geometry shading patches

[Zack Rusin-2]

yo,

after our discussions i hacked a new version of geometry shading support in
gallium. the new geometry shading syntax looks as follows:
        GEOM
        PROPERTY GS_INPUT_PRIMITIVE TRIANGLES
        PROPERTY GS_OUTPUT_PRIMITIVE TRIANGLE_STRIP
        DCL IN[][0], POSITION, CONSTANT
        DCL IN[][1], COLOR, CONSTANT
        DCL OUT[0], POSITION, CONSTANT
        DCL OUT[1], COLOR, CONSTANT
        0:MOV OUT[0], IN[0][0]
        1:MOV OUT[1], IN[0][1]
        2:EMIT
        3:MOV OUT[0], IN[1][0]
        4:MOV OUT[1], IN[1][1]
        5:EMIT
        6:MOV OUT[0], IN[2][0]
        7:MOV OUT[1], IN[2][1]
        8:EMIT
        9:ENDPRIM
        10:END
the attached patch series adds support for all of it throughout gallium
(including support for 2dimensional arrays in tgsi_sanity checking, tgsi_dump
and tgsi_text).
Right now don't support switching of primitives in the draw module (e.g. input
primitive being points and output primitive being a triangle_strip) but i'd
like to do that after committing the attached changes (also the semantic for
vertices which is a leftover from an old patch should be removed then if it's
ok with you).

Keith, most relevant patches for you are 0001 and 0005.
Michal the most relevant to you are 0007, 0010, 0011 and 0012.

z

[0003-gs-fix-drivers-so-they-work-with-geometry-shaders.patch]

---

From d25f9137b5c9c505623ce67e1528242acad6aedf Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Sat, 19 Dec 2009 13:54:53 -0500
Subject: [PATCH 03/12] gs: fix drivers so they work with geometry shaders

---
 src/gallium/drivers/llvmpipe/lp_state_derived.c |    4 ++--
 src/gallium/drivers/r300/r300_state_derived.c   |    6 +++---
 src/gallium/drivers/svga/svga_swtnl_state.c     |    4 ++--
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/src/gallium/drivers/llvmpipe/lp_state_derived.c b/src/gallium/drivers/llvmpipe/lp_state_derived.c
index fdc6a38..acfd7be 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_derived.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_derived.c
@@ -116,13 +116,13 @@ llvmpipe_get_vertex_info(struct llvmpipe_context *llvmpipe)
          }
 
          /* this includes texcoords and varying vars */
-         src = draw_find_vs_output(llvmpipe->draw,
+         src = draw_find_shader_output(llvmpipe->draw,
                                    lpfs->info.input_semantic_name[i],
                                    lpfs->info.input_semantic_index[i]);
          draw_emit_vertex_attr(vinfo, EMIT_4F, interp, src);
       }
 
-      llvmpipe->psize_slot = draw_find_vs_output(llvmpipe->draw,
+      llvmpipe->psize_slot = draw_find_shader_output(llvmpipe->draw,
                                                  TGSI_SEMANTIC_PSIZE, 0);
       if (llvmpipe->psize_slot > 0) {
          draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT,
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index 29bc701..727ae7a 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -71,9 +71,9 @@ static void r300_draw_emit_attrib(struct r300_context* r300,
     struct tgsi_shader_info* info = &r300->vs->info;
     int output;
 
-    output = draw_find_vs_output(r300->draw,
-                                 info->output_semantic_name[index],
-                                 info->output_semantic_index[index]);
+    output = draw_find_shader_output(r300->draw,
+                                     info->output_semantic_name[index],
+                                     info->output_semantic_index[index]);
     draw_emit_vertex_attr(&r300->vertex_info->vinfo, emit, interp, output);
 }
 
diff --git a/src/gallium/drivers/svga/svga_swtnl_state.c b/src/gallium/drivers/svga/svga_swtnl_state.c
index 25b8c2a..94b6ccc 100644
--- a/src/gallium/drivers/svga/svga_swtnl_state.c
+++ b/src/gallium/drivers/svga/svga_swtnl_state.c
@@ -156,7 +156,7 @@ int svga_swtnl_update_vdecl( struct svga_context *svga )
    memset(vdecl, 0, sizeof(vdecl));
 
    /* always add position */
-   src = draw_find_vs_output(draw, TGSI_SEMANTIC_POSITION, 0);
+   src = draw_find_shader_output(draw, TGSI_SEMANTIC_POSITION, 0);
    draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src);
    vinfo->attrib[0].emit = EMIT_4F;
    vdecl[0].array.offset = offset;
@@ -169,7 +169,7 @@ int svga_swtnl_update_vdecl( struct svga_context *svga )
    for (i = 0; i < fs->base.info.num_inputs; i++) {
       unsigned name = fs->base.info.input_semantic_name[i];
       unsigned index = fs->base.info.input_semantic_index[i];
-      src = draw_find_vs_output(draw, name, index);
+      src = draw_find_shader_output(draw, name, index);
       vdecl[nr_decls].array.offset = offset;
       vdecl[nr_decls].identity.usageIndex = fs->base.info.input_semantic_index[i];
 
--
1.6.5.7

---

[0001-gallium-add-geometry-shader-support-to-gallium.patch]

---

From 3adfbe16c537585c110f6339052a22be38d36cae Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Mon, 14 Dec 2009 17:11:46 -0500
Subject: [PATCH 01/12] gallium: add geometry shader support to gallium

