Merge branch 'sprite-coord'

20 files changed, 372 insertions, 127 deletions

diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index c339f72fe2..02af4d9280 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -8,6 +8,7 @@ C_SOURCES = \
 	cso_cache/cso_context.c \
 	cso_cache/cso_hash.c \
 	draw/draw_context.c \
+	draw/draw_fs.c \
 	draw/draw_gs.c \
 	draw/draw_pipe.c \
 	draw/draw_pipe_aaline.c \
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
index 144d929f68..48547c4b2c 100644
--- a/src/gallium/auxiliary/SConscript
+++ b/src/gallium/auxiliary/SConscript
@@ -54,6 +54,7 @@ source = [
     'cso_cache/cso_context.c',
     'cso_cache/cso_hash.c',
     'draw/draw_context.c',
+    'draw/draw_fs.c',
     'draw/draw_gs.c',
     'draw/draw_pipe.c',
     'draw/draw_pipe_aaline.c',
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index 937b093479..b07de76a49 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -413,6 +413,42 @@ draw_set_force_passthrough( struct draw_context *draw, boolean enable )
 }
 
 
+
+/**
+ * Allocate an extra vertex/geometry shader vertex attribute.
+ * This is used by some of the optional draw module stages such
+ * as wide_point which may need to allocate additional generic/texcoord
+ * attributes.
+ */
+int
+draw_alloc_extra_vertex_attrib(struct draw_context *draw,
+                               uint semantic_name, uint semantic_index)
+{
+   const int num_outputs = draw_current_shader_outputs(draw);
+   const int n = draw->extra_shader_outputs.num;
+
+   assert(n < Elements(draw->extra_shader_outputs.semantic_name));
+
+   draw->extra_shader_outputs.semantic_name[n] = semantic_name;
+   draw->extra_shader_outputs.semantic_index[n] = semantic_index;
+   draw->extra_shader_outputs.slot[n] = num_outputs + n;
+   draw->extra_shader_outputs.num++;
+
+   return draw->extra_shader_outputs.slot[n];
+}
+
+
+/**
+ * Remove all extra vertex attributes that were allocated with
+ * draw_alloc_extra_vertex_attrib().
+ */
+void
+draw_remove_extra_vertex_attribs(struct draw_context *draw)
+{
+   draw->extra_shader_outputs.num = 0;
+}
+
+
 /**
  * Ask the draw module for the location/slot of the given vertex attribute in
  * a post-transformed vertex.
@@ -446,12 +482,12 @@ draw_find_shader_output(const struct draw_context *draw,
          return i;
    }
 
-   /* XXX there may be more than one extra vertex attrib.
-    * For example, simulated gl_FragCoord and gl_PointCoord.
-    */
-   if (draw->extra_shader_outputs.semantic_name == semantic_name &&
-       draw->extra_shader_outputs.semantic_index == semantic_index) {
-      return draw->extra_shader_outputs.slot;
+   /* Search the extra vertex attributes */
+   for (i = 0; i < draw->extra_shader_outputs.num; i++) {
+      if (draw->extra_shader_outputs.semantic_name[i] == semantic_name &&
+          draw->extra_shader_outputs.semantic_index[i] == semantic_index) {
+         return draw->extra_shader_outputs.slot[i];
+      }
    }
 
    return 0;
@@ -470,16 +506,18 @@ draw_find_shader_output(const struct draw_context *draw,
 uint
 draw_num_shader_outputs(const struct draw_context *draw)
 {
-   uint count = draw->vs.vertex_shader->info.num_outputs;
+   uint count;
 
    /* If a geometry shader is present, its outputs go to the
     * driver, else the vertex shader's outputs.
     */
    if (draw->gs.geometry_shader)
       count = draw->gs.geometry_shader->info.num_outputs;
+   else
+      count = draw->vs.vertex_shader->info.num_outputs;
+
+   count += draw->extra_shader_outputs.num;
 
-   if (draw->extra_shader_outputs.slot > 0)
-      count++;
    return count;
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index 4f0d30123a..1f27cbf488 100644
--- a/src/gallium/auxiliary/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -46,6 +46,7 @@ struct draw_context;
 struct draw_stage;
 struct draw_vertex_shader;
 struct draw_geometry_shader;
+struct draw_fragment_shader;
 struct tgsi_sampler;
 
