vc4: Skip re-emitting the shader_rec if it's unchanged.

It's a bunch of work for us to emit it (and its uniforms), more work for
the kernel to validate it, and additional work for the CLE to read
it. Improves es2gears framerate by about 50%.

Signed-off-by: Eric Anholt <[email protected]>

1 files changed, 15 insertions, 1 deletions

diff --git a/src/gallium/drivers/vc4/vc4_context.h b/src/gallium/drivers/vc4/vc4_context.h
index 7faf5223630..30fb285eefe 100644
--- a/src/gallium/drivers/vc4/vc4_context.h
+++ b/src/gallium/drivers/vc4/vc4_context.h
@@ -67,7 +67,9 @@
 #define VC4_DIRTY_CLIP          (1 << 20)
 #define VC4_DIRTY_UNCOMPILED_VS (1 << 21)
 #define VC4_DIRTY_UNCOMPILED_FS (1 << 22)
-#define VC4_DIRTY_COMPILED_FS   (1 << 24)
+#define VC4_DIRTY_COMPILED_CS   (1 << 23)
+#define VC4_DIRTY_COMPILED_VS   (1 << 24)
+#define VC4_DIRTY_COMPILED_FS   (1 << 25)
 
 struct vc4_sampler_view {
         struct pipe_sampler_view base;
@@ -132,6 +134,12 @@ struct vc4_compiled_shader {
         struct vc4_ubo_range *ubo_ranges;
         uint32_t num_ubo_ranges;
         uint32_t ubo_size;
+        /**
+         * VC4_DIRTY_* flags that, when set in vc4->dirty, mean that the
+         * uniforms have to be rewritten (and therefore the shader state
+         * reemitted).
+         */
+        uint32_t uniform_dirty_bits;
 
         /** bitmask of which inputs are color inputs, for flat shade handling. */
         uint32_t color_inputs;
@@ -249,6 +257,11 @@ struct vc4_context {
          */
         bool draw_call_queued;
 
+        /** Maximum index buffer valid for the current shader_rec. */
+        uint32_t max_index;
+        /** Last index bias baked into the current shader_rec. */
+        uint32_t last_index_bias;
+
         struct primconvert_context *primconvert;
 
         struct hash_table *fs_cache, *vs_cache;
@@ -360,6 +373,7 @@ void vc4_simulator_init(struct vc4_screen *screen);
 int vc4_simulator_flush(struct vc4_context *vc4,
                         struct drm_vc4_submit_cl *args);
 
+void vc4_set_shader_uniform_dirty_flags(struct vc4_compiled_shader *shader);
 void vc4_write_uniforms(struct vc4_context *vc4,
                         struct vc4_compiled_shader *shader,
                         struct vc4_constbuf_stateobj *cb,