---
 src/gallium/auxiliary/cso_cache/cso_context.c      |   40 +++-
 src/gallium/auxiliary/cso_cache/cso_context.h      |    7 +
 src/gallium/auxiliary/draw/Makefile                |    1 +
 src/gallium/auxiliary/draw/SConscript              |    3 +-
 src/gallium/auxiliary/draw/draw_context.c          |   77 ++++-
 src/gallium/auxiliary/draw/draw_context.h          |   19 +-
 src/gallium/auxiliary/draw/draw_gs.c               |  338 ++++++++++++++++++++
 src/gallium/auxiliary/draw/draw_gs.h               |   76 +++++
 src/gallium/auxiliary/draw/draw_pipe_aaline.c      |   12 +-
 src/gallium/auxiliary/draw/draw_pipe_aapoint.c     |   12 +-
 src/gallium/auxiliary/draw/draw_pipe_clip.c        |    4 +-
 src/gallium/auxiliary/draw/draw_pipe_cull.c        |    2 +-
 src/gallium/auxiliary/draw/draw_pipe_offset.c      |    2 +-
 src/gallium/auxiliary/draw/draw_pipe_stipple.c     |    5 +-
 src/gallium/auxiliary/draw/draw_pipe_wide_line.c   |    2 +-
 src/gallium/auxiliary/draw/draw_pipe_wide_point.c  |   14 +-
 src/gallium/auxiliary/draw/draw_private.h          |   32 ++-
 .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c  |   50 +++-
 src/gallium/auxiliary/draw/draw_pt_post_vs.c       |    4 +-
 src/gallium/auxiliary/draw/draw_pt_util.c          |   16 +
 .../auxiliary/draw/draw_pt_varray_tmp_linear.h     |    4 +
 src/gallium/auxiliary/draw/draw_vs_varient.c       |    5 +-
 src/gallium/auxiliary/tgsi/tgsi_dump.c             |    4 +-
 src/gallium/auxiliary/tgsi/tgsi_exec.c             |   45 +++-
 src/gallium/auxiliary/tgsi/tgsi_exec.h             |   15 +-
 src/gallium/auxiliary/tgsi/tgsi_text.c             |    4 +-
 src/gallium/drivers/cell/ppu/cell_draw_arrays.c    |    2 +-
 src/gallium/drivers/cell/ppu/cell_state_derived.c  |    8 +-
 src/gallium/drivers/cell/ppu/cell_state_emit.c     |    2 +-
 src/gallium/drivers/i915/i915_context.c            |    2 +-
 src/gallium/drivers/i915/i915_state_derived.c      |   10 +-
 src/gallium/drivers/llvmpipe/lp_state_derived.c    |    2 +-
 src/gallium/drivers/llvmpipe/lp_state_fs.c         |    3 +-
 src/gallium/drivers/nv04/nv04_vbo.c                |    2 +-
 src/gallium/drivers/nv10/nv10_vbo.c                |    1 +
 src/gallium/drivers/nv20/nv20_state_emit.c         |   16 +-
 src/gallium/drivers/nv20/nv20_vbo.c                |    2 +-
 src/gallium/drivers/nv40/nv40_draw.c               |    5 +-
 src/gallium/drivers/softpipe/sp_context.c          |    4 +
 src/gallium/drivers/softpipe/sp_context.h          |    1 +
 src/gallium/drivers/softpipe/sp_draw_arrays.c      |   23 +-
 src/gallium/drivers/softpipe/sp_screen.c           |    2 +
 src/gallium/drivers/softpipe/sp_setup.c            |    2 +-
 src/gallium/drivers/softpipe/sp_state.h            |   10 +
 src/gallium/drivers/softpipe/sp_state_derived.c    |   10 +-
 src/gallium/drivers/softpipe/sp_state_fs.c         |   59 ++++
 src/gallium/drivers/svga/svga_swtnl_draw.c         |    2 +-
 src/gallium/include/pipe/p_context.h               |    6 +
 src/gallium/include/pipe/p_defines.h               |   30 +-
 src/gallium/include/pipe/p_shader_tokens.h         |   22 +-
 src/gallium/state_trackers/python/p_context.i      |   19 ++
 src/gallium/state_trackers/python/samples/gs.py    |  254 +++++++++++++++
 src/gallium/state_trackers/python/st_device.h      |    1 +
 src/mesa/state_tracker/st_draw_feedback.c          |    3 +-
 54 files changed, 1159 insertions(+), 137 deletions(-)
 create mode 100644 src/gallium/auxiliary/draw/draw_gs.c
 create mode 100644 src/gallium/auxiliary/draw/draw_gs.h
 create mode 100644 src/gallium/state_trackers/python/samples/gs.py

diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index 80bd0c9..2b16332 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -42,6 +42,7 @@
 #include "cso_cache/cso_context.h"
 #include "cso_cache/cso_cache.h"
 #include "cso_cache/cso_hash.h"
+#include "cso_context.h"
 
 struct cso_context {
    struct pipe_context *pipe;
@@ -85,8 +86,8 @@ struct cso_context {
    void *blend, *blend_saved;
    void *depth_stencil, *depth_stencil_saved;
    void *rasterizer, *rasterizer_saved;
-   void *fragment_shader, *fragment_shader_saved;
-   void *vertex_shader, *vertex_shader_saved;
+   void *fragment_shader, *fragment_shader_saved, *geometry_shader;
+   void *vertex_shader, *vertex_shader_saved, *geometry_shader_saved;
 
    struct pipe_framebuffer_state fb, fb_saved;
    struct pipe_viewport_state vp, vp_saved;
@@ -1027,3 +1028,38 @@ enum pipe_error cso_set_blend_color(struct cso_context *ctx,
    }
    return PIPE_OK;
 }
+
+enum pipe_error cso_set_geometry_shader_handle(struct cso_context *ctx,
+                                               void *handle)
+{
+   if (ctx->geometry_shader != handle) {
+      ctx->geometry_shader = handle;
+      ctx->pipe->bind_gs_state(ctx->pipe, handle);
+   }
+   return PIPE_OK;
+}
+
+void cso_delete_geometry_shader(struct cso_context *ctx, void *handle)
+{
+    if (handle == ctx->geometry_shader) {
+      /* unbind before deleting */
+      ctx->pipe->bind_gs_state(ctx->pipe, NULL);
+      ctx->geometry_shader = NULL;
+   }
+   ctx->pipe->delete_gs_state(ctx->pipe, handle);
+}
+
+void cso_save_geometry_shader(struct cso_context *ctx)
+{
+   assert(!ctx->geometry_shader_saved);
+   ctx->geometry_shader_saved = ctx->geometry_shader;
+}
+
+void cso_restore_geometry_shader(struct cso_context *ctx)
+{
+   if (ctx->geometry_shader_saved != ctx->geometry_shader) {
+      ctx->pipe->bind_gs_state(ctx->pipe, ctx->geometry_shader_saved);
+      ctx->geometry_shader = ctx->geometry_shader_saved;
+   }
+   ctx->geometry_shader_saved = NULL;
+}
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h
index e5b9217..b9e313e 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.h
+++ b/src/gallium/auxiliary/cso_cache/cso_context.h
@@ -146,6 +146,13 @@ void cso_save_vertex_shader(struct cso_context *cso);
 void cso_restore_vertex_shader(struct cso_context *cso);
 
 
+enum pipe_error cso_set_geometry_shader_handle(struct cso_context *ctx,
+                                               void *handle);
+void cso_delete_geometry_shader(struct cso_context *ctx, void *handle);
+void cso_save_geometry_shader(struct cso_context *cso);
+void cso_restore_geometry_shader(struct cso_context *cso);
+
+
 
 enum pipe_error cso_set_framebuffer(struct cso_context *cso,
                                     const struct pipe_framebuffer_state *fb);
diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile
index 5041dcc..2481674 100644
--- a/src/gallium/auxiliary/draw/Makefile
+++ b/src/gallium/auxiliary/draw/Makefile
@@ -5,6 +5,7 @@ LIBNAME = draw
 
 C_SOURCES = \
  draw_context.c \
+        draw_gs.c \
  draw_pipe.c \
  draw_pipe_aaline.c \
  draw_pipe_aapoint.c \
diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript
index 5f05aa3..a022c14 100644
--- a/src/gallium/auxiliary/draw/SConscript
+++ b/src/gallium/auxiliary/draw/SConscript
@@ -40,7 +40,8 @@ draw = env.ConvenienceLibrary(
  'draw_vs_llvm.c',
  'draw_vs_ppc.c',
  'draw_vs_sse.c',
- 'draw_vs_varient.c'
+ 'draw_vs_varient.c',
+                'draw_gs.c'
  ])
 
 auxiliaries.insert(0, draw)
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index cc5f7f0..667aa46 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -36,6 +36,7 @@
 #include "draw_context.h"
 #include "draw_vbuf.h"
 #include "draw_vs.h"
+#include "draw_gs.h"
 #include "draw_pt.h"
 #include "draw_pipe.h"
 
