vbo: Avoid extra validation of DrawElements.

This saves mapping the index buffer to get a bounds on the indices that
drivers just drop on the floor in the VBO case (cache win), saves a bonus
walk of the indices in the CheckArrayBounds case, and other miscellaneous
validation.  On intel it's a particularly a large win (50-100% in my app)
because even though we let the indices stay in both CPU and GPU caches, we
still end up waiting for the GPU to be done with the buffer before reading
from it.

Drivers that want the min/max_index fields must now check index_bounds_valid
and use vbo_get_minmax_index before using them.

3 files changed, 27 insertions, 3 deletions

diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index f69b122046..f2771cde09 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -81,7 +81,7 @@ _tnl_CreateContext( GLcontext *ctx )
    tnl->nr_blocks = 0;
 
    /* plug in the VBO drawing function */
-   vbo_set_draw_func(ctx, _tnl_draw_prims);
+   vbo_set_draw_func(ctx, _tnl_vbo_draw_prims);
 
    _math_init_transformation();
    _math_init_translate();
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
index 2ec65d5323..c64c2c2077 100644
--- a/src/mesa/tnl/t_draw.c
+++ b/src/mesa/tnl/t_draw.c
@@ -360,6 +360,20 @@ static void unmap_vbos( GLcontext *ctx,
 }
 
 
+void _tnl_vbo_draw_prims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLboolean index_bounds_valid,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+   if (!index_bounds_valid)
+      vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+
+   _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
 
 /* This is the main entrypoint into the slimmed-down software tnl
  * module.  In a regular swtnl driver, this can be plugged straight
@@ -393,7 +407,7 @@ void _tnl_draw_prims( GLcontext *ctx,
        */
       vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, 
 			min_index, max_index,
-			_tnl_draw_prims );
+			_tnl_vbo_draw_prims );
       return;
    }
    else if (max_index > max) {
@@ -411,7 +425,7 @@ void _tnl_draw_prims( GLcontext *ctx,
        */
       vbo_split_prims( ctx, arrays, prim, nr_prims, ib, 
 		       0, max_index,
-		       _tnl_draw_prims,
+		       _tnl_vbo_draw_prims,
 		       &limits );
    }
    else {
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index 4d628aa9a6..9c66d3b019 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -81,6 +81,16 @@ _tnl_draw_prims( GLcontext *ctx,
 		 GLuint min_index,
 		 GLuint max_index);
 
+void
+_tnl_vbo_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
+		     GLuint min_index,
+		     GLuint max_index);
+
 extern void
 _mesa_load_tracked_matrices(GLcontext *ctx);