On Fri, 2015-03-13 at 01:06 -0400, Ilia Mirkin wrote: > On Fri, Mar 13, 2015 at 12:46 AM, Matt Turner <matts...@gmail.com> wrote: > > Here are some ideas I think might be reasonable GSoC ideas. > > > > - GLSL linking in NIR > > - Would allow us to stop doing optimizations and other expensive > > things on GLSL IR > > > > - SSA in the i965/fs backend, and an SSA-based register allocator > > > > - Improve instruction scheduling in i965 (Nouveau has this on the > > Wiki as well. Maybe potential for code reuse) > > FWIW the nouveau thing would probably not involve a lot of reuse... > the instructions are fairly different, and a lot of the nouveau > project would have to be actually figuring out what _is_ a better > schedule rather than just implementing one. [Also nouveau presently > has *no* scheduling in place... so almost _anything_ would likely be > an improvement.] > > > > > - Maybe some OpenGL features (???), although lots of things are > > already in progress, like tessellation and compute shaders and others > > are probably on the critical path > > > > Are these good GSoC projects? Good -- being a combination of > > characteristics like usefulness, feasibility, well defined, and > > properly sized. > > > > Feel free to comment on them and to reply with your own. We can pick > > some we like and put them on the wiki. > > Another project that I think I mentioned on IRC is to go through the > ES extension list and add in all the "easy" ones (i.e. ones that are > already all or mostly implemented by desktop features), of which there > are a _ton_ (probably like 20). I'm just talking about the OES/EXT/KHR > ones, not the vendor ones. The majority would be modifying the various > paths to allow the logic to be executed in ES contexts and to > modify/add piglit tests verifying that the functionality can be used > in ES. Here are a bunch of differences between what the nvidia blob > exposes and mesa for the same hw: > > http://people.freedesktop.org/~imirkin/glxinfo/glxinfo.html#p=es&b=version&g=NVIDIA%20GF1xx%20%28GeForce%20400%2C%20500%29&diff > > Admittedly some of those are ES3.1-only, but far from all.
Just thought I'd point out there was an ES version of KHR_debug sent to the list [1] but didn't get reviewed or committed. The piglit tests were updated for ES and committed. http://lists.freedesktop.org/archives/mesa-dev/2014-September/067890.html > > -ilia > _______________________________________________ > mesa-dev mailing list > mesa-dev@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/mesa-dev _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
