with 2GB RAM, 2MB
L2 cache...just downloaded the pd-0.38-4 source from miller's site,
and only had to change the location of Wish Shell.app to get it
running: tiger has a Tk.framework (8.4.7, so my cg stuff is
shipping!), and the framework has Wish Shell.app embedded......only one "error":
consistency check failed: cos~: unexpected machine alignment
...and the audio ain't working, but that's probably an endian issue
in portaudio......then chris egged me on over chat to try GEM...at first we tried
his old G3 build of 0.90, but no luck...then I just grabbed cvs HEAD
and used the new configure/make build, and a few tweaks later, it was
working quite respectably! Not bad for such a big change...btw, yes,
I did include the "--with-MMX" configure flag :-Pwoot, jamie
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
On Jun 8, 2005, at 7:52 AM, Miller Puckette wrote:
The "cos~" problem I should have anticipated... if we're in MacOSX, the code assumes big-endian... I think. Proably the __FreeBSD_ test could also include MAXOSX, assuming it defines the constants BYTE_ORDER and LITTLE_ENDIAN in a unix-ish way...
...yep, I checked for the __LITTLE_ENDIAN__ and it is defined, so we
should got with that kind of check...of course it also has the
expected __SSE__, __SSE2__, i386, __i386, blahblahblah...I still feel
strange seeing that on a mac :-) But it's certainly another
testament to the portability of pd!
jamie
thanks jamie, you have made me feel so much better
On Jun 8, 2005, at 12:19 PM, james tittle wrote:
On Jun 8, 2005, at 7:52 AM, Miller Puckette wrote:
The "cos~" problem I should have anticipated... if we're in MacOSX, the code assumes big-endian... I think. Proably the __FreeBSD_ test could also include MAXOSX, assuming it defines the constants BYTE_ORDER and LITTLE_ENDIAN in a unix-ish way...
...yep, I checked for the __LITTLE_ENDIAN__ and it is defined, so we should got with that kind of check...of course it also has the expected __SSE__, __SSE2__, i386, __i386, blahblahblah...I still feel strange seeing that on a mac :-) But it's certainly another testament to the portability of pd!
jamie
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
Patrick Pagano, B.S., M.F.A Digital Media Specialist University of Floirda Digital Worlds Institute 352-294-2082
Or you re-learn all of your vector techniques for SSE3 :)
'Mac Ach perpetual SSE3 thread' ??? I kid, I kid!!
On Jun 8, 2005, at 3:46 PM, chris clepper wrote:
On Jun 8, 2005, at 11:24 AM, Pat Pagano wrote:
thanks jamie, you have made me feel so much better
Just wait to you see the GEM performance on Mactel. I'd suggest
holding onto those dual G4/G5 machines for a while until OSX86 gets
up to speed (literally).cgc
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/ listinfo/pd-list
On Jun 9, 2005, at 8:52 PM, vade wrote:
Or you re-learn all of your vector techniques for SSE3 :)
'Mac Ach perpetual SSE3 thread' ??? I kid, I kid!!
It will probably happen once the appropriate kit is lined up. The former I mean, the latter might be too much for BA and Herr Bettag to handle! ;)
cgc
-
chris clepper -
james tittle -
Miller Puckette -
Pat Pagano -
Patrick Pagano
-
vade