On Tue, 8 Nov 2011 14:32:09 -0800, Kenneth Graunke <kenn...@whitecape.org> wrote: > Like for the WM pull constants, we can merge the former prepare/emit > stages into one tracked state atom. Furthermore, the code that used to > handle the binding table was removed in the last commit, leaving some > rather silly looking short functions that can easily be folded in. > > Signed-off-by: Kenneth Graunke <kenn...@whitecape.org>
> - &brw_vs_constants, /* Before vs_surfaces and constant_buffer */
> - &brw_wm_pull_constants, /* Before brw_wm_binding_table */
> + /* Pull constants must be before brw_binding_table */
> + &brw_vs_pull_constants,
> + &brw_wm_pull_constants,
> +
> &gen6_vs_push_constants, /* Before vs_state */
> &gen6_wm_push_constants, /* Before wm_state */
>
> - &brw_vs_surfaces, /* must do before unit */
> &brw_renderbuffer_surfaces, /* must do before unit */
> &brw_texture_surfaces, /* must do before unit */
> &brw_binding_table,
I think the pull constants atoms should live down with the rest of the
surfaces.
> -/**
> - * Update the surface state for a VS constant buffer.
> - *
> - * Sets brw->vs.surf_bo[surf] and brw->vp->const_buffer.
> - */
> -static void
> -brw_update_vs_constant_surface( struct gl_context *ctx,
> - GLuint surf)
> -{
> - struct brw_context *brw = brw_context(ctx);
> - struct intel_context *intel = &brw->intel;
> - struct brw_vertex_program *vp =
> - (struct brw_vertex_program *) brw->vertex_program;
> - const struct gl_program_parameter_list *params =
> vp->program.Base.Parameters;
> -
> - /* If there's no constant buffer, then no surface BO is needed to point at
> - * it.
> - */
> - if (brw->vs.const_bo == NULL) {
> - brw->bind.surf_offset[surf] = 0;
> - return;
> - }
>
> + const int surf = SURF_INDEX_VERT_CONST_BUFFER;
> intel->vtbl.create_constant_surface(brw, brw->vs.const_bo,
> params->NumParameters,
> &brw->bind.surf_offset[surf]);
> -const struct brw_tracked_state brw_vs_surfaces = {
> +const struct brw_tracked_state brw_vs_pull_constants = {
> .dirty = {
> - .mesa = 0,
> - .brw = (BRW_NEW_VS_CONSTBUF |
> - BRW_NEW_BATCH),
> - .cache = 0
> + .mesa = (_NEW_PROGRAM_CONSTANTS),
> + .brw = (BRW_NEW_VERTEX_PROGRAM),
> + .cache = CACHE_NEW_VS_PROG,
> },
> - .emit = brw_upload_vs_surfaces,
> + .emit = brw_upload_vs_pull_constants,
> };
Looks like BRW_NEW_BATCH got misplaced, which is required since you're
uploading state batch data (create_constant_surface()).
Hmm. You know, I ought to add some debug code that checks that
!bo_references(intel->batch.bo, old_batch_bo).
pgpm732kX1QIo.pgp
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
