MLAA For Mesa Is Ready For Testing

[curaga]

It looks like you're running swrast, the oldest software driver. Other than that, I can't say what's broken there.

[nightcast2000]

Hi there curaga,
I think you are indeed right I was using the swrast driver for some reason..
Anyway, I got the latest commits from Mesa git and compiled mesa with:

./autogen.sh --enable-debug --enable-xcb --enable-glx-tls --enable-asm --enable-xorg --with-dri-drivers=nouveau --with-gallium-drivers=nouveau,swrast --with-expat=/usr/lib  --disable-egl --enable-gbm --enable-gl-osmesa --enable-gallium-gbm --enable-shared-glapi

After I compiled it, I exported the parameters: export LD_LIBRARY_PATH=/home/steve/mesa/lib
export LIBGL_DRIVERS_PATH=/home/steve/mesa/lib/gallium
export LIBGL_DEBUG=verbose

Here is the output of glxinfo: http://pastebin.com/WSWMg71a

I then cd'ed over to mlaatest/32 and ran ./mlaatest again I did get some errors and the whole thing froze again. I took a picture with my mobile phone which should be attacted to this post.

I've tried to run the test again to redirect the output to a txt file,but all I got is this:

ATTENTION: default value of option vblank_mode overridden by environment.
ATTENTION: default value of option vblank_mode overridden by environment.
nvfx_screen_get_param:94 -  Warning: unknown PIPE_CAP 29

I hope this helps, I understand the nvfx driver is certainly not ready at all, but I hope the information I have provided will be useful.

Regards,
Nightcast2000

[curaga]

I think your experience is similar to Michael's - gf6 nouveau has been broken for him too for a while. PIPE_CAP 29 is SM3, which your card should support, but the driver isn't updated to know it yet.

Thanks for the glxinfo, looks like nvfx doesn't expose ARB_shader_texture_lod yet, even though your hw would support it. So even if the driver were stable, it wouldn't run the effect right now.

It's a bit sad that this at least for now locks out gf6 and gf7, the generations able to run it but on the nvfx driver. Fermi on the nvc0 driver is known to work.

[oibaf]

I added the pp branch to mesa and I see the new driconf options. However when I enable them nothing change (I tried also the removing of color channel). RV530 on a r300g.

EDIT: same result with llvmpipe

If I get it to work will be OK to add to my PPA?

[curaga]

The pp branch only does llvmpipe + softpipe right now. I'm currently waiting for a Brazos netbook to ship to do some hw testing.

Sure, you can ship it - though with only software support right now, it's not too useful. It should not break anything else.


With llvmpipe, try: PP_DEBUG=1 pp_nored=1 glxgears
If you don't see the effect, please post the output.

[oibaf]

The pp branch only does llvmpipe + softpipe right now. I'm currently waiting for a Brazos netbook to ship to do some hw testing.

Sure, you can ship it - though with only software support right now, it's not too useful. It should not break anything else.


With llvmpipe, try: PP_DEBUG=1 pp_nored=1 glxgears
If you don't see the effect, please post the output.

OK, that works, thanks. Are the driconf settings still not properly working?

[curaga]

They are for me, but llvmpipe uses "dri" as the driver name, not dri2, for some reason.

[oibaf]

They are for me, but llvmpipe uses "dri" as the driver name, not dri2, for some reason.

OK, it works.

Strange however that when I set it with driconf it always ends up with driver="r300" even when .drirc is missing and when using 'LIBGL_ALWAYS_SOFTWARE=1 driconf' (also it correctly detect llvmpipe in the GUI)...

[oibaf]

Sure, you can ship it - though with only software support right now, it's not too useful. It should not break anything else.

OK, added to my PPA:

http://phoronix.com/forums/showthread.php?50038-Updated-and-Optimized-Ubuntu-Free-Graphics-Drivers&p=219460#post219460

[curaga]

FWIW, the queue is now bound to hw drivers as well. Anyone want to test the color filters?