On 09/06/2010 06:58 PM, Luca Barbieri wrote: >> And prog_optimize can't do anything if there are temporaries with indirect >> addressing. We need either the temporary array file or to do register >> allocation in GLSL IR. > > Sure, but I think this can/should be done on top of this change, and > thus isn't an argument against it. > But you said that this particular shader doesn't use indirect TEMP addressing, so I think this patch would be useful after all, if we don't want to do register allocation 3 times (GLSL IR, prog_optimize, and the driver (nv50 uses linear scan)).
> Instead of a temporary register file, just adding bounds to indirect > addressing instructions would be slightly more expressive. Not sure > which is better. > I'd still like a separate declaration for each array in addition to addressing bounds, which would certainly be helpful as well though. > DirectX 10 has indirect temp declarations with "dcl_indexableTemp". > > Anyway, pre-DirectX 10 cards such as nv30/nv40 often don't support > indirect temps at all, so they will need to be lowered somewhere, > ideally in a GLSL lowering pass, so that we can kill some of the > conditional assignments if we know that the index cannot take that > value. > _______________________________________________ > 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
