Remove debug, fix up culling for swz triangles

7 files changed, 123 insertions, 69 deletions

diff --git a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c
index b23dd914e5..21d10dd640 100644
--- a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c
@@ -104,22 +104,6 @@ static void dump(struct intel_context *intel,
 }
 
 
-const GLuint foo[] = {
-0x7d800003, 0x00000000, 0x00000000, 0x001f003f, 0x00000000, 0x7d040ff7, 0x08300000, 0x04040000,
-0xffffffff, 0x00000000, 0x00902440, 0x00000002, 0x00000006, 0x00000000, 0x7d071202, 0x2c20002b,
- 0x2c201003, 0x00000006, 0x7f280005, 0x43000000, 0x42000000, 0x00000000, 0x42000000, 0x00000000,
- 0x00000000, 0x7f820000, 0x00010000, 0xffff0002, 0x05000000, 0x00000000, 0x00000000, 0x00000000
-};
-
-static void dump_foo( struct intel_context *intel )
-{
-   GLuint offset = 0xeabcf000;
-   _mesa_printf("FOO\n");
-   dump(intel, foo, offset, offset + sizeof(foo), 0);
-   _mesa_printf("END_FOO\n");
-}
-
-
 static void
 intel_dump_batchbuffer(struct intel_batchbuffer *batch, 
 		       GLubyte *batch_map,
@@ -244,9 +228,6 @@ intel_batchbuffer_alloc(struct intel_context *intel)
 
    batch->max_relocs = page_size / sizeof(struct buffer_reloc);
 
-   //dump_foo(intel);
-   (void) dump_foo;
-   
    batch->intel = intel;
 
    driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1,
@@ -449,8 +430,9 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch,
 
    if (used == 0)
       return batch->last_fence;
-
-   _mesa_printf("%s used %d relocs: %d\n", __FUNCTION__, used, batch->nr_relocs);
+   
+   if (0) 
+      _mesa_printf("%s used %d relocs: %d\n", __FUNCTION__, used, batch->nr_relocs);
 
    assert(used < SEGMENT_SZ);
 
diff --git a/src/mesa/drivers/dri/i915tex/intel_cmdstream.c b/src/mesa/drivers/dri/i915tex/intel_cmdstream.c
index 57de58cbeb..48068dd766 100644
--- a/src/mesa/drivers/dri/i915tex/intel_cmdstream.c
+++ b/src/mesa/drivers/dri/i915tex/intel_cmdstream.c
@@ -58,8 +58,9 @@ void intel_cmdstream_use_batch_range( struct intel_context *intel,
  */
 void intel_cmdstream_reset( struct intel_context *intel )
 {
-   _mesa_printf("%s used %x of %x\n", __FUNCTION__, 
-		intel->cmdstream.used, intel->cmdstream.size);
+   if (0) 
+      _mesa_printf("%s used %x of %x\n", __FUNCTION__, 
+		   intel->cmdstream.used, intel->cmdstream.size);
 
    intel->cmdstream.map = intel->batch->map + intel->cmdstream.offset;
    intel->cmdstream.used = 0;
diff --git a/src/mesa/drivers/dri/i915tex/intel_frame_tracker.c b/src/mesa/drivers/dri/i915tex/intel_frame_tracker.c
index 072efee74b..71c94fe2d0 100644
--- a/src/mesa/drivers/dri/i915tex/intel_frame_tracker.c
+++ b/src/mesa/drivers/dri/i915tex/intel_frame_tracker.c
@@ -276,9 +276,9 @@ void intel_frame_set_mode( struct intel_frame_tracker *ft,
    if (ft->mode == new_mode) 
       return;
 
-   _mesa_printf("transiton %s -> %s\n",
-		mode_name[ft->mode],
-		mode_name[new_mode]);
+   if (0) _mesa_printf("transiton %s -> %s\n",
+		       mode_name[ft->mode],
+		       mode_name[new_mode]);
 
    switch (ft->mode) {
    case INTEL_FT_SWAP_BUFFERS:
diff --git a/src/mesa/drivers/dri/i915tex/intel_swrast.c b/src/mesa/drivers/dri/i915tex/intel_swrast.c
index 2819bb6566..9fa658151b 100644
--- a/src/mesa/drivers/dri/i915tex/intel_swrast.c
+++ b/src/mesa/drivers/dri/i915tex/intel_swrast.c
@@ -82,9 +82,6 @@ static void *swrender_allocate_vertices( struct intel_render *render,
 
    assert(vertex_size == swrender->vf->vertex_stride);
 
-   _mesa_printf("%s %d %d %p\n", __FUNCTION__, vertex_size, nr_vertices,
-		swrender->hw_verts); 
-
    return swrender->hw_verts;
 }
 
@@ -246,7 +243,6 @@ static void swrender_draw_prim( struct intel_render *render,
 
    case GL_LINE_STRIP:
       for (i = 0; i+1 < nr; i++) {
-	 _mesa_printf("line %d %d\n", i, i+1);
 	 line( swrender, 
 	       get_vertex(swrender, start+i),
 	       get_vertex(swrender, start+i+1) );
@@ -408,7 +404,7 @@ static void swrender_start_render( struct intel_render *render,
    struct swrast_render *swrender = swrast_render( render );
    struct intel_context *intel = swrender->intel;
 
-   _mesa_printf("%s\n", __FUNCTION__);
+   //_mesa_printf("%s\n", __FUNCTION__);
 
    /* Wait for pending flip.
     */
diff --git a/src/mesa/drivers/dri/i915tex/intel_swz.c b/src/mesa/drivers/dri/i915tex/intel_swz.c
index 7641261d8e..7ce07904c9 100644
--- a/src/mesa/drivers/dri/i915tex/intel_swz.c
+++ b/src/mesa/drivers/dri/i915tex/intel_swz.c
@@ -119,7 +119,6 @@ static void init_zones( struct swz_render *swz )
    space = intel->vtbl.get_state_emit_size( intel, 
 					    swz->initial_driver_flags );
 
-   _mesa_printf("draw origin: %d,%d\n", intel->drawX, intel->drawY);
 
    for (i = y = 0; y < swz->zone_height; y++)
    {
@@ -189,7 +188,7 @@ static void swz_start_render( struct intel_render *render,
    swz->nr_zones = swz->zone_width * swz->zone_height;
    assert(swz->nr_zones < MAX_ZONES);
    
-   if (1) _mesa_printf("%s %dx%d --> %dx%d\n", 
+   if (0) _mesa_printf("%s %dx%d --> %dx%d\n", 
 		       __FUNCTION__,
 		       intel->driDrawable->w,
 		       intel->driDrawable->h,
@@ -295,7 +294,7 @@ static void swz_flush( struct intel_render *render,
    GLuint i = 0;
    GLuint x, y;
 
-   _mesa_printf("%s finished: %d\n", __FUNCTION__, finished);
+   if (0) _mesa_printf("%s finished: %d\n", __FUNCTION__, finished);
 
    if (swz->started_binning) 
    {
@@ -432,7 +431,6 @@ static void swz_clear( struct intel_render *render,
    }
    else 
    {
-      _mesa_printf("%d, %d\n", x2, y2);
       swz_clear_rect( render, mask, x1, y1, x2, y2 );
    }
 
@@ -443,7 +441,6 @@ static void swz_clear( struct intel_render *render,
 static void swz_destroy_context( struct intel_render *render )
 {
    struct swz_render *swz = swz_render( render );
-   _mesa_printf("%s\n", __FUNCTION__);
 
    _mesa_free(swz->last_driver_state);
    _mesa_free(swz->initial_driver_state);
diff --git a/src/mesa/drivers/dri/i915tex/intel_swz.h b/src/mesa/drivers/dri/i915tex/intel_swz.h
index d6dd447c57..7a875005f3 100644
--- a/src/mesa/drivers/dri/i915tex/intel_swz.h
+++ b/src/mesa/drivers/dri/i915tex/intel_swz.h
@@ -80,6 +80,7 @@ struct swz_render {
    GLuint vbo_offset;
 
    GLfloat xoff, yoff;
+   GLfloat cull;
 
    struct intel_hw_dirty reset_state;
 
@@ -330,7 +331,7 @@ static INLINE void zone_loadreg_imm( struct swz_zone *zone,
 {
    GLuint *out = zone_get_dwords( zone, 3 );
 
-   _mesa_printf("zone_loadreg_imm: %x / %08x\n", reg, value);
+   //_mesa_printf("zone_loadreg_imm: %x / %08x\n", reg, value);
 
    out[0] = MI_LOAD_REGISTER_IMM;
    out[1] = reg;
@@ -354,7 +355,6 @@ static INLINE void zone_get_space( struct swz_render *swz,
    GLubyte *newptr = intel_cmdstream_alloc_block( swz->intel );
 
    if (newptr == NULL) {
-      _mesa_printf("FLUSHING ****************\n");
       intel_frame_flush_and_restart( swz->intel->ft );
    }
    else {
diff --git a/src/mesa/drivers/dri/i915tex/intel_swz_prims.c b/src/mesa/drivers/dri/i915tex/intel_swz_prims.c
index 63e9fd6156..a84399eaa1 100644
--- a/src/mesa/drivers/dri/i915tex/intel_swz_prims.c
+++ b/src/mesa/drivers/dri/i915tex/intel_swz_prims.c
@@ -65,6 +65,20 @@ static void invalidate_bins( struct swz_render *swz )
 
    swz->draws++;
 
+   switch(intel->draw_state.cull_mode) {
+   case WINDING_NONE:
+   case WINDING_BOTH:
+      swz->cull = 0;
+      break;
+   case WINDING_CW:
+      swz->cull = -1;
+      break;
+   case WINDING_CCW:
+      swz->cull = 1;
+      break;
+   }
+      
+
    {
       /* XXX: only want to do this once per VB, not once per prim...
        */
@@ -174,25 +188,85 @@ static void tri( struct swz_render *swz,
    i1 += swz->vbo_offset;
    i2 += swz->vbo_offset;
 
-#if 0
-   GLfloat ex = v0[0] - v2[0];
-   GLfloat ey = v0[1] - v2[1];
-   GLfloat fx = v1[0] - v2[0];
-   GLfloat fy = v1[1] - v2[1];
+   x0 = v0[0];
+   x1 = v0[0];
+   y0 = v0[1];
+   y1 = v0[1];
+
+   if (x0 > v1[0]) x0 = v1[0];
+   if (x0 > v2[0]) x0 = v2[0];
+   if (y0 > v1[1]) y0 = v1[1];
+   if (y0 > v2[1]) y0 = v2[1];
+
+   if (x1 < v1[0]) x1 = v1[0];
+   if (x1 < v2[0]) x1 = v2[0];
+   if (y1 < v1[1]) y1 = v1[1];
+   if (y1 < v2[1]) y1 = v2[1];
+
+   zone_x0 = x0 + swz->xoff;
+   zone_x1 = x1 + swz->xoff;
+   zone_y0 = y0 + swz->yoff;
+   zone_y1 = y1 + swz->yoff;
+
+   zone_x0 /= ZONE_WIDTH;
+   zone_x1 /= ZONE_WIDTH;
+   zone_y0 /= ZONE_HEIGHT;
+   zone_y1 /= ZONE_HEIGHT;
+
+   if (zone_x0 < 0) zone_x0 = 0;
+   if (zone_y0 < 0) zone_y0 = 0;
+   if (zone_x1 >= swz->zone_width) zone_x1 = swz->zone_width-1;
+   if (zone_y1 >= swz->zone_height) zone_y1 = swz->zone_height-1;
    
-   GLfloat det = ex * fy - ey * fx;
-
-   if (det >= 0) 
-      return;
-#endif
-
-   /* Calculate bounds: NOTE: reading back from the vbo - must be
-    * declared with appropriate flags.
-    *
-    * Almost all of the slowdown of swz relative to other rendering is
-    * attributable to this calculation.  Need an optimized sse version
-    * to get performance back.
+
+   if (0) _mesa_printf("tri (%f..%f)x(%f..%f) --> (%d..%d)x(%d..%d)\n", 
+		       x0, x1, y0, y1,
+		       zone_x0, zone_x1, zone_y0, zone_y1 );
+
+
+   /* Emit to each zone:
     */
+   for (y = zone_y0; y <= zone_y1; y++) {
+      struct swz_zone *zone = &swz->zone[y * swz->zone_width + zone_x0];
+
+      for (x = zone_x0; x <= zone_x1; x++, zone++) {
+	 zone_update_state(swz, zone, ZONE_TRIS, ZONE_PRIM_SPACE );
+	 zone_emit_tri(zone, i0, i1, i2);
+	 ASSERT(intel_cmdstream_space(zone->ptr) >= ZONE_WRAP_SPACE);
+      }
+   }
+}
+
+
+static void cull_tri( struct swz_render *swz,
+		      GLuint i0, 
+		      GLuint i1, 
+		      GLuint i2 )
+{
+   GLfloat x0, x1, y0, y1;
+   GLint zone_x0, zone_x1, zone_y0, zone_y1;
+   GLint x,y;
+
+   const GLfloat *v0 = get_vertex(swz, i0);
+   const GLfloat *v1 = get_vertex(swz, i1);
+   const GLfloat *v2 = get_vertex(swz, i2);
+
+   i0 += swz->vbo_offset;
+   i1 += swz->vbo_offset;
+   i2 += swz->vbo_offset;
+
+   {
+      GLfloat ex = v0[0] - v2[0];
+      GLfloat ey = v0[1] - v2[1];
+      GLfloat fx = v1[0] - v2[0];
+      GLfloat fy = v1[1] - v2[1];
+   
+      GLfloat det = ex * fy - ey * fx;
+
+      if (det * swz->cull > 0) 
+	 return;
+   }
+
    x0 = v0[0];
    x1 = v0[0];
    y0 = v0[1];
@@ -377,7 +451,7 @@ void swz_clear_rect( struct intel_render *render,
    GLint zone_x0, zone_x1, zone_y0, zone_y1;
    GLint x, y;
 
-   _mesa_printf("%s %d..%d %d..%d\n", __FUNCTION__, x0, x1, y0, y1);
+   if (0) _mesa_printf("%s %d..%d %d..%d\n", __FUNCTION__, x0, x1, y0, y1);
    
    assert( swz->started_binning );
 
@@ -417,7 +491,7 @@ void swz_zone_init( struct intel_render *render,
    GLint zone_x0, zone_x1, zone_y0, zone_y1;
    GLint x, y;
 
-   _mesa_printf("%s %d..%d %d..%d\n", __FUNCTION__, x0, x1, y0, y1);
+   if (0) _mesa_printf("%s %d..%d %d..%d\n", __FUNCTION__, x0, x1, y0, y1);
 
    assert( swz->started_binning );
 
@@ -446,7 +520,8 @@ void swz_zone_init( struct intel_render *render,
 
 
 
-
+typedef void (*trifunc)( struct swz_render *swz,
+			 GLuint, GLuint, GLuint );
 
 
 static void swz_draw_prim( struct intel_render *render,
@@ -459,6 +534,8 @@ static void swz_draw_prim( struct intel_render *render,
 
    invalidate_bins( swz );
 
+   trifunc do_tri = swz->cull ? cull_tri : tri;
+
    switch (swz->prim) {
    case GL_POINTS:
       for (i = 0; i < nr; i++) {
@@ -477,7 +554,6 @@ static void swz_draw_prim( struct intel_render *render,
 
    case GL_LINE_STRIP:
       for (i = 0; i+1 < nr; i++) {
-	 _mesa_printf("line %d %d\n", i, i+1);
 	 line( swz, 
 	       start+i,
 	       start+i+1 );
@@ -486,7 +562,7 @@ static void swz_draw_prim( struct intel_render *render,
 
    case GL_TRIANGLES:
       for (i = 0; i+2 < nr; i += 3) {
-	 tri( swz, 
+	 do_tri( swz, 
 	      start+i,
 	      start+i+1,
 	      start+i+2 );
@@ -496,12 +572,12 @@ static void swz_draw_prim( struct intel_render *render,
    case GL_TRIANGLE_STRIP:
       for (i = 0; i+2 < nr; i++) {
 	 if (i & 1) 
-	    tri( swz, 
+	    do_tri( swz, 
 		 start+i+1,
 		 start+i+0,
 		 start+i+2 );
 	 else
-	    tri( swz, 
+	    do_tri( swz, 
 		 start+i+0,
 		 start+i+1,
 		 start+i+2 );
@@ -510,7 +586,7 @@ static void swz_draw_prim( struct intel_render *render,
 
    case GL_TRIANGLE_FAN:
       for (i = 0; i+2 < nr; i++) {
-	 tri( swz, 
+	 do_tri( swz, 
 	      start+0,
 	      start+i+1,
 	      start+i+2 );
@@ -519,7 +595,7 @@ static void swz_draw_prim( struct intel_render *render,
 
    case GL_POLYGON:
       for (i = 0; i+2 < nr; i++) {
-	 tri( swz, 
+	 do_tri( swz, 
 	      start+i+1,
 	      start+i+2,
 	      start+0 );
@@ -543,6 +619,8 @@ static void swz_draw_indexed_prim( struct intel_render *render,
 
    invalidate_bins( swz );
 
+   trifunc do_tri = swz->cull ? cull_tri : tri;
+
    switch (swz->prim) {
    case GL_POINTS:
       for (i = 0; i < nr; i++) {
@@ -569,7 +647,7 @@ static void swz_draw_indexed_prim( struct intel_render *render,
 
    case GL_TRIANGLES:
       for (i = 0; i+2 < nr; i += 3) {
-	 tri( swz, 
+	 do_tri( swz, 
 	      indices[i],
 	      indices[i+1],
 	      indices[i+2] );
@@ -579,12 +657,12 @@ static void swz_draw_indexed_prim( struct intel_render *render,
    case GL_TRIANGLE_STRIP:
       for (i = 0; i+2 < nr; i++) {
 	 if (i & 1) 
-	    tri( swz, 
+	    do_tri( swz, 
 		 indices[i+1],
 		 indices[i+0],
 		 indices[i+2] );
 	 else
-	    tri( swz, 
+	    do_tri( swz, 
 		 indices[i+0],
 		 indices[i+1],
 		 indices[i+2] );
@@ -593,7 +671,7 @@ static void swz_draw_indexed_prim( struct intel_render *render,
 
    case GL_TRIANGLE_FAN:
       for (i = 0; i+2 < nr; i++) {
-	 tri( swz, 
+	 do_tri( swz, 
 	      indices[0],
 	      indices[i+1],
 	      indices[i+2] );
@@ -602,7 +680,7 @@ static void swz_draw_indexed_prim( struct intel_render *render,
 
    case GL_POLYGON:
       for (i = 0; i+2 < nr; i++) {
-	 tri( swz, 
+	 do_tri( swz, 
 	      indices[i+1],
 	      indices[i+2],
 	      indices[0] );