@@ -67,6 +68,9 @@ struct draw_context *draw_create( void )
    if (!draw_vs_init( draw ))
       goto fail;
 
+   if (!draw_gs_init( draw ))
+      goto fail;
+
    return draw;
 
 fail:
@@ -231,11 +235,19 @@ draw_set_mapped_vertex_buffer(struct draw_context *draw,
 
 void
 draw_set_mapped_constant_buffer(struct draw_context *draw,
-                                const void *buffer,
+                                unsigned shader_type,
+                                const void *buffer,
                                 unsigned size )
 {
-   draw->pt.user.constants = buffer;
-   draw_vs_set_constants( draw, (const float (*)[4])buffer, size );
+   debug_assert(shader_type == PIPE_SHADER_VERTEX ||
+                shader_type == PIPE_SHADER_GEOMETRY);
+   if (shader_type == PIPE_SHADER_VERTEX) {
+      draw->pt.user.vs_constants = buffer;
+      draw_vs_set_constants( draw, (const float (*)[4])buffer, size );
+   } else if (shader_type == PIPE_SHADER_GEOMETRY) {
+      draw->pt.user.gs_constants = buffer;
+      draw_gs_set_constants( draw, (const float (*)[4])buffer, size );
+   }
 }
 
 
@@ -298,7 +310,7 @@ draw_set_force_passthrough( struct draw_context *draw, boolean enable )
  * a post-transformed vertex.
  *
  * With this function, drivers that use the draw module should have no reason
- * to track the current vertex shader.
+ * to track the current vertex/geometry shader.
  *
  * Note that the draw module may sometimes generate vertices with extra
  * attributes (such as texcoords for AA lines).  The driver can call this
@@ -309,43 +321,59 @@ draw_set_force_passthrough( struct draw_context *draw, boolean enable )
  * work for the drivers.
  */
 int
-draw_find_vs_output(const struct draw_context *draw,
-                    uint semantic_name, uint semantic_index)
+draw_find_shader_output(const struct draw_context *draw,
+                        uint semantic_name, uint semantic_index)
 {
    const struct draw_vertex_shader *vs = draw->vs.vertex_shader;
+   const struct draw_geometry_shader *gs = draw->gs.geometry_shader;
    uint i;
-   for (i = 0; i < vs->info.num_outputs; i++) {
-      if (vs->info.output_semantic_name[i] == semantic_name &&
-          vs->info.output_semantic_index[i] == semantic_index)
+   const struct tgsi_shader_info *info = &vs->info;
+
+   if (gs)
+      info = &gs->info;
+
+   for (i = 0; i < info->num_outputs; i++) {
+      if (info->output_semantic_name[i] == semantic_name &&
+          info->output_semantic_index[i] == semantic_index)
          return i;
    }
 
    /* XXX there may be more than one extra vertex attrib.
     * For example, simulated gl_FragCoord and gl_PointCoord.
     */
-   if (draw->extra_vp_outputs.semantic_name == semantic_name &&
-       draw->extra_vp_outputs.semantic_index == semantic_index) {
-      return draw->extra_vp_outputs.slot;
+   if (draw->extra_shader_outputs.semantic_name == semantic_name &&
+       draw->extra_shader_outputs.semantic_index == semantic_index) {
+      return draw->extra_shader_outputs.slot;
    }
+
    return 0;
 }
 
 
 /**
- * Return number of vertex shader outputs.
+ * Return number of the shader outputs.
+ *
+ * If geometry shader is present, its output will be returned,
+ * if not vertex shader is used.
  */
 uint
-draw_num_vs_outputs(const struct draw_context *draw)
+draw_num_shader_outputs(const struct draw_context *draw)
 {
    uint count = draw->vs.vertex_shader->info.num_outputs;
-   if (draw->extra_vp_outputs.slot > 0)
+
+   /* if geometry shader is present, its outputs go to te
+    * driver, not the vertex shaders */
+   if (draw->gs.geometry_shader)
+      count = draw->gs.geometry_shader->info.num_outputs;
+
+   if (draw->extra_shader_outputs.slot > 0)
       count++;
    return count;
 }
 
 
 /**
- * Provide TGSI sampler objects for vertex shaders that use texture fetches.
+ * Provide TGSI sampler objects for vertex/geometry shaders that use texture fetches.
  * This might only be used by software drivers for the time being.
  */
 void
@@ -355,6 +383,8 @@ draw_texture_samplers(struct draw_context *draw,
 {
    draw->vs.num_samplers = num_samplers;
    draw->vs.samplers = samplers;
+   draw->gs.num_samplers = num_samplers;
+   draw->gs.samplers = samplers;
 }
 
 
@@ -421,3 +451,18 @@ void draw_do_flush( struct draw_context *draw, unsigned flags )
       draw->flushing = FALSE;
    }
 }
+
+
+int draw_current_shader_outputs(struct draw_context *draw)
+{
+   if (draw->gs.geometry_shader)
+      return draw->gs.num_gs_outputs;
+   return draw->vs.num_vs_outputs;
+}
+
+int draw_current_shader_position_output(struct draw_context *draw)
+{
+   if (draw->gs.geometry_shader)
+      return draw->gs.position_output;
+   return draw->vs.position_output;
+}
diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index 465b8f1..b716209 100644
--- a/src/gallium/auxiliary/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -45,6 +45,7 @@ struct pipe_context;
 struct draw_context;
 struct draw_stage;
 struct draw_vertex_shader;
+struct draw_geometry_shader;
 struct tgsi_sampler;
 
 
@@ -85,11 +86,11 @@ draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe
 
 
 int
-draw_find_vs_output(const struct draw_context *draw,
-                    uint semantic_name, uint semantic_index);
+draw_find_shader_output(const struct draw_context *draw,
+                        uint semantic_name, uint semantic_index);
 
 uint
-draw_num_vs_outputs(const struct draw_context *draw);
+draw_num_shader_outputs(const struct draw_context *draw);
 
 
 void
@@ -112,6 +113,17 @@ void draw_delete_vertex_shader(struct draw_context *draw,
                                struct draw_vertex_shader *dvs);
 
 
+/*
+ * Geometry shader functions
+ */
+struct draw_geometry_shader *
+draw_create_geometry_shader(struct draw_context *draw,
+                            const struct pipe_shader_state *shader);
+void draw_bind_geometry_shader(struct draw_context *draw,
+                               struct draw_geometry_shader *dvs);
+void draw_delete_geometry_shader(struct draw_context *draw,
+                                 struct draw_geometry_shader *dvs);
+
 
 /*
  * Vertex data functions
@@ -140,6 +152,7 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw,
                                    unsigned attr, const void *buffer);
 
 void draw_set_mapped_constant_buffer(struct draw_context *draw,
+                                     unsigned shader_type,
                                      const void *buffer,
                                      unsigned size );
 
diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
new file mode 100644
index 0000000..1154228
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -0,0 +1,338 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMWare Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "draw_gs.h"
+
+#include "draw_private.h"
+#include "draw_context.h"
+
+#include "tgsi/tgsi_parse.h"
+#include "tgsi/tgsi_exec.h"
+
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#define MAX_PRIM_VERTICES 6
+/* fixme: move it from here */
+#define MAX_PRIMITIVES 64
+
+boolean
+draw_gs_init( struct draw_context *draw )
+{
+   draw->gs.machine = tgsi_exec_machine_create();
+   if (!draw->gs.machine)
+      return FALSE;
+
+   draw->gs.machine->Primitives = align_malloc(
+      MAX_PRIMITIVES * sizeof(struct tgsi_exec_vector), 16);
+   if (!draw->gs.machine->Primitives)
+      return FALSE;
+   memset(draw->gs.machine->Primitives, 0,
+          MAX_PRIMITIVES * sizeof(struct tgsi_exec_vector));
+
+   return TRUE;
+}
+
+
+void draw_gs_set_constants( struct draw_context *draw,
+                            const float (*constants)[4],
+                            unsigned size )
+{
+}
+
+
+struct draw_geometry_shader *
+draw_create_geometry_shader(struct draw_context *draw,
+                            const struct pipe_shader_state *state)
+{
+   struct draw_geometry_shader *gs;
+   int i;
+
+   gs = CALLOC_STRUCT(draw_geometry_shader);
+
+   if (!gs)
+      return NULL;
+
+   gs->state = *state;
+   gs->state.tokens = tgsi_dup_tokens(state->tokens);
+   if (!gs->state.tokens) {
+      FREE(gs);
+      return NULL;
+   }
+
+   tgsi_scan_shader(state->tokens, &gs->info);
+
+   /* setup the defaults */
+   gs->input_primitive = PIPE_PRIM_TRIANGLES;
+   gs->output_primitive = PIPE_PRIM_TRIANGLE_STRIP;
+   gs->max_output_vertices = 32;
+
+   for (i = 0; i < gs->info.num_properties; ++i) {
+      if (gs->info.properties[i].name ==
+          TGSI_PROPERTY_GS_INPUT_PRIM)
+         gs->input_primitive = gs->info.properties[i].data[0];
+      else if (gs->info.properties[i].name ==
+               TGSI_PROPERTY_GS_OUTPUT_PRIM)
+         gs->output_primitive = gs->info.properties[i].data[0];
+      else if (gs->info.properties[i].name ==
+               TGSI_PROPERTY_GS_MAX_VERTICES)
+         gs->max_output_vertices = gs->info.properties[i].data[0];
+   }
+
+   gs->machine = draw->gs.machine;
+
+   if (gs)
+   {
+      uint i;
+      for (i = 0; i < gs->info.num_outputs; i++) {
+         if (gs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION &&
+             gs->info.output_semantic_index[i] == 0)
+            gs->position_output = i;
+      }
+   }
+
+   return gs;
+}
+
+void draw_bind_geometry_shader(struct draw_context *draw,
+                               struct draw_geometry_shader *dgs)
+{
+   draw_do_flush(draw, DRAW_FLUSH_STATE_CHANGE);
+
+   if (dgs) {
+      draw->gs.geometry_shader = dgs;
+      draw->gs.num_gs_outputs = dgs->info.num_outputs;
+      draw->gs.position_output = dgs->position_output;
+      draw_geometry_shader_prepare(dgs, draw);
+   }
+   else {
+      draw->gs.geometry_shader = NULL;
+      draw->gs.num_gs_outputs = 0;
+   }
+}
+
+void draw_delete_geometry_shader(struct draw_context *draw,
+                                 struct draw_geometry_shader *dgs)
+{
+   FREE(dgs);
+}
+
+static INLINE int num_vertices_for_prim(int prim)
+{
+   switch(prim) {
+   case PIPE_PRIM_POINTS:
+      return 1;
+   case PIPE_PRIM_LINES:
+      return 2;
+   case PIPE_PRIM_LINE_LOOP:
+      return 2;
+   case PIPE_PRIM_LINE_STRIP:
+      return 2;
+   case PIPE_PRIM_TRIANGLES:
+      return 3;
+   case PIPE_PRIM_TRIANGLE_STRIP:
+      return 3;
+   case PIPE_PRIM_TRIANGLE_FAN:
+      return 3;
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      return 4;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      return 6;
+   default:
+      assert(!"Bad geometry shader input");
+      return 0;
+   }
+}
+
+static void draw_fetch_geometry_input(struct draw_geometry_shader *shader,
+                                      int start_primitive,
+                                      int num_primitives,
+                                      const float (*input_ptr)[4],
+                                      unsigned input_vertex_stride,
+                                      unsigned inputs_from_vs)
+{
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned slot, vs_slot, k, j;
+   unsigned num_vertices = num_vertices_for_prim(shader->input_primitive);
+   int idx = 0;
+
+   for (slot = 0, vs_slot = 0; slot < shader->info.num_inputs; slot++) {
+      debug_printf("Slot = %d (semantic = %d)\n", slot,
+                   shader->info.input_semantic_name[slot]);
+      if (shader->info.input_semantic_name[slot] ==
+          TGSI_SEMANTIC_VERTICES) {
+         for (j = 0; j < num_primitives; ++j) {
+            machine->Inputs[idx].xyzw[0].f[j] = (float)num_vertices;
+            machine->Inputs[idx].xyzw[1].f[j] = (float)num_vertices;
+            machine->Inputs[idx].xyzw[2].f[j] = (float)num_vertices;
+            machine->Inputs[idx].xyzw[3].f[j] = (float)num_vertices;
+         }
+         ++idx;
+      } else {
+         for (j = 0; j < num_primitives; ++j) {
+            int vidx = idx;
+            const float (*prim_ptr)[4];
+            debug_printf("    %d) Prim (num_verts = %d)\n", start_primitive + j,
+                         num_vertices);
+            prim_ptr = (const float (*)[4])(
+               (const char *)input_ptr +
+               (j * num_vertices * input_vertex_stride));
+
+            for (k = 0; k < num_vertices; ++k, ++vidx) {
+               const float (*input)[4];
+               input = (const float (*)[4])(
+                  (const char *)prim_ptr + (k * input_vertex_stride));
+               debug_printf("\t%d)(%d) Input vert:\n", vidx, k);
+#if 1
+               assert(!util_is_inf_or_nan(input[vs_slot][0]));
+               assert(!util_is_inf_or_nan(input[vs_slot][1]));
+               assert(!util_is_inf_or_nan(input[vs_slot][2]));
+               assert(!util_is_inf_or_nan(input[vs_slot][3]));
+#endif
+               machine->Inputs[vidx].xyzw[0].f[j] = input[vs_slot][0];
+               machine->Inputs[vidx].xyzw[1].f[j] = input[vs_slot][1];
+               machine->Inputs[vidx].xyzw[2].f[j] = input[vs_slot][2];
+               machine->Inputs[vidx].xyzw[3].f[j] = input[vs_slot][3];
+#if 0
+               debug_printf("\t\t%d %f %f %f %f\n", slot,
+                            machine->Inputs[vidx].xyzw[0].f[j],
+                            machine->Inputs[vidx].xyzw[1].f[j],
+                            machine->Inputs[vidx].xyzw[2].f[j],
+                            machine->Inputs[vidx].xyzw[3].f[j]);
+#endif
+            }
+         }
+         ++vs_slot;
+         idx += num_vertices;
+      }
+   }
+}
+
+static INLINE void
+draw_geometry_fetch_outputs(struct draw_geometry_shader *shader,
+                            int num_primitives,
+                            float (*output)[4],
+                            unsigned vertex_size)
+{
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned prim_idx, j, slot;
+
+   /* Unswizzle all output results.
+    */
+   /* FIXME: handle all the primitives produced by the gs, not just
+    * the first one
+    unsigned prim_count =
+    mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0];*/
+   for (prim_idx = 0; prim_idx < num_primitives; ++prim_idx) {
+      unsigned num_verts_per_prim = machine->Primitives[0];
+      for (j = 0; j < num_verts_per_prim; j++) {
+         int idx = (prim_idx * num_verts_per_prim + j) *
+                   shader->info.num_outputs;
+#ifdef DEBUG_OUTPUTS
+         debug_printf("%d) Output vert:\n", idx);
+#endif
+         for (slot = 0; slot < shader->info.num_outputs; slot++) {
+            output[slot][0] = machine->Outputs[idx + slot].xyzw[0].f[prim_idx];
+            output[slot][1] = machine->Outputs[idx + slot].xyzw[1].f[prim_idx];
+            output[slot][2] = machine->Outputs[idx + slot].xyzw[2].f[prim_idx];
+            output[slot][3] = machine->Outputs[idx + slot].xyzw[3].f[prim_idx];
+#ifdef DEBUG_OUTPUTS
+            debug_printf("\t%d: %f %f %f %f\n", slot,
+                         output[slot][0],
+                         output[slot][1],
+                         output[slot][2],
+                         output[slot][3]);
+#endif
+            debug_assert(!util_is_inf_or_nan(output[slot][0]));
+         }
+         output = (float (*)[4])((char *)output + vertex_size);
+      }
+   }
+}
+
+void draw_geometry_shader_run(struct draw_geometry_shader *shader,
+                              const float (*input)[4],
+                              float (*output)[4],
+                              const float (*constants)[4],
+                              unsigned count,
+                              unsigned input_stride,
+                              unsigned vertex_size)
+{
+   struct tgsi_exec_machine *machine = shader->machine;
+   unsigned int i;
+   unsigned num_vertices = num_vertices_for_prim(shader->input_primitive);
+   unsigned num_primitives = count/num_vertices;
+   unsigned inputs_from_vs = 0;
+
+   machine->Consts = constants;
+
+   for (i = 0; i < shader->info.num_inputs; ++i) {
+      if (shader->info.input_semantic_name[i] != TGSI_SEMANTIC_VERTICES &&
+          shader->info.input_semantic_name[i] != TGSI_SEMANTIC_PRIMID)
+         ++inputs_from_vs;
+   }
+
+   for (i = 0; i < num_primitives; ++i) {
+      unsigned int max_primitives = 1;
+
+      draw_fetch_geometry_input(shader, i, max_primitives, input,
+                                input_stride, inputs_from_vs);
+
+      tgsi_set_exec_mask(machine,
+                         1,
+                         max_primitives > 1,
+                         max_primitives > 2,
+                         max_primitives > 3);
+
+      /* run interpreter */
+      tgsi_exec_machine_run(machine);
+
+      draw_geometry_fetch_outputs(shader, max_primitives,
+                                  output, vertex_size);
+   }
+}
+
+void draw_geometry_shader_delete(struct draw_geometry_shader *shader)
+{
+   FREE((void*) shader->state.tokens);
+   FREE(shader);
+}
+
+void draw_geometry_shader_prepare(struct draw_geometry_shader *shader,
+                                  struct draw_context *draw)
+{
+    if (shader->machine->Tokens != shader->state.tokens) {
+       tgsi_exec_machine_bind_shader(shader->machine,
+                                     shader->state.tokens,
+                                     draw->gs.num_samplers,
+                                     draw->gs.samplers);
+    }
+}
diff --git a/src/gallium/auxiliary/draw/draw_gs.h b/src/gallium/auxiliary/draw/draw_gs.h
new file mode 100644
index 0000000..d6a97d9
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_gs.h
@@ -0,0 +1,76 @@
+/**************************************************************************
+ *
+ * Copyright 2009 VMWare Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#ifndef DRAW_GS_H
+#define DRAW_GS_H
+
+#include "draw_context.h"
+#include "draw_private.h"
+
+
+#define MAX_TGSI_PRIMITIVES 4
+
+struct draw_context;
+
+/**
+ * Private version of the compiled geometry shader
+ */
+struct draw_geometry_shader {
+   struct draw_context *draw;
+
+   struct tgsi_exec_machine *machine;
+
+   /* This member will disappear shortly:*/
+   struct pipe_shader_state state;
+
+   struct tgsi_shader_info info;
+   unsigned position_output;
+
+   unsigned max_output_vertices;
+   unsigned input_primitive;
+   unsigned output_primitive;
+
+   /* Extracted from shader:
+    */
+   const float (*immediates)[4];
+};
+
+void draw_geometry_shader_run(struct draw_geometry_shader *shader,
+                              const float (*input)[4],
+                              float (*output)[4],
+                              const float (*constants)[4],
+                              unsigned count,
+                              unsigned input_stride,
+                              unsigned output_stride);
+
+void draw_geometry_shader_prepare(struct draw_geometry_shader *shader,
+                                  struct draw_context *draw);
+
+void draw_geometry_shader_delete(struct draw_geometry_shader *shader);
+
+
+#endif
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
index 1437542..4585dcd 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c
@@ -660,13 +660,13 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header)
    }
 
