radeonsi/gfx10: don't compile the GS copy shader if it's 100% not needed

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index 1eacef63e7223872fec581f30e29499f967fb55a..070ccd2ad1775a1cc611ebd51aebd88999726e33 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -369,6 +369,7 @@ struct si_shader_selector {
        unsigned        gsvs_vertex_size;
        unsigned        max_gsvs_emit_size;
        unsigned        enabled_streamout_buffer_mask;
+       bool            tess_turns_off_ngg;
 
        /* PS parameters. */
        unsigned        color_attr_index[2];

diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index f913b541e61c047bf63196b2ba0e89cd804144af..b381b3953bc54e5aa74fd79b0129966af9b334c8 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -2533,12 +2533,9 @@ static void si_init_shader_selector_async(void *job, int thread_index)
                }
        }
 
-       /* The GS copy shader is always pre-compiled.
-        *
-        * TODO-GFX10: We could compile the GS copy shader on demand, since it
-        * is only used in the (rare) non-NGG case.
-        */
-       if (sel->type == PIPE_SHADER_GEOMETRY) {
+       /* The GS copy shader is always pre-compiled. */
+       if (sel->type == PIPE_SHADER_GEOMETRY &&
+           (sscreen->info.chip_class <= GFX9 || sel->tess_turns_off_ngg)) {
                sel->gs_copy_shader = si_generate_gs_copy_shader(sscreen, compiler, sel, debug);
                if (!sel->gs_copy_shader) {
                        fprintf(stderr, "radeonsi: can't create GS copy shader\n");
@@ -2714,6 +2711,13 @@ 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. */
+               sel->tess_turns_off_ngg =
+                       (sscreen->info.family == CHIP_NAVI10 ||
+                        sscreen->info.family == CHIP_NAVI12 ||
+                        sscreen->info.family == CHIP_NAVI14) &&
+                       sel->gs_num_invocations * sel->gs_max_out_vertices > 256;
                break;
 
        case PIPE_SHADER_TESS_CTRL:
@@ -2980,9 +2984,8 @@ static bool si_update_ngg(struct si_context *sctx)
 
        bool new_ngg = true;
 
-       /* EN_MAX_VERT_OUT_PER_GS_INSTANCE does not work with tesselation. */
        if (sctx->gs_shader.cso && sctx->tes_shader.cso &&
-           sctx->gs_shader.cso->gs_num_invocations * sctx->gs_shader.cso->gs_max_out_vertices > 256)
+           sctx->gs_shader.cso->tess_turns_off_ngg)
                new_ngg = false;
 
        if (new_ngg != sctx->ngg) {