draw: Only run prepare when state, prim and opt changes

In bad applications like ipers which does a lot of draw calls with
no state changes this helps to greatly reduce time spent in prepare.
In ipers around 7% of CPU was spent in various prepare functions,
after this commit no prepare function show on the profile.

This commit also has the added benefit of now grouping all pipelined
drawing into a single draw call if the driver uses vbuf_render.

Reviewed-by: Stéphane Marchesin <[email protected]>
Tested-by: Stéphane Marchesin <[email protected]>

5 files changed, 66 insertions, 10 deletions

diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index 4ce44455084..3c0b1aa39c9 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -355,6 +355,10 @@ draw_set_vertex_elements(struct draw_context *draw,
 {
    assert(count <= PIPE_MAX_ATTRIBS);
 
+   /* We could improve this by only flushing the frontend and the fetch part
+    * of the middle. This would avoid recalculating the emit keys.*/
+   draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+
    memcpy(draw->pt.vertex_element, elements, count * sizeof(elements[0]));
    draw->pt.nr_vertex_elements = count;
 }
@@ -654,6 +658,8 @@ void draw_do_flush( struct draw_context *draw, unsigned flags )
 
       draw_pipeline_flush( draw, flags );
 
+      draw_pt_flush( draw, flags );
+
       draw->flushing = FALSE;
    }
 }
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 1a0286de891..c3eca973e92 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -63,6 +63,7 @@ struct draw_stage;
 struct vbuf_render;
 struct tgsi_exec_machine;
 struct tgsi_sampler;
+struct draw_pt_front_end;
 
 
 /**
@@ -137,6 +138,12 @@ struct draw_context
    /* Support prototype passthrough path:
     */
    struct {
+      /* Current active frontend */
+      struct draw_pt_front_end *frontend;
+      unsigned prim;
+      unsigned opt;
+      unsigned eltSize; /* saved eltSize for flushing */
+
       struct {
          struct draw_pt_middle_end *fetch_emit;
          struct draw_pt_middle_end *fetch_shade_emit;
@@ -391,6 +398,7 @@ void draw_remove_extra_vertex_attribs(struct draw_context *draw);
 boolean draw_pt_init( struct draw_context *draw );
 void draw_pt_destroy( struct draw_context *draw );
 void draw_pt_reset_vertex_ids( struct draw_context *draw );
+void draw_pt_flush( struct draw_context *draw, unsigned flags );
 
 
 /*******************************************************************************
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index 9a017fd0567..025d53989e4 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -52,7 +52,7 @@ DEBUG_GET_ONCE_BOOL_OPTION(draw_no_fse, "DRAW_NO_FSE", FALSE)
  *     - backend  -- the vbuf_render provided by the driver.
  */
 static boolean
-draw_pt_arrays(struct draw_context *draw, 
+draw_pt_arrays(struct draw_context *draw,
                unsigned prim,
                unsigned start, 
                unsigned count)
@@ -106,17 +106,56 @@ draw_pt_arrays(struct draw_context *draw,
          middle = draw->pt.middle.general;
    }
 
-   frontend = draw->pt.front.vsplit;
+   frontend = draw->pt.frontend;
+
+   if (frontend ) {
+      if (draw->pt.prim != prim || draw->pt.opt != opt) {
+         /* In certain conditions switching primitives requires us to flush
+          * and validate the different stages. One example is when smooth
+          * lines are active but first drawn with triangles and then with
+          * lines.
+          */
+         draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE );
+         frontend = NULL;
+      } else if (draw->pt.eltSize != draw->pt.user.eltSize) {
+         /* Flush draw state if eltSize changed.
+          * This could be improved so only the frontend is flushed since it
+          * converts all indices to ushorts and the fetch part of the middle
+          * always perpares both linear and indexed.
+          */
+         frontend->flush( frontend, DRAW_FLUSH_STATE_CHANGE );
+         frontend = NULL;
+      }
+   }
 
-   frontend->prepare( frontend, prim, middle, opt );
+   if (!frontend) {
+      frontend = draw->pt.front.vsplit;
 
-   frontend->run(frontend, start, count);
+      frontend->prepare( frontend, prim, middle, opt );
 
-   frontend->finish( frontend );
+      draw->pt.frontend = frontend;
+      draw->pt.eltSize = draw->pt.user.eltSize;
+      draw->pt.prim = prim;
+      draw->pt.opt = opt;
+   }
+
+   frontend->run( frontend, start, count );
 
    return TRUE;
 }
 
+void draw_pt_flush( struct draw_context *draw, unsigned flags )
+{
+   if (draw->pt.frontend) {
+      draw->pt.frontend->flush( draw->pt.frontend, flags );
+
+      /* don't prepare if we only are flushing the backend */
+      if (!(flags & DRAW_FLUSH_BACKEND))
+         draw->pt.frontend = NULL;
+   }
+}
+
+
 
 boolean draw_pt_init( struct draw_context *draw )
 {
diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h
index 9a45845521a..2c2efdc1c59 100644
--- a/src/gallium/auxiliary/draw/draw_pt.h
+++ b/src/gallium/auxiliary/draw/draw_pt.h
@@ -73,7 +73,7 @@ struct draw_pt_front_end {
                 unsigned start,
                 unsigned count );
 
-   void (*finish)( struct draw_pt_front_end * );
+   void (*flush)( struct draw_pt_front_end *, unsigned flags );
    void (*destroy)( struct draw_pt_front_end * );
 };
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_vsplit.c b/src/gallium/auxiliary/draw/draw_pt_vsplit.c
index c19dcd9e1f7..0fed057efbe 100644
--- a/src/gallium/auxiliary/draw/draw_pt_vsplit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_vsplit.c
@@ -178,11 +178,14 @@ static void vsplit_prepare(struct draw_pt_front_end *frontend,
 }
 
 
-static void vsplit_finish(struct draw_pt_front_end *frontend)
+static void vsplit_flush(struct draw_pt_front_end *frontend, unsigned flags)
 {
    struct vsplit_frontend *vsplit = (struct vsplit_frontend *) frontend;
-   vsplit->middle->finish(vsplit->middle);
-   vsplit->middle = NULL;
+
+   if (!(flags & DRAW_FLUSH_BACKEND)) {
+      vsplit->middle->finish(vsplit->middle);
+      vsplit->middle = NULL;
+   }
 }
 
 
@@ -202,7 +205,7 @@ struct draw_pt_front_end *draw_pt_vsplit(struct draw_context *draw)
 
    vsplit->base.prepare = vsplit_prepare;
    vsplit->base.run     = NULL;
-   vsplit->base.finish  = vsplit_finish;
+   vsplit->base.flush   = vsplit_flush;
    vsplit->base.destroy = vsplit_destroy;
    vsplit->draw = draw;