    /* update vertex attrib info */
-   aaline->tex_slot = draw->vs.num_vs_outputs;
-   aaline->pos_slot = draw->vs.position_output;
+   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_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-   draw->extra_vp_outputs.semantic_index = aaline->fs->generic_attrib;
-   draw->extra_vp_outputs.slot = aaline->tex_slot;
+   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;
 
    /* how many samplers? */
    /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */
@@ -707,7 +707,7 @@ aaline_flush(struct draw_stage *stage, unsigned flags)
                                        aaline->state.texture);
    draw->suspend_flushing = FALSE;
 
-   draw->extra_vp_outputs.slot = 0;
+   draw->extra_shader_outputs.slot = 0;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
index 75130a8..d86717e 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c
@@ -687,14 +687,14 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
    bind_aapoint_fragment_shader(aapoint);
 
    /* update vertex attrib info */
-   aapoint->tex_slot = draw->vs.num_vs_outputs;
+   aapoint->tex_slot = draw_current_shader_outputs(draw);
    assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */
 
-   aapoint->pos_slot = draw->vs.position_output;
+   aapoint->pos_slot = draw_current_shader_position_output(draw);
 
-   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-   draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib;
-   draw->extra_vp_outputs.slot = aapoint->tex_slot;
+   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;
 
    /* find psize slot in post-transform vertex */
    aapoint->psize_slot = -1;
@@ -731,7 +731,7 @@ aapoint_flush(struct draw_stage *stage, unsigned flags)
    aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs);
    draw->suspend_flushing = FALSE;
 
