On Wed, Jan 24, 2018 at 11:13 AM, Kenneth Graunke <kenn...@whitecape.org>
wrote:
> On Wednesday, January 24, 2018 2:03:41 AM PST Jason Ekstrand wrote:
> > Instead of keeping an array of booleans, we now hang onto an array of
> > isl_aux_usage enums. This means that the thing we are passing from
> > brw_draw.c to surface state setup is the thing that surface state setup
> > actually needs instead of an input to compute what it needs.
> >
> > Cc: mesa-sta...@lists.freedesktop.org
> [snip]
> > @@ -497,7 +498,8 @@ brw_predraw_resolve_inputs(struct brw_context *brw,
> bool rendering)
> > }
> >
> > static void
> > -brw_predraw_resolve_framebuffer(struct brw_context *brw)
> > +brw_predraw_resolve_framebuffer(struct brw_context *brw,
> > + bool *draw_aux_buffer_disabled)
> > {
> > struct gl_context *ctx = &brw->ctx;
> > struct intel_renderbuffer *depth_irb;
> > @@ -550,7 +552,8 @@ brw_predraw_resolve_framebuffer(struct brw_context
> *brw)
> > enum isl_aux_usage aux_usage =
> > intel_miptree_render_aux_usage(brw, irb->mt, isl_format,
> > blend_enabled,
> > - brw->draw_aux_buffer_disabled[
> i]);
> > + draw_aux_buffer_disabled[i]);
> > + brw->draw_aux_usage[i] = aux_usage;
>
> I'm concerned this might break compute shaders. Sure, they don't have
> renderbuffers. However, the code to emit SURFACE_STATEs for textures
> calls brw_aux_surface_disabled which brw->draw_aux_usage[]. It only
> accesses elements 0..fb->_NumColorDrawBuffers-1, but I suspect there's
> still a framebuffer bound when doing a compute dispatch - it's just
> not relevant.
>
Compute shaders will never call this as it is the framebuffer funciton.
They only call predraw_resolve_inputs.
> This appears to be the only place that sets brw->draw_aux_usage[],
> and it isn't called for compute shaders.
>
> I'd say to just whack it back to "don't override things", but that
> means knowing what the actual usage is, which sounds painful...
>
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
