Internal Multi-Threading Arrives For Gallium3D's Direct3D 9 "Nine"

[phoronix]

Phoronix: Internal Multi-Threading Arrives For Gallium3D's Direct3D 9 "Nine"

Prolific Gallium Nine developer Axel Davy has published a massive patch-set today implementing support for internal multi-threading in this Direct3D 9 state tracker...

http://www.phoronix.com/scan.php?page=news_item&px=Gallium-Nine-Internal-MT

[cRaZy-bisCuiT]

Well, that's awesome! Games that already perform quite good with Gallium NINE may outperform Wine + CSMT by a far margin now. csmt_force=1 is meant as a kernel parameter?

[LinuxID10T]

This is fucking beautiful. Too bad the OpenGL driver still isn't properly multithreaded...

[shmerl]

How far is the multithreaded command stream for radeonsi itself?

[artivision]

There are references on some games for D3D9E renderers. What is D3D9E? Doesn't seam related to EX.

[haagch]

Happy testing, wine-staging csmt has been long available so it is more stable than a brand new nine code. It might take years to test with all the games and hardware to achieve the same stability.

Is there reason to believe it will be significantly more unstable?
The versions in the nine csmt branch have been working without issue for me so far.

That said, with nine master currently csgo crashes after a short time of gameplay.

=>0 0xf751d3ff __memcpy_sse2_unaligned+0x2f() in libc.so.6 (0x200ae7c8)
  1 0xe7eed01f in winepulse (+0xd01e) (0x200ae7c8)

unaligned... hm...
Hopefully that's not another stack alignment issue.

[trek]

Happy testing, wine-staging csmt has been long available so it is more stable than a brand new nine code. It might take years to test with all the games and hardware to achieve the same stability.

frankly I tested three games with csmt and nine and where nine is able to run fine and fast every game, with csmt I found: 1 game does not start and others run half-speed of nine, but 1 game (bioshock demo) has visual glitches with or without csmt but not with nine

so I strongly suggest to try nine for games

[geearf]

Awesome work!

Now I'm curious, was it easier to implement csmt in Nine than standard wine (it's been years I think)? and if so, in a few words why?

[mannerov]

csmt_force=1 is meant as a kernel parameter?

no, that's a drirc option, meaning you can put it in you drirc file, or as env var.

[mannerov]

Awesome work!

Now I'm curious, was it easier to implement csmt in Nine than standard wine (it's been years I think)? and if so, in a few words why?

I think what's blocking for csmt in wine is that the devs haven't had time so far to make the cleanups they wanted to the patchset, and instead just rebased it years over years over their wine changes. I don't think however it must have been harder to implement that it has been for nine.