-   draw->extra_vp_outputs.slot = 0;
+   draw->extra_shader_outputs.slot = 0;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c
index 0670268..205cda5 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_clip.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c
@@ -114,8 +114,8 @@ static void interp( const struct clipper *clip,
     const struct vertex_header *out,
     const struct vertex_header *in )
 {
-   const unsigned nr_attrs = clip->stage.draw->vs.num_vs_outputs;
-   const unsigned pos_attr = clip->stage.draw->vs.position_output;
+   const unsigned nr_attrs = draw_current_shader_outputs(clip->stage.draw);
+   const unsigned pos_attr = draw_current_shader_position_output(clip->stage.draw);
    unsigned j;
 
    /* Vertex header.
diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c
index 0a70483..11b39db 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_cull.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c
@@ -55,7 +55,7 @@ static INLINE struct cull_stage *cull_stage( struct draw_stage *stage )
 static void cull_tri( struct draw_stage *stage,
       struct prim_header *header )
 {
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
 
    /* Window coords: */
    const float *v0 = header->v[0]->data[pos];
diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c
index 40798a5..e829492 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_offset.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c
@@ -63,7 +63,7 @@ static INLINE struct offset_stage *offset_stage( struct draw_stage *stage )
 static void do_offset_tri( struct draw_stage *stage,
    struct prim_header *header )
 {
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    struct offset_stage *offset = offset_stage(stage);  
    float inv_det = 1.0f / header->det;
 
diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
index 6e921ba..70fbab9 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c
@@ -73,7 +73,8 @@ screen_interp( struct draw_context *draw,
                const struct vertex_header *v1 )
 {
    uint attr;
-   for (attr = 0; attr < draw->vs.num_vs_outputs; attr++) {
+   int num_outputs = draw_current_shader_outputs(draw);
+   for (attr = 0; attr < num_outputs; attr++) {
       const float *val0 = v0->data[attr];
       const float *val1 = v1->data[attr];
       float *newv = dst->data[attr];
@@ -121,7 +122,7 @@ stipple_line(struct draw_stage *stage, struct prim_header *header)
    struct stipple_stage *stipple = stipple_stage(stage);
    struct vertex_header *v0 = header->v[0];
    struct vertex_header *v1 = header->v[1];
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    const float *pos0 = v0->data[pos];
    const float *pos1 = v1->data[pos];
    float start = 0;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
index f32cbef..3073c87 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c
@@ -59,7 +59,7 @@ static void wideline_line( struct draw_stage *stage,
                            struct prim_header *header )
 {
    /*const struct wideline_stage *wide = wideline_stage(stage);*/
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    const float half_width = 0.5f * stage->draw->rasterizer->line_width;
 
    struct prim_header tri;
diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
index 7d76a7d..8dc50c0 100644
--- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
+++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c
@@ -112,7 +112,7 @@ static void set_texcoords(const struct widepoint_stage *wide,
 
    if (wide->point_coord_fs_input >= 0) {
       /* put gl_PointCoord into the extra vertex slot */
-      uint slot = wide->stage.draw->extra_vp_outputs.slot;
+      uint slot = wide->stage.draw->extra_shader_outputs.slot;
       v->data[slot][0] = tc[0];
       v->data[slot][1] = tc[1];
       v->data[slot][2] = 0.0F;
@@ -130,7 +130,7 @@ static void widepoint_point( struct draw_stage *stage,
                              struct prim_header *header )
 {
    const struct widepoint_stage *wide = widepoint_stage(stage);
-   const unsigned pos = stage->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(stage->draw);
    const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite;
    float half_size;
    float left_adj, right_adj, bot_adj, top_adj;
@@ -257,13 +257,13 @@ static void widepoint_first_point( struct draw_stage *stage,
       wide->point_coord_fs_input = find_pntc_input_attrib(draw);
 
       /* setup extra vp output (point coord implemented as a texcoord) */
-      draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
-      draw->extra_vp_outputs.semantic_index = 0;
-      draw->extra_vp_outputs.slot = draw->vs.num_vs_outputs;
+      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_vp_outputs.slot = 0;
+      draw->extra_shader_outputs.slot = 0;
    }
 
    wide->psize_slot = -1;
@@ -287,7 +287,7 @@ static void widepoint_flush( struct draw_stage *stage, unsigned flags )
 {
    stage->point = widepoint_first_point;
    stage->next->flush( stage->next, flags );
-   stage->draw->extra_vp_outputs.slot = 0;
+   stage->draw->extra_shader_outputs.slot = 0;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index 3850ced..e490415 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -152,8 +152,9 @@ struct draw_context
          /** vertex arrays */
          const void *vbuffer[PIPE_MAX_ATTRIBS];
         
-         /** constant buffer (for vertex shader) */
-         const void *constants;
+         /** constant buffer (for vertex/geometry shader) */
+         const void *vs_constants;
+         const void *gs_constants;
       } user;
 
       boolean test_fse;         /* enable FSE even though its not correct (eg for softpipe) */
@@ -211,6 +212,18 @@ struct draw_context
       struct translate_cache *emit_cache;
    } vs;
 
+   struct {
+      struct draw_geometry_shader *geometry_shader;
+      uint num_gs_outputs;  /**< convenience, from geometry_shader */
+      uint position_output;
+
+      /** TGSI program interpreter runtime state */
+      struct tgsi_exec_machine *machine;
+
+      uint num_samplers;
+      struct tgsi_sampler **samplers;
+   } gs;
+
    /* Clip derived state:
     */
    float plane[12][4];
@@ -222,7 +235,7 @@ struct draw_context
       uint semantic_name;
       uint semantic_index;
       int slot;
-   } extra_vp_outputs;
+   } extra_shader_outputs;
 
    unsigned reduced_prim;
 
@@ -245,6 +258,19 @@ void draw_vs_set_constants( struct draw_context *,
 
 
 
+/*******************************************************************************
+ * Geometry shading code:
+ */
+boolean draw_gs_init( struct draw_context *draw );
+void draw_gs_set_constants( struct draw_context *,
+                            const float (*constants)[4],
+                            unsigned size );
+
+/*******************************************************************************
+ * Common shading code:
+ */
+int draw_current_shader_outputs(struct draw_context *draw);
+int draw_current_shader_position_output(struct draw_context *draw);
 
 /*******************************************************************************
  * Vertex processing (was passthrough) code:
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
index 9321137..1a9df4c 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
@@ -32,6 +32,7 @@
 #include "draw/draw_vertex.h"
 #include "draw/draw_pt.h"
 #include "draw/draw_vs.h"
+#include "draw/draw_gs.h"
 #include "translate/translate.h"
 
 
@@ -119,7 +120,8 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle,
 {
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
-   struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   struct draw_vertex_shader *vshader = draw->vs.vertex_shader;
+   struct draw_geometry_shader *gshader = draw->gs.geometry_shader;
    unsigned opt = fpme->opt;
    unsigned alloc_count = align( fetch_count, 4 );
 
@@ -147,13 +149,21 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle,
     */
    if (opt & PT_SHADE)
    {
-      shader->run_linear(shader,
- (const float (*)[4])pipeline_verts->data,
- (      float (*)[4])pipeline_verts->data,
- (const float (*)[4])draw->pt.user.constants,
- fetch_count,
- fpme->vertex_size,
- fpme->vertex_size);
+      vshader->run_linear(vshader,
+                          (const float (*)[4])pipeline_verts->data,
+                          (      float (*)[4])pipeline_verts->data,
+                          (const float (*)[4])draw->pt.user.vs_constants,
+                          fetch_count,
+                          fpme->vertex_size,
+                          fpme->vertex_size);
+      if (gshader)
+         draw_geometry_shader_run(gshader,
+                                  (const float (*)[4])pipeline_verts->data,
+                                  (      float (*)[4])pipeline_verts->data,
+                                  (const float (*)[4])draw->pt.user.gs_constants,
+                                  fetch_count,
+                                  fpme->vertex_size,
+                                  fpme->vertex_size);
    }
 
    if (draw_pt_post_vs_run( fpme->post_vs,
@@ -196,6 +206,7 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle,
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
    struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   struct draw_geometry_shader *geometry_shader = draw->gs.geometry_shader;
    unsigned opt = fpme->opt;
    unsigned alloc_count = align( count, 4 );
 
@@ -226,10 +237,19 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle,
       shader->run_linear(shader,
  (const float (*)[4])pipeline_verts->data,
  (      float (*)[4])pipeline_verts->data,
- (const float (*)[4])draw->pt.user.constants,
+ (const float (*)[4])draw->pt.user.vs_constants,
  count,
  fpme->vertex_size,
  fpme->vertex_size);
+
+      if (geometry_shader)
+         draw_geometry_shader_run(geometry_shader,
+                                  (const float (*)[4])pipeline_verts->data,
+                                  (      float (*)[4])pipeline_verts->data,
+                                  (const float (*)[4])draw->pt.user.gs_constants,
+                                  count,
+                                  fpme->vertex_size,
+                                  fpme->vertex_size);
    }
 
    if (draw_pt_post_vs_run( fpme->post_vs,
@@ -270,6 +290,7 @@ static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle
    struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
    struct draw_vertex_shader *shader = draw->vs.vertex_shader;
+   struct draw_geometry_shader *geometry_shader = draw->gs.geometry_shader;
    unsigned opt = fpme->opt;
    unsigned alloc_count = align( count, 4 );
 
@@ -296,10 +317,19 @@ static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle
       shader->run_linear(shader,
  (const float (*)[4])pipeline_verts->data,
  (      float (*)[4])pipeline_verts->data,
- (const float (*)[4])draw->pt.user.constants,
+ (const float (*)[4])draw->pt.user.vs_constants,
  count,
  fpme->vertex_size,
  fpme->vertex_size);
+
+      if (geometry_shader)
+         draw_geometry_shader_run(geometry_shader,
+                                  (const float (*)[4])pipeline_verts->data,
+                                  (      float (*)[4])pipeline_verts->data,
+                                  (const float (*)[4])draw->pt.user.gs_constants,
+                                  count,
+                                  fpme->vertex_size,
+                                  fpme->vertex_size);
    }
 
    if (draw_pt_post_vs_run( fpme->post_vs,
diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
index 9dfb478..5515182 100644
--- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c
+++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
@@ -100,7 +100,7 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs,
    struct vertex_header *out = vertices;
    const float *scale = pvs->draw->viewport.scale;
    const float *trans = pvs->draw->viewport.translate;
-   const unsigned pos = pvs->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(pvs->draw);
    unsigned clipped = 0;
    unsigned j;
 
@@ -190,7 +190,7 @@ static boolean post_vs_viewport( struct pt_post_vs *pvs,
    struct vertex_header *out = vertices;
    const float *scale = pvs->draw->viewport.scale;
    const float *trans = pvs->draw->viewport.translate;
-   const unsigned pos = pvs->draw->vs.position_output;
+   const unsigned pos = draw_current_shader_position_output(pvs->draw);
    unsigned j;
 
    if (0) debug_printf("%s\n", __FUNCTION__);
diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c
index b61fa29..17c3b8c 100644
--- a/src/gallium/auxiliary/draw/draw_pt_util.c
+++ b/src/gallium/auxiliary/draw/draw_pt_util.c
@@ -50,16 +50,32 @@ void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr)
       *first = 2;
       *incr = 1;
       break;
+   case PIPE_PRIM_LINES_ADJACENCY:
+      *first = 4;
+      *incr = 2;
+      break;
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      *first = 4;
+      *incr = 1;
+      break;
    case PIPE_PRIM_TRIANGLES:
       *first = 3;
       *incr = 3;
       break;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+      *first = 6;
+      *incr = 3;
+      break;
    case PIPE_PRIM_TRIANGLE_STRIP:
    case PIPE_PRIM_TRIANGLE_FAN:
    case PIPE_PRIM_POLYGON:
       *first = 3;
       *incr = 1;
       break;
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      *first = 6;
+      *incr = 1;
+      break;
    case PIPE_PRIM_QUADS:
       *first = 4;
       *incr = 4;
diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
index 010c7a1..f0aec5f 100644
--- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
+++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h
@@ -36,6 +36,10 @@ static void FUNC(struct draw_pt_front_end *frontend,
    case PIPE_PRIM_TRIANGLE_STRIP:
    case PIPE_PRIM_QUADS:
    case PIPE_PRIM_QUAD_STRIP:
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
       for (j = 0; j < count;) {
          unsigned remaining = count - j;
          unsigned nr = trim( MIN2(varray->driver_fetch_max, remaining), first, incr );
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index 7ee567d..d166925 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -147,11 +147,12 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs,
                               temp_buffer,
                               temp_buffer,
-                              (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
+                              (const float (*)[4])vsvg->base.vs->draw->pt.user.vs_constants,
                               count,
                               temp_vertex_stride,
                               temp_vertex_stride);
 
+   /* FIXME: geometry shading? */
 
    if (vsvg->base.key.clip) {
       /* not really handling clipping, just do the rhw so we can
@@ -207,7 +208,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient,
    vsvg->base.vs->run_linear( vsvg->base.vs,
                               temp_buffer,
                               temp_buffer,
-                              (const float (*)[4])vsvg->base.vs->draw->pt.user.constants,
+                              (const float (*)[4])vsvg->base.vs->draw->pt.user.vs_constants,
                               count,
                               temp_vertex_stride,
                               temp_vertex_stride);
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index 5e7e5d2..bb4f564 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -122,7 +122,9 @@ static const char *semantic_names[] =
    "GENERIC",
    "NORMAL",
    "FACE",
-   "EDGEFLAG"
+   "EDGEFLAG",
+   "VERTICES_IN",
+   "PRIM_ID"
 };
 
 static const char *immediate_type_names[] =
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 123117c..e2f2878 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -372,6 +372,7 @@ tgsi_exec_machine_create( void )
    memset(mach, 0, sizeof(*mach));
 
    mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR];
+   mach->MaxGeometryShaderOutputs = TGSI_MAX_TOTAL_VERTICES;
    mach->Predicates = &mach->Temps[TGSI_EXEC_TEMP_P0];
 
    /* Setup constants. */
@@ -1468,6 +1469,15 @@ store_dest(
       index = mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0]
          + reg->Register.Index;
       dst = &mach->Outputs[offset + index].xyzw[chan_index];
+#if 0
+      if (TGSI_PROCESSOR_GEOMETRY == mach->Processor) {
+         fprintf(stderr, "STORING OUT[%d] mask(%d), = (", index, execmask);
+         for (i = 0; i < QUAD_SIZE; i++)
+            if (execmask & (1 << i))
+               fprintf(stderr, "%f, ", chan->f[i]);
+         fprintf(stderr, ")\n");
+      }
+#endif
       break;
 
    case TGSI_FILE_TEMPORARY:
@@ -1638,6 +1648,35 @@ exec_kilp(struct tgsi_exec_machine *mach,
    mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask;
 }
 
