On Fri, 2015-07-24 at 16:20 +0300, Francisco Jerez wrote:
> Iago Toral Quiroga <ito...@igalia.com> writes:
>
> > Larger registers should have been moved to scratch (like GRF array access)
> > or split to size 1 by the split_virtual_grfs pass.
>
> Not necessarily. split_virtual_grfs() won't be able to split stuff
> which is read or written at once by the same instruction -- E.g. by
> send-from-GRF instructions as used for surface messages on e.g. your
> SSBO implementation. :)
>
> We should probably eventually migrate other messages too like the ones
> used for texturing and framebuffer writes to use proper sends from
> GRF...
Okay, in that case I'll include patches to add support for spilling
registers with size > 1 as well.
Thanks,
Iago
> > ---
> > src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp | 3 ++-
> > 1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
> > b/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
> > index cff5406..80ab813 100644
> > --- a/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
> > +++ b/src/mesa/drivers/dri/i965/brw_vec4_reg_allocate.cpp
> > @@ -271,7 +271,8 @@ vec4_visitor::evaluate_spill_costs(float *spill_costs,
> > bool *no_spill)
> >
> > for (unsigned i = 0; i < this->alloc.count; i++) {
> > spill_costs[i] = 0.0;
> > - no_spill[i] = alloc.sizes[i] != 1;
> > + no_spill[i] = false;
> > + assert(this->alloc.sizes[i] == 1);
> > }
> >
> > /* Calculate costs for spilling nodes. Call it a cost of 1 per
> > --
> > 1.9.1
> >
> > _______________________________________________
> > 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
