Re: [Gnash] speed issues

[Bernhard Rosenkraenzer]

On Saturday, 4. March 2006 21:20, Rob Savoye wrote:
> There shouldn't be any special options needed for OpenGL, it's all hardwired
> for now. Could you somehow be getting the  wrong opengl library at link
> time ?

No, I'm even getting r300 specific warnings -- e.g. if I run
LIBGL_DEBUG=verbose gnash gnash/testsuite/movies.all/elvis.swf
it results in
libGL: XF86DRIGetClientDriverName: 5.0.3 r300 (screen 0)
libGL: OpenDriver: trying /usr/lib/dri/r300_dri.so
drmOpenByBusid: Searching for BusID pci:0000:01:00.0
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is 11, (OK)
drmOpenByBusid: drmOpenMinor returns 11
drmOpenByBusid: drmGetBusid reports pci:0000:01:00.0
libGL error:
Can't open configuration file /etc/drirc: No such file or directory.
libGL error:
Can't open configuration file /home/arklinux/.drirc: No such file or 
directory.
*********************************WARN_ONCE*********************************
File r300_render.c function r300Fallback line 805
fallback:ctx->Line.SmoothFlag
***************************************************************************
*********************************WARN_ONCE*********************************
File r300_state.c function r300Enable line 457
TODO - double side stencil !
***************************************************************************
*********************************WARN_ONCE*********************************
File r300_ioctl.c function r300Clear line 555
CB_DPATH has been enabled.
Please let me know if this introduces new instabilities.
***************************************************************************

(The same box runs Tuxracer and even Nexuiz well, so I know the various 
warnings r300 produces in LIBGL_DEBUG=verbose mode are harmless).

strace-ing gnash spends most of its time in a loop doing [8 == FD 
of /dev/dri/card0]
ioctl(8, 0x6444, 0) = 0 [30 times]
ioctl(8, 0xc0086451, [some address]) = 0
ioctl(8, 0xc010643a, [some address]) = 0
ioctl(8, 0x6447, 0) = 0
gettimeofday({timestamp, timestamp}, NULL) = 0
sched_yield() = 0
select(0, NULL, NULL, NULL, { 0, 31000}) = 0
gettimeofday([timestamp, timestamp], NULL) = 0
(start over with 30 x ioctl(8, 0x6444, 0))

I've tried it again on a different box (Radeon 9200), and it works as expected 
there -- so this is probably some interaction between gnash and the r300 
driver (or gnash making use of some OpenGL operation that is implemented for 
r200, but falls back to software rendering for r300 -- but I think if it were 
that, the r300 driver would warn about it - it's usually pretty verbose).

strace output is slightly different on the r200 - all the ioctl(8, 0x6444, 0)s 
are gone.

I'll try again with a current CVS snapshot of the r300 driver.

Regards,
bero