radeonsi: skip unnecessary si_update_shaders calls

Small decrease in draw call overhead.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

4 files changed, 27 insertions, 7 deletions

diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 326b8191475..41e94487c8a 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -260,6 +260,7 @@ struct si_context {
 	struct si_vertex_element	*vertex_elements;
 	unsigned			sprite_coord_enable;
 	bool				flatshade;
+	bool				do_update_shaders;
 
 	/* shader descriptors */
 	struct si_descriptors		vertex_buffers;
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index c5b61e94114..47fc7a0b14b 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -572,6 +572,7 @@ static void si_bind_blend_state(struct pipe_context *ctx, void *state)
 	struct si_context *sctx = (struct si_context *)ctx;
 	si_pm4_bind_state(sctx, blend, (struct si_state_blend *)state);
 	si_mark_atom_dirty(sctx, &sctx->cb_render_state);
+	sctx->do_update_shaders = true;
 }
 
 static void si_delete_blend_state(struct pipe_context *ctx, void *state)
@@ -869,6 +870,7 @@ static void si_bind_rs_state(struct pipe_context *ctx, void *state)
 	si_update_poly_offset_state(sctx);
 
 	si_mark_atom_dirty(sctx, &sctx->clip_regs);
+	sctx->do_update_shaders = true;
 }
 
 static void si_delete_rs_state(struct pipe_context *ctx, void *state)
@@ -1018,6 +1020,7 @@ static void si_bind_dsa_state(struct pipe_context *ctx, void *state)
 		sctx->stencil_ref.dsa_part = dsa->stencil_ref;
 		si_mark_atom_dirty(sctx, &sctx->stencil_ref.atom);
 	}
+	sctx->do_update_shaders = true;
 }
 
 static void si_delete_dsa_state(struct pipe_context *ctx, void *state)
@@ -2388,6 +2391,7 @@ static void si_set_framebuffer_state(struct pipe_context *ctx,
 	}
 
 	sctx->need_check_render_feedback = true;
+	sctx->do_update_shaders = true;
 }
 
 static void si_emit_framebuffer_state(struct si_context *sctx, struct r600_atom *atom)
@@ -2628,6 +2632,7 @@ static void si_set_min_samples(struct pipe_context *ctx, unsigned min_samples)
 		return;
 
 	sctx->ps_iter_samples = min_samples;
+	sctx->do_update_shaders = true;
 
 	if (sctx->framebuffer.nr_samples > 1)
 		si_mark_atom_dirty(sctx, &sctx->msaa_config);
@@ -3267,6 +3272,7 @@ static void si_bind_vertex_elements(struct pipe_context *ctx, void *state)
 
 	sctx->vertex_elements = v;
 	sctx->vertex_buffers_dirty = true;
+	sctx->do_update_shaders = true;
 }
 
 static void si_delete_vertex_element(struct pipe_context *ctx, void *state)
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 24fa682926c..30b644ee42b 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -867,7 +867,7 @@ void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
 	struct si_context *sctx = (struct si_context *)ctx;
 	struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
 	struct pipe_index_buffer ib = {};
-	unsigned mask, dirty_fb_counter, dirty_tex_counter;
+	unsigned mask, dirty_fb_counter, dirty_tex_counter, rast_prim;
 
 	if (!info->count && !info->indirect &&
 	    (info->indexed || !info->count_from_stream_output))
@@ -911,15 +911,21 @@ void si_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
 	 * draw_vbo recursively, and before si_update_shaders, which uses
 	 * current_rast_prim for this draw_vbo call. */
 	if (sctx->gs_shader.cso)
-		sctx->current_rast_prim = sctx->gs_shader.cso->gs_output_prim;
+		rast_prim = sctx->gs_shader.cso->gs_output_prim;
 	else if (sctx->tes_shader.cso)
-		sctx->current_rast_prim =
-			sctx->tes_shader.cso->info.properties[TGSI_PROPERTY_TES_PRIM_MODE];
+		rast_prim = sctx->tes_shader.cso->info.properties[TGSI_PROPERTY_TES_PRIM_MODE];
 	else
-		sctx->current_rast_prim = info->mode;
+		rast_prim = info->mode;
 
-	if (!si_update_shaders(sctx) ||
-	    !si_upload_graphics_shader_descriptors(sctx))
+	if (rast_prim != sctx->current_rast_prim) {
+		sctx->current_rast_prim = rast_prim;
+		sctx->do_update_shaders = true;
+	}
+
+	if (sctx->do_update_shaders && !si_update_shaders(sctx))
+		return;
+
+	if (!si_upload_graphics_shader_descriptors(sctx))
 		return;
 
 	if (info->indexed) {
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index 47cf496ae41..87d0b7d9022 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -1360,6 +1360,7 @@ static void si_bind_vs_shader(struct pipe_context *ctx, void *state)
 
 	sctx->vs_shader.cso = sel;
 	sctx->vs_shader.current = sel ? sel->first_variant : NULL;
+	sctx->do_update_shaders = true;
 	si_mark_atom_dirty(sctx, &sctx->clip_regs);
 	r600_update_vs_writes_viewport_index(&sctx->b, si_get_vs_info(sctx));
 }
@@ -1375,6 +1376,7 @@ static void si_bind_gs_shader(struct pipe_context *ctx, void *state)
 
 	sctx->gs_shader.cso = sel;
 	sctx->gs_shader.current = sel ? sel->first_variant : NULL;
+	sctx->do_update_shaders = true;
 	si_mark_atom_dirty(sctx, &sctx->clip_regs);
 	sctx->last_rast_prim = -1; /* reset this so that it gets updated */
 
@@ -1394,6 +1396,7 @@ static void si_bind_tcs_shader(struct pipe_context *ctx, void *state)
 
 	sctx->tcs_shader.cso = sel;
 	sctx->tcs_shader.current = sel ? sel->first_variant : NULL;
+	sctx->do_update_shaders = true;
 
 	if (enable_changed)
 		sctx->last_tcs = NULL; /* invalidate derived tess state */
@@ -1410,6 +1413,7 @@ static void si_bind_tes_shader(struct pipe_context *ctx, void *state)
 
 	sctx->tes_shader.cso = sel;
 	sctx->tes_shader.current = sel ? sel->first_variant : NULL;
+	sctx->do_update_shaders = true;
 	si_mark_atom_dirty(sctx, &sctx->clip_regs);
 	sctx->last_rast_prim = -1; /* reset this so that it gets updated */
 
@@ -1431,6 +1435,7 @@ static void si_bind_ps_shader(struct pipe_context *ctx, void *state)
 
 	sctx->ps_shader.cso = sel;
 	sctx->ps_shader.current = sel ? sel->first_variant : NULL;
+	sctx->do_update_shaders = true;
 	si_mark_atom_dirty(sctx, &sctx->cb_render_state);
 }
 
@@ -2199,6 +2204,8 @@ bool si_update_shaders(struct si_context *sctx)
 		if (!si_update_spi_tmpring_size(sctx))
 			return false;
 	}
+
+	sctx->do_update_shaders = false;
 	return true;
 }