From: Marek Olšák Date: Tue, 30 Jun 2020 20:57:23 +0000 (-0400) Subject: radeonsi: prevent a gfx10_ngg_calculate_subgroup_info failure for TES+NGG GS X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=081691b5ae2281e047b75beaa7462c448e74da04;p=mesa.git radeonsi: prevent a gfx10_ngg_calculate_subgroup_info failure for TES+NGG GS arb_tessellation_shader-tes-gs-max-output -small -scan 1 50 -auto -fbo doesn't pass, but at least all shaders are compiled successfully. Acked-by: Pierre-Eric Pelloux-Prayer Part-of: --- diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c index d1a1aa725cd..01ed224b282 100644 --- a/src/gallium/drivers/radeonsi/si_state_shaders.c +++ b/src/gallium/drivers/radeonsi/si_state_shaders.c @@ -2651,9 +2651,15 @@ static void *si_create_shader_selector(struct pipe_context *ctx, sel->gs_input_verts_per_prim = u_vertices_per_prim(sel->info.properties[TGSI_PROPERTY_GS_INPUT_PRIM]); - /* EN_MAX_VERT_OUT_PER_GS_INSTANCE does not work with tesselation. */ + /* EN_MAX_VERT_OUT_PER_GS_INSTANCE does not work with tesselation so + * we can't split workgroups. Disable ngg if any of the following conditions is true: + * - num_invocations * gs_max_out_vertices > 256 + * - LDS usage is too high + */ sel->tess_turns_off_ngg = sscreen->info.chip_class >= GFX10 && - sel->gs_num_invocations * sel->gs_max_out_vertices > 256; + (sel->gs_num_invocations * sel->gs_max_out_vertices > 256 || + sel->gs_num_invocations * sel->gs_max_out_vertices * + (sel->info.num_outputs * 4 + 1) > 6500 /* max dw per GS primitive */); break; case PIPE_SHADER_TESS_CTRL: