On 22 November 2014 at 00:16, Andrei Slăvoiu andrei.slavoiu@gmail.com wrote:
I don't think we want this. We're never going to use EXT_geometry_shader4, and in fact are likely to drop ARB_geometry_shader4 in favour of just using core contexts. We should just do the work to make core contexts work. Matteo has some initial patches if you'd like to help out with that.
On Monday 24 November 2014 09:39:14 Henri Verbeet wrote:
I don't think we want this. We're never going to use EXT_geometry_shader4, and in fact are likely to drop ARB_geometry_shader4 in favour of just using core contexts. We should just do the work to make core contexts work. Matteo has some initial patches if you'd like to help out with that.
The sole purpose of this is to allow the fallback code to identify DX10 capable cards *now*, before the core contexts work is done. Do you think the code for using core contexts could become a reality in a reasonable time frame? This year? Early next year?
BTW, where can I find those patches? I don't remember seeing something like that on the mailing list.
2014-11-24 13:03 GMT+01:00 Andrei Slavoiu andrei.slavoiu@gmail.com:
On Monday 24 November 2014 09:39:14 Henri Verbeet wrote:
I don't think we want this. We're never going to use EXT_geometry_shader4, and in fact are likely to drop ARB_geometry_shader4 in favour of just using core contexts. We should just do the work to make core contexts work. Matteo has some initial patches if you'd like to help out with that.
The sole purpose of this is to allow the fallback code to identify DX10 capable cards *now*, before the core contexts work is done. Do you think the code for using core contexts could become a reality in a reasonable time frame? This year? Early next year?
Not going to happen this year, hopefully at some point next year. I agree with Henri that it isn't probably worth to add this kind of stuff just for the fallback.
BTW, where can I find those patches? I don't remember seeing something like that on the mailing list.
I haven't published them anywhere yet. They are still in quite a rough shape at the moment and I know I need to rewrite a few parts. Also patches are pretty much in the order in which I written them and I've got a few commits fixing stuff in earlier ones (which obviously need to be merged), or patches which need to be splitted and such. I guess I could put the series somewhere publicly accessible like GitHub, after some cleanup to achieve a minimum level of decency, but in general you would have to expect a lot of rebases and stuff like that even then.
My general plan is to start sending the initial patches "soon". There are a few prerequisite changes / additions to the Mac driver, opengl32 and wined3d which are quite independent from the rest and hopefully should be in submission-quality in a few days or so. After them it will be probably the turn of removing the use of builtin FFP variables in our GLSL shaders, which is one of those parts of my patch series which still needs significant cleanup...
În ziua de Lun 24 Noi 2014, la 21:54:02, Matteo Bruni a scris:
Not going to happen this year, hopefully at some point next year. I agree with Henri that it isn't probably worth to add this kind of stuff just for the fallback.
If it's not ok to add support for new extensions just to improve the fall-back mechanism then how about reverting to the old behaviour of treating all DX9 SM3 cards as DX10? While it was wrong, no one complained about it. But they do complain about the new behaviour. Most likely because most DX9 era cards are already explicitly recognized while most cards hitting the fall-back are newer ones that are actually DX10 or even DX11 capable.
I haven't published them anywhere yet.
Need any help with them? The least I could do is provide early testing on the radeonsi driver.
On 25 November 2014 at 00:34, Andrei Slăvoiu andrei.slavoiu@gmail.com wrote:
În ziua de Lun 24 Noi 2014, la 21:54:02, Matteo Bruni a scris:
Not going to happen this year, hopefully at some point next year. I agree with Henri that it isn't probably worth to add this kind of stuff just for the fallback.
If it's not ok to add support for new extensions just to improve the fall-back mechanism then how about reverting to the old behaviour of treating all DX9 SM3 cards as DX10? While it was wrong, no one complained about it. But they do complain about the new behaviour. Most likely because most DX9 era cards are already explicitly recognized while most cards hitting the fall-back are newer ones that are actually DX10 or even DX11 capable.
I don't have the impression that this is a big issue for most people. We certainly don't seem to get a lot of bug reports about this, and the workaround is fairly easy.
-
Andrei Slavoiu -
Henri Verbeet -
Matteo Bruni