Merge branch 'mesa_7_5_branch'

[mesa.git] / src / mesa / drivers / dri / i965 / brw_draw.c

diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 0b6499943772a8e32a2291050642cbf4e9cdcece..682094ff139aaa1025c99fbf63a3ceaceb71133f 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -127,6 +127,7 @@ static void brw_emit_prim(struct brw_context *brw,
                          uint32_t hw_prim)
 {
    struct brw_3d_primitive prim_packet;
+   struct intel_context *intel = &brw->intel;
 
    if (INTEL_DEBUG & DEBUG_PRIMS)
       _mesa_printf("PRIM: %s %d %d\n", _mesa_lookup_enum_by_nr(prim->mode), 
@@ -140,16 +141,35 @@ static void brw_emit_prim(struct brw_context *brw,
 
    prim_packet.verts_per_instance = trim(prim->mode, prim->count);
    prim_packet.start_vert_location = prim->start;
+   if (prim->indexed)
+      prim_packet.start_vert_location += brw->ib.start_vertex_offset;
    prim_packet.instance_count = 1;
    prim_packet.start_instance_location = 0;
    prim_packet.base_vert_location = 0;
 
    /* Can't wrap here, since we rely on the validated state. */
    brw->no_batch_wrap = GL_TRUE;
+
+   /* If we're set to always flush, do it before and after the primitive emit.
+    * We want to catch both missed flushes that hurt instruction/state cache
+    * and missed flushes of the render cache as it heads to other parts of
+    * the besides the draw code.
+    */
+   if (intel->always_flush_cache) {
+      BEGIN_BATCH(1, IGNORE_CLIPRECTS);
+      OUT_BATCH(intel->vtbl.flush_cmd());
+      ADVANCE_BATCH();
+   }
    if (prim_packet.verts_per_instance) {
       intel_batchbuffer_data( brw->intel.batch, &prim_packet,
                              sizeof(prim_packet), LOOP_CLIPRECTS);
    }
+   if (intel->always_flush_cache) {
+      BEGIN_BATCH(1, IGNORE_CLIPRECTS);
+      OUT_BATCH(intel->vtbl.flush_cmd());
+      ADVANCE_BATCH();
+   }
+
    brw->no_batch_wrap = GL_FALSE;
 }
 
@@ -169,19 +189,13 @@ static void brw_merge_inputs( struct brw_context *brw,
       brw->vb.inputs[i].glarray = arrays[i];
 
       if (arrays[i]->StrideB != 0)
-        brw->vb.info.varying |= 1 << i;
-
         brw->vb.info.sizes[i/16] |= (brw->vb.inputs[i].glarray->Size - 1) <<
            ((i%16) * 2);
    }
 
-   /* Raise statechanges if input sizes and varying have changed: 
-    */
+   /* Raise statechanges if input sizes have changed. */
    if (memcmp(brw->vb.info.sizes, old.sizes, sizeof(old.sizes)) != 0)
       brw->state.dirty.brw |= BRW_NEW_INPUT_DIMENSIONS;
-
-   if (brw->vb.info.varying != old.varying)
-      brw->state.dirty.brw |= BRW_NEW_INPUT_VARYING;
 }
 
 /* XXX: could split the primitive list to fallback only on the
@@ -393,9 +407,13 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
       retval = GL_TRUE;
    }
 
+   if (intel->always_flush_batch)
+      intel_batchbuffer_flush(intel->batch);
  out:
    UNLOCK_HARDWARE(intel);
 
+   brw_state_cache_check_size(brw);
+
    if (warn)
       fprintf(stderr, "i965: Single primitive emit potentially exceeded "
              "available aperture space\n");
@@ -406,54 +424,31 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    return retval;
 }
 
-static GLboolean brw_need_rebase( GLcontext *ctx,
-                                 const struct gl_client_array *arrays[],
-                                 const struct _mesa_index_buffer *ib,
-                                 GLuint min_index )
-{
-   if (min_index == 0) 
-      return GL_FALSE;
-
-   if (ib) {
-      if (!vbo_all_varyings_in_vbos(arrays))
-        return GL_TRUE;
-      else
-        return GL_FALSE;
-   }
-   else {
-      /* Hmm.  This isn't quite what I wanted.  BRW can actually
-       * handle the mixed case well enough that we shouldn't need to
-       * rebase.  However, it's probably not very common, nor hugely
-       * expensive to do it this way:
-       */
-      if (!vbo_all_varyings_in_vbos(arrays))
-        return GL_TRUE;
-      else
-        return GL_FALSE;
-   }
-}
-                                 
-
 void brw_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 )
 {
    GLboolean retval;
 
-   /* Decide if we want to rebase.  If so we end up recursing once
-    * only into this function.
-    */
-   if (brw_need_rebase( ctx, arrays, ib, min_index )) {
-      vbo_rebase_prims( ctx, arrays, 
-                       prim, nr_prims, 
-                       ib, min_index, max_index, 
-                       brw_draw_prims );
-      
-      return;
+   if (!vbo_all_varyings_in_vbos(arrays)) {
+      if (!index_bounds_valid)
+        vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+
+      /* Decide if we want to rebase.  If so we end up recursing once
+       * only into this function.
+       */
+      if (min_index != 0) {
+        vbo_rebase_prims(ctx, arrays,
+                         prim, nr_prims,
+                         ib, min_index, max_index,
+                         brw_draw_prims );
+        return;
+      }
    }
 
    /* Make a first attempt at drawing: