On Tue, Aug 5, 2014 at 3:23 PM, Ilia Mirkin <imir...@alum.mit.edu> wrote:
> On Tue, Aug 5, 2014 at 4:14 PM, Bryan Cain <bryanca...@gmail.com> wrote: > > On Mon, Aug 4, 2014 at 11:54 PM, Ilia Mirkin <imir...@alum.mit.edu> > wrote: > >> > >> Another apporach I've tried is to just use a TEMP register as the > >> indirect offset here. Unfortunately this gets destroyed by > >> st_glsl_to_tgsi's various optimizations which assume that temp > >> registers can't be reladdr's and so messes up the values. I started > >> adding support for that, but then quickly realized that was probably > >> not the right thing to do. > >> > >> So... should I increase the number of address registers to 1? Or is > >> there some other simple approach that I'm missing? > >> > >> Thanks, > >> > >> -ilia > > > > > > The problem you are having with the optimizations in glsl_to_tgsi is > > because, as you say, it doesn't check the reladdr fields when calculating > > live intervals. I'm currently in the middle of reworking the live > interval > > calculation - if you don't mind waiting a day or two for me to finish the > > patch set and send it to the list, I can add that change to the reworked > > function in the patch set. > > I can certainly wait a little while. However from everything I can > tell, all current reladdr's are in the ADDR register file, so perhaps > I should just stick with that? > Either way is workable from glsl_to_tgsi's standpoint. I assume the gallium interface people have a better idea of what the preferred solution is. [Resending because I forgot to CC the list before.]
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev