radeonsi: optimize scissor states

- convert 16 states to 1 atom
- only emit 1 scissor if VIEWPORT_INDEX isn't written
- use only one packet when emitting consecutive scissors

Reviewed-by: Alex Deucher <[email protected]>
Acked-by: Christian König <[email protected]>

1 files changed, 20 insertions, 0 deletions

diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index b223e060be3..5a9ef29a549 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -760,6 +760,23 @@ static void *si_create_tes_state(struct pipe_context *ctx,
 	return si_create_shader_state(ctx, state, PIPE_SHADER_TESS_EVAL);
 }
 
+/**
+ * Normally, we only emit 1 viewport and 1 scissor if no shader is using
+ * the VIEWPORT_INDEX output, and emitting the other viewports and scissors
+ * is delayed. When a shader with VIEWPORT_INDEX appears, this should be
+ * called to emit the rest.
+ */
+static void si_update_viewports_and_scissors(struct si_context *sctx)
+{
+	struct tgsi_shader_info *info = si_get_vs_info(sctx);
+
+	if (!info || !info->writes_viewport_index)
+		return;
+
+	if (sctx->scissors.dirty_mask)
+	    si_mark_atom_dirty(sctx, &sctx->scissors.atom);
+}
+
 static void si_bind_vs_shader(struct pipe_context *ctx, void *state)
 {
 	struct si_context *sctx = (struct si_context *)ctx;
@@ -770,6 +787,7 @@ static void si_bind_vs_shader(struct pipe_context *ctx, void *state)
 
 	sctx->vs_shader = sel;
 	si_mark_atom_dirty(sctx, &sctx->clip_regs);
+	si_update_viewports_and_scissors(sctx);
 }
 
 static void si_bind_gs_shader(struct pipe_context *ctx, void *state)
@@ -787,6 +805,7 @@ static void si_bind_gs_shader(struct pipe_context *ctx, void *state)
 
 	if (enable_changed)
 		si_shader_change_notify(sctx);
+	si_update_viewports_and_scissors(sctx);
 }
 
 static void si_bind_tcs_shader(struct pipe_context *ctx, void *state)
@@ -821,6 +840,7 @@ static void si_bind_tes_shader(struct pipe_context *ctx, void *state)
 		si_shader_change_notify(sctx);
 		sctx->last_tes_sh_base = -1; /* invalidate derived tess state */
 	}
+	si_update_viewports_and_scissors(sctx);
 }
 
 static void si_make_dummy_ps(struct si_context *sctx)