[Mesa-dev] [PATCH] draw: fix incorrect vertex size computation in LLVM drawing code
Jose Fonseca
jfonseca at vmware.com
Mon Jan 20 09:28:14 PST 2014
Good catch Brian. Change looks good to me AFAICT.
We do need more this sort of assertions in the draw module. I never fail to get surprised at how intricate and sensitive draw module is -- mostly due the fact that it reaches out to the upstream gallium interface from within the driver. I do hope one day the state trackers just use geometry shaders for all those things, so that the draw module is a straighforward VS/GS processor, and does not need to create its own fragment sahders. (It is unfortunate that with util/u_blitter we replicate the bad example of reaching out to gallium interface from inside the driver.)
Jose
----- Original Message -----
> We were calling draw_total_vs_outputs() too early. The call to
> draw_pt_emit_prepare() could result in the vertex size changing.
> So call draw_total_vs_outputs() after draw_pt_emit_prepare().
>
> This fix would seem to be needed for the non-LLVM code as well,
> but it's not obvious. Instead, I added an assertion there to
> try to catch this problem if it were to occur there.
>
> Bugzilla:
> https://urldefense.proofpoint.com/v1/url?u=https://bugs.freedesktop.org/show_bug.cgi?id%3D72926&k=oIvRg1%2BdGAgOoM1BIlLLqw%3D%3D%0A&r=NMr9uy2iTjWVixC0wOcYCWEIYhfo80qKwRgdodpoDzA%3D%0A&m=91zldQ8lO5OhCBRVy4vcYnfNUSKL3Bo1lcxmHhpMqWQ%3D%0A&s=2dc25b5d2d9cdcc89cb242f8bdd09f59761026dd200cea7eb9dd3278928e3603
> Cc: 10.0 <mesa-stable at lists.freedesktop.org>
> ---
> .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 7 +++++--
> .../draw/draw_pt_fetch_shade_pipeline_llvm.c | 22
> ++++++++++++--------
> 2 files changed, 18 insertions(+), 11 deletions(-)
>
> diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
> b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
> index 8fcc170..2c5c4cd 100644
> --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
> +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
> @@ -72,8 +72,8 @@ static void fetch_pipeline_prepare( struct
> draw_pt_middle_end *middle,
>
> const unsigned gs_out_prim = (gs ? gs->output_primitive :
> u_assembled_prim(prim));
> - unsigned nr = MAX2( vs->info.num_inputs,
> - draw_total_vs_outputs(draw) );
> + unsigned nr_vs_outputs = draw_total_vs_outputs(draw);
> + unsigned nr = MAX2(vs->info.num_inputs, nr_vs_outputs);
>
> if (gs) {
> nr = MAX2(nr, gs->info.num_outputs + 1);
> @@ -129,6 +129,9 @@ static void fetch_pipeline_prepare( struct
> draw_pt_middle_end *middle,
> /* No need to prepare the shader.
> */
> vs->prepare(vs, draw);
> +
> + /* Make sure that the vertex size didn't change at any point above */
> + assert(nr_vs_outputs == draw_total_vs_outputs(draw));
> }
>
>
> diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
> b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
> index 9f17241..60ec528 100644
> --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
> +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
> @@ -141,19 +141,11 @@ llvm_middle_end_prepare( struct draw_pt_middle_end
> *middle,
> struct draw_geometry_shader *gs = draw->gs.geometry_shader;
> const unsigned out_prim = gs ? gs->output_primitive :
> u_assembled_prim(in_prim);
> - const unsigned nr = MAX2(vs->info.num_inputs,
> - draw_total_vs_outputs(draw));
> + unsigned nr;
>
> fpme->input_prim = in_prim;
> fpme->opt = opt;
>
> - /* Always leave room for the vertex header whether we need it or
> - * not. It's hard to get rid of it in particular because of the
> - * viewport code in draw_pt_post_vs.c.
> - */
> - fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 *
> sizeof(float);
> -
> -
> draw_pt_post_vs_prepare( fpme->post_vs,
> draw->clip_xy,
> draw->clip_z,
> @@ -177,6 +169,18 @@ llvm_middle_end_prepare( struct draw_pt_middle_end
> *middle,
> *max_vertices = 4096;
> }
>
> + /* Get the number of float[4] attributes per vertex.
> + * Note: this must be done after draw_pt_emit_prepare() since that
> + * can effect the vertex size.
> + */
> + nr = MAX2(vs->info.num_inputs, draw_total_vs_outputs(draw));
> +
> + /* Always leave room for the vertex header whether we need it or
> + * not. It's hard to get rid of it in particular because of the
> + * viewport code in draw_pt_post_vs.c.
> + */
> + fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 *
> sizeof(float);
> +
> /* return even number */
> *max_vertices = *max_vertices & ~1;
>
> --
> 1.7.10.4
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://urldefense.proofpoint.com/v1/url?u=http://lists.freedesktop.org/mailman/listinfo/mesa-dev&k=oIvRg1%2BdGAgOoM1BIlLLqw%3D%3D%0A&r=NMr9uy2iTjWVixC0wOcYCWEIYhfo80qKwRgdodpoDzA%3D%0A&m=91zldQ8lO5OhCBRVy4vcYnfNUSKL3Bo1lcxmHhpMqWQ%3D%0A&s=beda6ecc8e4e1c8c7778d049ee75198ed00145f1ec140f1a3233a2dc5a93c732
>
More information about the mesa-dev
mailing list