 #define DRAW_MAX_TEXTURE_LEVELS 13  /* 4K x 4K for now */
@@ -138,6 +139,17 @@ void draw_delete_vertex_shader(struct draw_context *draw,
 
 
 /*
+ * Fragment shader functions
+ */
+struct draw_fragment_shader *
+draw_create_fragment_shader(struct draw_context *draw,
+                            const struct pipe_shader_state *shader);
+void draw_bind_fragment_shader(struct draw_context *draw,
+                               struct draw_fragment_shader *dvs);
+void draw_delete_fragment_shader(struct draw_context *draw,
+                                 struct draw_fragment_shader *dvs);
+
+/*
  * Geometry shader functions
  */
 struct draw_geometry_shader *
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
index eac21110be..d1aba76309 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -688,10 +688,9 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header)
    aaline->tex_slot = draw_current_shader_outputs(draw);
    aaline->pos_slot = draw_current_shader_position_output(draw);;
 
-   /* advertise the extra post-transformed vertex attribute */
-   draw->extra_shader_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-   draw->extra_shader_outputs.semantic_index = aaline->fs->generic_attrib;
-   draw->extra_shader_outputs.slot = aaline->tex_slot;
+   /* allocate the extra post-transformed vertex attribute */
+   (void) draw_alloc_extra_vertex_attrib(draw, TGSI_SEMANTIC_GENERIC,
+                                         aaline->fs->generic_attrib);
 
    /* how many samplers? */
    /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
@@ -744,7 +743,7 @@ aaline_flush(struct draw_stage *stage, unsigned flags)
 
    draw->suspend_flushing = FALSE;
 
-   draw->extra_shader_outputs.slot = 0;
+   draw_remove_extra_vertex_attribs(draw);
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
index d406a86ccb..5ea552f51c 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -701,9 +701,9 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
 
    aapoint->pos_slot = draw_current_shader_position_output(draw);
 
-   draw->extra_shader_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-   draw->extra_shader_outputs.semantic_index = aapoint->fs->generic_attrib;
-   draw->extra_shader_outputs.slot = aapoint->tex_slot;
+   /* allocate the extra post-transformed vertex attribute */
+   (void) draw_alloc_extra_vertex_attrib(draw, TGSI_SEMANTIC_GENERIC,
+                                         aapoint->fs->generic_attrib);
 
    /* find psize slot in post-transform vertex */
    aapoint->psize_slot = -1;
@@ -754,7 +754,7 @@ aapoint_flush(struct draw_stage *stage, unsigned flags)
 
    draw->suspend_flushing = FALSE;
 
-   draw->extra_shader_outputs.slot = 0;
+   draw_remove_extra_vertex_attribs(draw);
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c
index 8b92543987..c575a8ac7c 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_validate.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c
@@ -172,7 +172,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
    wide_lines = (rast->line_width > draw->pipeline.wide_line_threshold
                  && !rast->line_smooth);
 
-   /* drawing large points? */
+   /* drawing large/sprite points (but not AA points)? */
    if (rast->sprite_coord_enable && draw->pipeline.point_sprite)
       wide_points = TRUE;
    else if (rast->point_smooth && draw->pipeline.aapoint)
@@ -207,7 +207,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
       precalc_flat = TRUE;
    }
 
-   if (wide_points || rast->sprite_coord_enable) {
+   if (wide_points) {
       draw->pipeline.wide_point->next = next;
       next = draw->pipeline.wide_point;
    }
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index ee2945c7c9..3646c6a714 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -57,26 +57,24 @@
 #include "util/u_memory.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_shader_tokens.h"
+#include "draw_fs.h"
 #include "draw_vs.h"
 #include "draw_pipe.h"
 
 
 struct widepoint_stage {
-   struct draw_stage stage;
+   struct draw_stage stage;  /**< base class */
 
    float half_point_size;
 
    float xbias;
    float ybias;
 
