On Mon, Aug 16, 2010 at 10:54 AM, Luca Barbieri <l...@luca-barbieri.com> wrote: > Regarding OpenCL, it was found that nv50 has the normalization bit in > the sampler view, while r600 has it in the shader instruction. > > Thus, for nv50, the are two options: > 1. Move the bit from the sampler state to the sampler view -> have the > OpenCL state tracker create two sampler views, one for each > normalization. > 2. Leave things as is -> have the nv50 driver create two "hardware > sampler views" (aka TIC entries) internally > > Right now I think nv50 just doesn't create sampler views as persistent > GPU objects (and reemits them on sampler changes), but I'm not really > sure about it. > > And for r600: > 1. Have a normalization bit in the shader -> have the OpenCL state > tracker recreate the shader if necessary > 2. Don't have the normalization bit in the shader -> have the r600 > driver recompile the shader if necessary
You could also include two tex clauses in the shader and use a shader bool to select which one you want at run time, then you wouldn't have to recompile. Alex > > If any Intel cards supports compute, that's also something to consider. > > Note however that all of this is independent from the resource flags. > _______________________________________________ > 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