i965: Flush pipeline on EndTransformFeedback.

A common use case for transform feedback is to perform one draw
operation that writes transform feedback output to a buffer, followed
by a second draw operation that consumes that buffer as vertex input.
Since vertex input is consumed at an earlier pipeline stage than
writing transform feedback output, we need to flush the pipeline to
ensure that the transform feedback output is completely written before
the data is consumed.

In an ideal world, we would do some dependency tracking, so that we
would only flush the pipeline if the next draw call was about to
consume data generated by a previous draw call in the same batch.
However, since we don't have that sort of dependency tracking
infrastructure right now, we just unconditionally flush the buffer
every time glEndTransformFeedback() is called.  This will cause a
performance hit compared to the ideal case (since we will sometimes
flush the pipeline unnecessarily), but fortunately the performance hit
will be confined to circumstances where transform feedback is in use.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>

src/mesa/drivers/dri/i965/brw_context.c

@@ -117,6 +117,7 @@ static void brwInitDriverFunctions( struct dd_function_table *functions )
    brw_init_queryobj_functions(functions);
 
    functions->PrepareExecBegin = brwPrepareExecBegin;
+   functions->EndTransformFeedback = brw_end_transform_feedback;
 }
 
 bool

src/mesa/drivers/dri/i965/brw_context.h

@@ -1076,6 +1076,11 @@ brw_update_sol_surface(struct brw_context *brw,
 bool
 brw_fprog_uses_noperspective(const struct gl_fragment_program *fprog);
 
+/* gen6_sol.c */
+void
+brw_end_transform_feedback(struct gl_context *ctx,
+                           struct gl_transform_feedback_object *obj);
+
 
 
 /*======================================================================

src/mesa/drivers/dri/i965/gen6_sol.c

@@ -27,6 +27,7 @@
  */
 
 #include "brw_context.h"
+#include "intel_batchbuffer.h"
 #include "brw_defines.h"
 
 static void
@@ -69,3 +70,18 @@ const struct brw_tracked_state gen6_sol_surface = {
    },
    .emit = gen6_update_sol_surfaces,
 };
+
+void
+brw_end_transform_feedback(struct gl_context *ctx,
+                           struct gl_transform_feedback_object *obj)
+{
+   /* After EndTransformFeedback, it's likely that the client program will try
+    * to draw using the contents of the transform feedback buffer as vertex
+    * input.  In order for this to work, we need to flush the data through at
+    * least the GS stage of the pipeline, and flush out the render cache.  For
+    * simplicity, just do a full flush.
+    */
+   struct brw_context *brw = brw_context(ctx);
+   struct intel_context *intel = &brw->intel;
+   intel_batchbuffer_emit_mi_flush(intel);
+}