-   uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS];
-   uint texcoord_enable[PIPE_MAX_SHADER_OUTPUTS];
-   uint num_texcoords;
-   uint texcoord_mode;
+   /** for automatic texcoord generation/replacement */
+   uint num_texcoord_gen;
+   uint texcoord_gen_slot[PIPE_MAX_SHADER_OUTPUTS];
 
    int psize_slot;
-
-   int point_coord_fs_input;  /**< input for pointcoord */
 };
 
 
@@ -96,30 +94,20 @@ widepoint_stage( struct draw_stage *stage )
 static void set_texcoords(const struct widepoint_stage *wide,
                           struct vertex_header *v, const float tc[4])
 {
+   const struct draw_context *draw = wide->stage.draw;
+   const struct pipe_rasterizer_state *rast = draw->rasterizer;
+   const uint texcoord_mode = rast->sprite_coord_mode;
    uint i;
-   for (i = 0; i < wide->num_texcoords; i++) {
-      if (wide->texcoord_enable[i]) {
-         uint j = wide->texcoord_slot[i];
-         v->data[j][0] = tc[0];
-         if (wide->texcoord_mode == PIPE_SPRITE_COORD_LOWER_LEFT)
-            v->data[j][1] = 1.0f - tc[1];
-         else
-            v->data[j][1] = tc[1];
-         v->data[j][2] = tc[2];
-         v->data[j][3] = tc[3];
-      }
-   }
 
-   if (wide->point_coord_fs_input >= 0) {
-      /* put gl_PointCoord into the extra vertex slot */
-      uint slot = wide->stage.draw->extra_shader_outputs.slot;
+   for (i = 0; i < wide->num_texcoord_gen; i++) {
+      const uint slot = wide->texcoord_gen_slot[i];
       v->data[slot][0] = tc[0];
-      if (wide->texcoord_mode == PIPE_SPRITE_COORD_LOWER_LEFT)
+      if (texcoord_mode == PIPE_SPRITE_COORD_LOWER_LEFT)
          v->data[slot][1] = 1.0f - tc[1];
       else
          v->data[slot][1] = tc[1];
-      v->data[slot][2] = 0.0F;
-      v->data[slot][3] = 1.0F;
+      v->data[slot][2] = tc[2];
+      v->data[slot][3] = tc[3];
    }
 }
 
@@ -201,18 +189,9 @@ static void widepoint_point( struct draw_stage *stage,
 }
 
 
-static int
-find_pntc_input_attrib(struct draw_context *draw)
-{
-   /* Scan the fragment program's input decls to find the pointcoord
-    * attribute.  The xy components will store the point coord.
-    */
-   return 0; /* XXX fix this */
-}
-
-
-static void widepoint_first_point( struct draw_stage *stage, 
-			      struct prim_header *header )
+static void
+widepoint_first_point(struct draw_stage *stage, 
+                      struct prim_header *header)
 {
    struct widepoint_stage *wide = widepoint_stage(stage);
    struct draw_context *draw = stage->draw;
@@ -244,31 +223,49 @@ static void widepoint_first_point( struct draw_stage *stage,
       stage->point = draw_pipe_passthrough_point;
    }
 
+   draw_remove_extra_vertex_attribs(draw);
+
    if (rast->point_quad_rasterization) {
-      /* find vertex shader texcoord outputs */
-      const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
-      uint i, j = 0;
-      wide->texcoord_mode = rast->sprite_coord_mode;
-      for (i = 0; i < vs->info.num_outputs; i++) {
-         if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) {
-            wide->texcoord_slot[j] = i;
-            wide->texcoord_enable[j] = (rast->sprite_coord_enable >> j) & 1;
-            j++;
+      const struct draw_fragment_shader *fs = draw->fs.fragment_shader;
+      uint i;
+
+      wide->num_texcoord_gen = 0;
+
+      /* Loop over fragment shader inputs looking for generic inputs
+       * for which bit 'k' in sprite_coord_enable is set.
+       */
+      for (i = 0; i < fs->info.num_inputs; i++) {
+         if (fs->info.input_semantic_name[i] == TGSI_SEMANTIC_GENERIC) {
+            const int generic_index = fs->info.input_semantic_index[i];
+            /* Note that sprite_coord enable is a bitfield of
+             * PIPE_MAX_SHADER_OUTPUTS bits.
+             */
+            if (generic_index < PIPE_MAX_SHADER_OUTPUTS &&
+                (rast->sprite_coord_enable & (1 << generic_index))) {
+               /* OK, this generic attribute needs to be replaced with a
+                * texcoord (see above).
+                */
+               int slot = draw_find_shader_output(draw,
+                                                  TGSI_SEMANTIC_GENERIC,
+                                                  generic_index);
+
+               if (slot > 0) {
+                  /* there's already a post-vertex shader attribute
+                   * for this fragment shader input attribute.
+                   */
+               }
+               else {
+                  /* need to allocate a new post-vertex shader attribute */
+                  slot = draw_alloc_extra_vertex_attrib(draw,
+                                                        TGSI_SEMANTIC_GENERIC,
+                                                        generic_index);
+               }
+
+               /* add this slot to the texcoord-gen list */
+               wide->texcoord_gen_slot[wide->num_texcoord_gen++] = slot;
+            }
          }
       }
-      wide->num_texcoords = j;
-
-      /* find fragment shader PointCoord input */
-      wide->point_coord_fs_input = find_pntc_input_attrib(draw);
-
-      /* setup extra vp output (point coord implemented as a texcoord) */
-      draw->extra_shader_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-      draw->extra_shader_outputs.semantic_index = 0;
-      draw->extra_shader_outputs.slot = draw_current_shader_outputs(draw);
-   }
-   else {
-      wide->point_coord_fs_input = -1;
-      draw->extra_shader_outputs.slot = 0;
    }
 
    wide->psize_slot = -1;
@@ -295,7 +292,8 @@ static void widepoint_flush( struct draw_stage *stage, unsigned flags )
 
    stage->point = widepoint_first_point;
    stage->next->flush( stage->next, flags );
-   stage->draw->extra_shader_outputs.slot = 0;
+
+   draw_remove_extra_vertex_attribs(draw);
 
    /* restore original rasterizer state */
    if (draw->rast_handle) {
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 362f563ba6..d417f825a0 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -250,6 +250,11 @@ struct draw_context
       struct tgsi_sampler **samplers;
    } gs;
 
+   /** Fragment shader state */
+   struct {
+      struct draw_fragment_shader *fragment_shader;
+   } fs;
+
    /** Stream output (vertex feedback) state */
    struct {
       struct pipe_stream_output_state state;
@@ -266,9 +271,10 @@ struct draw_context
    /* If a prim stage introduces new vertex attributes, they'll be stored here
     */
    struct {
-      uint semantic_name;
-      uint semantic_index;
-      int slot;
+      uint num;
+      uint semantic_name[10];
+      uint semantic_index[10];
+      uint slot[10];
    } extra_shader_outputs;
 
    unsigned reduced_prim;
@@ -362,6 +368,11 @@ void draw_gs_destroy( struct draw_context *draw );
 uint draw_current_shader_outputs(const struct draw_context *draw);
 uint draw_current_shader_position_output(const struct draw_context *draw);
 
+int draw_alloc_extra_vertex_attrib(struct draw_context *draw,
+                                   uint semantic_name, uint semantic_index);
+void draw_remove_extra_vertex_attribs(struct draw_context *draw);
+
+
 /*******************************************************************************
  * Vertex processing (was passthrough) code:
  */
diff --git a/src/gallium/docs/source/cso/rasterizer.rst b/src/gallium/docs/source/cso/rasterizer.rst
index ee3419ccfc..d547055096 100644
--- a/src/gallium/docs/source/cso/rasterizer.rst
+++ b/src/gallium/docs/source/cso/rasterizer.rst
@@ -124,17 +124,24 @@ Points
 sprite_coord_enable
 ^^^^^^^^^^^^^^^^^^^
 
-Specifies if a texture unit has its texture coordinates replaced or not. This
-is a packed bitfield containing the enable for all texcoords -- if all bits
-are zero, point sprites are effectively disabled. 
+Controls automatic texture coordinate generation for rendering sprite points.
+
+When bit k in the sprite_coord_enable bitfield is set, then generic
+input k to the fragment shader will get an automatically computed
+texture coordinate.
+
+The texture coordinate will be of the form (s, t, 0, 1) where s varies
+from 0 to 1 from left to right while t varies from 0 to 1 according to
+the state of 'sprite_coord_mode' (see below).
 
 If any bit is set, then point_smooth MUST be disabled (there are no
 round sprites) and point_quad_rasterization MUST be true (sprites are
 always rasterized as quads).  Any mismatch between these states should
 be considered a bug in the state-tracker.
 
-If enabled, the four vertices of the resulting quad will be assigned
-texture coordinates, according to sprite_coord_mode.
+This feature is implemented in the :ref:`Draw` module but may also be
+implemented natively by GPUs or implemented with a geometry shader.
+
 
 sprite_coord_mode
 ^^^^^^^^^^^^^^^^^
@@ -145,6 +152,7 @@ have coordinates (0,0,0,1). For PIPE_SPRITE_COORD_UPPER_LEFT, the upper-left
 vertex will have coordinates (0,0,0,1).
 This state is used by :ref:`Draw` to generate texcoords.
 
+
 point_quad_rasterization
 ^^^^^^^^^^^^^^^^^^^^^^^^
 
diff --git a/src/gallium/docs/source/pipeline.txt b/src/gallium/docs/source/pipeline.txt
new file mode 100644
index 0000000000..fd1fbe9c76
--- /dev/null
+++ b/src/gallium/docs/source/pipeline.txt
@@ -0,0 +1,128 @@
+XXX this could be converted/formatted for Sphinx someday.
+XXX do not use tabs in this file.
+
+
+
+            position                     ]
+            primary/secondary colors     ]
+            generics (normals,           ]
+               texcoords, fog)           ] User vertices / arrays
+            point size                   ]
+            edge flag                    ]
+            primitive ID                 } System-generated values
+            vertex ID                    }
+              | | |
+              V V V
+      +-------------------+
+      |  Vertex shader    |
+      +-------------------+
+              | | |
+              V V V
+            position
+            clip distance
+            generics
+            front/back & primary/secondary colors
+            point size
+            edge flag
+            primitive ID
+              | | |
+              V V V
+      +------------------------+
+      |     Geometry shader    |
+      | (consume vertex ID)    |
+      | (may change prim type) |
+      +------------------------+
+              | | |
+              V V V
+            [...]
+            fb layer
+              | | |
+              V V V
+      +--------------------------+
+      |         Clipper          |
+      | (consume clip distances) |
+      +--------------------------+
+              | | |
+              V V V
+      +-------------------+
+      |  Polygon Culling  |
+      +-------------------+
+              | | |
+              V V V
+      +-----------------------+
+      |    Choose front or    |
+      |    back face color    |
+      | (consume other color) |
+      +-----------------------+
+              | | |
+              V V V
+            [...]
+            primary/secondary colors only
+              | | |
+              V V V
+      +-------------------+
+      |   Polygon Offset  |
+      +-------------------+
+              | | |
+              V V V
+      +----------------------+
+      | Unfilled polygons    |
+      | (consume edge flags) |
+      | (change prim type)   |
+      +----------------------+
+              | | |
+              V V V
+            position
+            generics
+            primary/secondary colors
+            point size
+            primitive ID
+            fb layer
+              | | |
+              V V V
+  +---------------------------------+ 
+  | Optional Draw module helpers    |
+  | * Polygon Stipple               |
+  | * Line Stipple                  |
+  | * Line AA/smooth (as tris)      |
+  | * Wide lines (as tris)          |
+  | * Wide points/sprites (as tris) |
+  | * Point AA/smooth (as tris)     |
+  | (NOTE: these stages may emit    |
+  |  new/extra generic attributes   |
+  |  such as texcoords)             |
+  +---------------------------------+
+              | | |
+              V V V
+            position                     ]
+            generics (+ new/extra ones)  ]
+            primary/secondary colors     ] Software rast vertices
+            point size                   ]
+            primitive ID                 ]
+            fb layer                     ]
+              | | |
+              V V V
+      +---------------------+
+      | Triangle/Line/Point |
+      |    Rasterization    |
+      +---------------------+
+              | | |
+              V V V
+            generic attribs
+            primary/secondary colors
+            primitive ID
+            fragment win coord pos   } System-generated values
+            front/back face flag     }
+              | | |
+              V V V
+      +-------------------+
+      |  Fragment shader  |
+      +-------------------+
+              | | |
+              V V V
+            zero or more colors
+            zero or one Z value
+
+
+NOTE: The instance ID is not shown.  It can be imagined to be a global variable
+accessible to all shader stages.
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
index 6674d281d1..ea7002aafc 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -490,12 +490,14 @@ void
 lp_setup_set_point_state( struct lp_setup_context *setup,
                           float point_size,                          
                           boolean point_size_per_vertex,
-                          uint sprite)
+                          uint sprite_coord_enable,
+                          uint sprite_coord_origin)
 {
    LP_DBG(DEBUG_SETUP, "%s\n", __FUNCTION__);
 
    setup->point_size = point_size;
-   setup->sprite = sprite;
+   setup->sprite_coord_enable = sprite_coord_enable;
+   setup->sprite_coord_origin = sprite_coord_origin;
    setup->point_size_per_vertex = point_size_per_vertex;
 }
 
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.h b/src/gallium/drivers/llvmpipe/lp_setup.h
index b94061b7d4..81ff43f8ad 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.h
+++ b/src/gallium/drivers/llvmpipe/lp_setup.h
@@ -107,7 +107,8 @@ void
 lp_setup_set_point_state( struct lp_setup_context *setup,
                           float point_size,                          
                           boolean point_size_per_vertex,
-                          uint sprite);
+                          uint sprite_coord_enable,
+                          uint sprite_coord_origin);
 
 void
 lp_setup_set_fs_inputs( struct lp_setup_context *setup,
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_context.h b/src/gallium/drivers/llvmpipe/lp_setup_context.h
index 80b356476a..8506ed2dc9 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_context.h
+++ b/src/gallium/drivers/llvmpipe/lp_setup_context.h
@@ -73,7 +73,7 @@ struct lp_setup_context
    uint prim;
    uint vertex_size;
    uint nr_vertices;
-   uint sprite;
+   uint sprite_coord_enable, sprite_coord_origin;
    uint vertex_buffer_size;
    void *vertex_buffer;
 
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_point.c b/src/gallium/drivers/llvmpipe/lp_setup_point.c
index 5538987151..774a3c80da 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_point.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_point.c
@@ -52,26 +52,30 @@ struct point_info {
 /**
  * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
  */
-static void constant_coef( struct lp_setup_context *setup,
-                           struct lp_rast_triangle *point,
-                           unsigned slot,
-                           const float value,
-                           unsigned i )
+static void
+constant_coef(struct lp_setup_context *setup,
+              struct lp_rast_triangle *point,
+              unsigned slot,
+              const float value,
+              unsigned i)
 {
    point->inputs.a0[slot][i] = value;
    point->inputs.dadx[slot][i] = 0.0f;
    point->inputs.dady[slot][i] = 0.0f;
 }
 
-static void perspective_coef( struct lp_setup_context *setup,
-                              struct lp_rast_triangle *point,
-                              const struct point_info *info,
-                              unsigned slot,
-                              unsigned vert_attr,
-                              unsigned i)
+
+static void
+perspective_coef(struct lp_setup_context *setup,
+                 struct lp_rast_triangle *point,
+                 const struct point_info *info,
+                 unsigned slot,
+                 unsigned vert_attr,
+                 unsigned i,
+                 unsigned sprite_coord_origin)
 {
-   if (i == 0) {   
-      float dadx = FIXED_ONE / (float)info->dx12;  
+   if (i == 0) {
+      float dadx = FIXED_ONE / (float)info->dx12;
       float dady =  0.0f;
       point->inputs.dadx[slot][i] = dadx;
       point->inputs.dady[slot][i] = dady;
@@ -79,30 +83,30 @@ static void perspective_coef( struct lp_setup_context *setup,
                                   (dadx * ((float)info->v0[0][0] - setup->pixel_offset) +
                                    dady * ((float)info->v0[0][1] - setup->pixel_offset)));
    }
-
    else if (i == 1) {
-      float dadx =  0.0f; 
-      float dady =  FIXED_ONE / (float)info->dx12;
-   
+      float dadx = 0.0f;
+      float dady = FIXED_ONE / (float)info->dx12;
+
+      if (sprite_coord_origin == PIPE_SPRITE_COORD_LOWER_LEFT) {
+         dady = -dady;
+      }
+
       point->inputs.dadx[slot][i] = dadx;
       point->inputs.dady[slot][i] = dady;
       point->inputs.a0[slot][i] = (0.5 -
-                                  (dadx * ((float)info->v0[0][0] - setup->pixel_offset) +
-                                   dady * ((float)info->v0[0][1] - setup->pixel_offset)));
+                                   (dadx * ((float)info->v0[0][0] - setup->pixel_offset) +
+                                    dady * ((float)info->v0[0][1] - setup->pixel_offset)));
    }
-
    else if (i == 2) {
       point->inputs.a0[slot][i] = 0.0f;
       point->inputs.dadx[slot][i] = 0.0f;
       point->inputs.dady[slot][i] = 0.0f;
    }
-      
    else if (i == 3) {
       point->inputs.a0[slot][i] = 1.0f;
       point->inputs.dadx[slot][i] = 0.0f;
       point->inputs.dady[slot][i] = 0.0f;
    }
-
 }
 
 
@@ -144,6 +148,7 @@ setup_point_fragcoord_coef(struct lp_setup_context *setup,
    }
 }
 
+
 /**
  * Compute the point->coef[] array dadx, dady, a0 values.
  */
@@ -152,6 +157,7 @@ setup_point_coefficients( struct lp_setup_context *setup,
                           struct lp_rast_triangle *point,
                           const struct point_info *info)
 {
+   const struct lp_fragment_shader *shader = setup->fs.current.variant->shader;
    unsigned fragcoord_usage_mask = TGSI_WRITEMASK_XYZ;
    unsigned slot;
 
@@ -172,17 +178,25 @@ setup_point_coefficients( struct lp_setup_context *setup,
          fragcoord_usage_mask |= usage_mask;
          break;
 
+      case LP_INTERP_LINEAR:
+         /* Sprite tex coords may use linear interpolation someday */
+         /* fall-through */
+
       case LP_INTERP_PERSPECTIVE:
-         /* For point sprite textures */        
-         if (setup->fs.current.variant->shader->info.input_semantic_name[slot] 
-             == TGSI_SEMANTIC_GENERIC) 
-         {
-            int index = setup->fs.current.variant->shader->info.input_semantic_index[slot];
-            
-            if (setup->sprite & (1 << index)) {
+         /* check if the sprite coord flag is set for this attribute.
+          * If so, set it up so it up so x any y vary from 0 to 1.
+          */
+         if (shader->info.input_semantic_name[slot] == TGSI_SEMANTIC_GENERIC) {
+            const int index = shader->info.input_semantic_index[slot];
+            /* Note that sprite_coord enable is a bitfield of
+             * PIPE_MAX_SHADER_OUTPUTS bits.
+             */
+            if (index < PIPE_MAX_SHADER_OUTPUTS &&
+                (setup->sprite_coord_enable & (1 << index))) {
                for (i = 0; i < NUM_CHANNELS; i++)
                   if (usage_mask & (1 << i))
-                     perspective_coef(setup, point, info, slot+1, vert_attr, i);
+                     perspective_coef(setup, point, info, slot+1, vert_attr, i,
+                                      setup->sprite_coord_origin);
                fragcoord_usage_mask |= TGSI_WRITEMASK_W;
                break;                     
             }
@@ -203,6 +217,7 @@ setup_point_coefficients( struct lp_setup_context *setup,
                               fragcoord_usage_mask);
 }
 
+
 static INLINE int
 subpixel_snap(float a)
 {
@@ -322,8 +337,9 @@ try_setup_point( struct lp_setup_context *setup,
 }
 
 
-static void lp_setup_point( struct lp_setup_context *setup,
-                           const float (*v0)[4] )
+static void 
+lp_setup_point(struct lp_setup_context *setup,
+               const float (*v0)[4])
 {
    if (!try_setup_point( setup, v0 ))
    {
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index e54dd9f0a3..fb673db6d0 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -886,6 +886,7 @@ static void *
 llvmpipe_create_fs_state(struct pipe_context *pipe,
                          const struct pipe_shader_state *templ)
 {
+   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
    struct lp_fragment_shader *shader;
    int nr_samplers;
 
@@ -902,6 +903,12 @@ llvmpipe_create_fs_state(struct pipe_context *pipe,
    /* we need to keep a local copy of the tokens */
    shader->base.tokens = tgsi_dup_tokens(templ->tokens);
 
+   shader->draw_data = draw_create_fragment_shader(llvmpipe->draw, templ);
+   if (shader->draw_data == NULL) {
+      FREE((void *) shader->base.tokens);
+      return NULL;
+   }
+
    nr_samplers = shader->info.file_max[TGSI_FILE_SAMPLER] + 1;
 
    shader->variant_key_size = Offset(struct lp_fragment_shader_variant_key,
@@ -938,6 +945,9 @@ llvmpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 
    draw_flush(llvmpipe->draw);
 
+   draw_bind_fragment_shader(llvmpipe->draw,
+                             (llvmpipe->fs ? llvmpipe->fs->draw_data : NULL));
+
    llvmpipe->fs = fs;
 
    llvmpipe->dirty |= LP_NEW_FS;
@@ -995,6 +1005,8 @@ llvmpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
       li = next;
    }
 
+   draw_delete_fragment_shader(llvmpipe->draw, shader->draw_data);
+
    assert(shader->variants_cached == 0);
    FREE((void *) shader->base.tokens);
    FREE(shader);
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.h b/src/gallium/drivers/llvmpipe/lp_state_fs.h
index 2914e7d7ef..4999b8dca1 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.h
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.h
@@ -100,6 +100,8 @@ struct lp_fragment_shader
 
    struct lp_fs_variant_list_item variants;
 
+   struct draw_fragment_shader *draw_data;
+
    /* For debugging/profiling purposes */
    unsigned variant_key_size;
    unsigned no;
diff --git a/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c b/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
index 0bad7320f3..dbd73812e4 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_rasterizer.c
@@ -78,8 +78,9 @@ llvmpipe_bind_rasterizer_state(struct pipe_context *pipe, void *handle)
       lp_setup_set_point_state( llvmpipe->setup,
                    llvmpipe->rasterizer->point_size,
                    llvmpipe->rasterizer->point_size_per_vertex,
-                   llvmpipe->rasterizer->sprite_coord_enable);
-       }
+                   llvmpipe->rasterizer->sprite_coord_enable,
+                   llvmpipe->rasterizer->sprite_coord_mode);
+   }
 
    llvmpipe->dirty |= LP_NEW_RASTERIZER;
 }
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index 39d204de8a..c5d61f840f 100644
--- a/src/gallium/drivers/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -70,6 +70,8 @@ struct sp_fragment_shader {
 
    struct tgsi_shader_info info;
 
+   struct draw_fragment_shader *draw_shader;
+
    boolean origin_lower_left; /**< fragment shader uses lower left position origin? */
    boolean pixel_center_integer; /**< fragment shader uses integer pixel center? */
 
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index ded242d3dc..50bc2eea5f 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -60,7 +60,15 @@ softpipe_create_fs_state(struct pipe_context *pipe,
       state = softpipe_create_fs_exec( softpipe, templ );
    }
 
-   assert(state);
+   if (!state)
+      return NULL;
+
+   /* draw's fs state */
+   state->draw_shader = draw_create_fragment_shader(softpipe->draw, templ);
+   if (!state->draw_shader) {
+      state->delete( state );
+      return NULL;
+   }
 
    /* get/save the summary info for this shader */
    tgsi_scan_shader(templ->tokens, &state->info);
@@ -90,6 +98,9 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
 
    softpipe->fs = fs;
 
+   draw_bind_fragment_shader(softpipe->draw,
+                             (softpipe->fs ? softpipe->fs->draw_shader : NULL));
+
    softpipe->dirty |= SP_NEW_FS;
 }
 
@@ -109,6 +120,8 @@ softpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
       tgsi_exec_machine_bind_shader(softpipe->fs_machine, NULL, 0, NULL);
    }
 
+   draw_delete_fragment_shader(softpipe->draw, state->draw_shader);
+
    state->delete( state );
 }