+static void
+emit_vertex(struct tgsi_exec_machine *mach)
+{
+   /* FIXME: check for exec mask correctly
+   unsigned i;
+   for (i = 0; i < QUAD_SIZE; ++i) {
+         if ((mach->ExecMask & (1 << i)))
+   */
+   if (mach->ExecMask) {
+      mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += mach->NumOutputs;
+      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++;
+   }
+}
+
+static void
+emit_primitive(struct tgsi_exec_machine *mach)
+{
+   unsigned *prim_count = &mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0];
+   /* FIXME: check for exec mask correctly
+   unsigned i;
+   for (i = 0; i < QUAD_SIZE; ++i) {
+         if ((mach->ExecMask & (1 << i)))
+   */
+   if (mach->ExecMask) {
+      ++(*prim_count);
+      debug_assert((*prim_count * mach->NumOutputs) < mach->MaxGeometryShaderOutputs);
+      mach->Primitives[*prim_count] = 0;
+   }
+}
 
 /*
  * Fetch a four texture samples using STR texture coordinates.
@@ -3087,13 +3126,11 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_EMIT:
-      mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16;
-      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++;
+      emit_vertex(mach);
       break;
 
    case TGSI_OPCODE_ENDPRIM:
-      mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++;
-      mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0;
+      emit_primitive(mach);
       break;
 
    case TGSI_OPCODE_BGNFOR:
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index fd94c1b..afaf5c3 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -191,6 +191,14 @@ struct tgsi_exec_labels
  */
 #define TGSI_EXEC_MAX_CONST_BUFFER  4096
 
+/* The maximum number of vertices per primitive */
+#define TGSI_MAX_PRIM_VERTICES 6
+
+/* The maximum number of primitives to be generated */
+#define TGSI_MAX_PRIMITIVES 64
+
+/* The maximum total number of vertices */
+#define TGSI_MAX_TOTAL_VERTICES (TGSI_MAX_PRIM_VERTICES * TGSI_MAX_PRIMITIVES * PIPE_MAX_ATTRIBS)
 
 /** function call/activation record */
 struct tgsi_call_record
@@ -201,7 +209,6 @@ struct tgsi_call_record
    uint ReturnAddr;
 };
 
-
 /**
  * Run-time virtual machine state for executing TGSI shader.
  */
@@ -214,8 +221,8 @@ struct tgsi_exec_machine
 
    float                         Imms[TGSI_EXEC_NUM_IMMEDIATES][4];
 
-   struct tgsi_exec_vector       Inputs[PIPE_MAX_ATTRIBS];
-   struct tgsi_exec_vector       Outputs[PIPE_MAX_ATTRIBS];
+   struct tgsi_exec_vector       Inputs[TGSI_MAX_PRIM_VERTICES * PIPE_MAX_ATTRIBS];
+   struct tgsi_exec_vector       Outputs[TGSI_MAX_TOTAL_VERTICES];
 
    struct tgsi_exec_vector       *Addrs;
    struct tgsi_exec_vector       *Predicates;
@@ -229,6 +236,8 @@ struct tgsi_exec_machine
 
    /* GEOMETRY processor only. */
    unsigned                      *Primitives;
+   unsigned                       NumOutputs;
+   unsigned                       MaxGeometryShaderOutputs;
 
    /* FRAGMENT processor only. */
    const struct tgsi_interp_coef *InterpCoefs;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index f000958..84e863b 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -791,7 +791,9 @@ static const char *semantic_names[TGSI_SEMANTIC_COUNT] =
    "PSIZE",
    "GENERIC",
    "NORMAL",
-   "FACE"
+   "FACE",
+   "VERTICES_IN",
+   "PRIM_ID"
 };
 
 static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] =
diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
index 5cc1d4d..01bea0f 100644
--- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
@@ -59,7 +59,7 @@ cell_map_constant_buffers(struct cell_context *sp)
       }
    }
 
-   draw_set_mapped_constant_buffer(sp->draw,
+   draw_set_mapped_constant_buffer(sp->draw, PIPE_SHADER_VERTEX,
                                    sp->mapped_constants[PIPE_SHADER_VERTEX],
                                    sp->constants[PIPE_SHADER_VERTEX].buffer->size);
 }
diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c
index efc4f78..b723e79 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_derived.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c
@@ -66,7 +66,7 @@ calculate_vertex_layout( struct cell_context *cell )
    vinfo->num_attribs = 0;
 
    /* we always want to emit vertex pos */
-   src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_POSITION, 0);
+   src = draw_find_shader_output(cell->draw, TGSI_SEMANTIC_POSITION, 0);
    assert(src >= 0);
    draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src);
 
@@ -82,14 +82,14 @@ calculate_vertex_layout( struct cell_context *cell )
          break;
 
       case TGSI_SEMANTIC_COLOR:
-         src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_COLOR,
+         src = draw_find_shader_output(cell->draw, TGSI_SEMANTIC_COLOR,
                                    fs->info.input_semantic_index[i]);
          assert(src >= 0);
          draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
          break;
 
       case TGSI_SEMANTIC_FOG:
-         src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_FOG, 0);
+         src = draw_find_shader_output(cell->draw, TGSI_SEMANTIC_FOG, 0);
 #if 1
          if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */
             src = 0;
@@ -100,7 +100,7 @@ calculate_vertex_layout( struct cell_context *cell )
 
       case TGSI_SEMANTIC_GENERIC:
          /* this includes texcoords and varying vars */
-         src = draw_find_vs_output(cell->draw, TGSI_SEMANTIC_GENERIC,
+         src = draw_find_shader_output(cell->draw, TGSI_SEMANTIC_GENERIC,
                               fs->info.input_semantic_index[i]);
          assert(src >= 0);
          draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c
index ac5fafe..5b87286 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_emit.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c
@@ -331,7 +331,7 @@ cell_emit_state(struct cell_context *cell)
       const struct draw_context *const draw = cell->draw;
       struct cell_shader_info info;
 
-      info.num_outputs = draw_num_vs_outputs(draw);
+      info.num_outputs = draw_num_shader_outputs(draw);
       info.declarations = (uintptr_t) draw->vs.machine.Declarations;
       info.num_declarations = draw->vs.machine.NumDeclarations;
       info.instructions = (uintptr_t) draw->vs.machine.Instructions;
diff --git a/src/gallium/drivers/i915/i915_context.c b/src/gallium/drivers/i915/i915_context.c
index 94c8aee..949f046 100644
--- a/src/gallium/drivers/i915/i915_context.c
+++ b/src/gallium/drivers/i915/i915_context.c
@@ -84,7 +84,7 @@ i915_draw_range_elements(struct pipe_context *pipe,
    }
 
 
-   draw_set_mapped_constant_buffer(draw,
+   draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX,
                                    i915->current.constants[PIPE_SHADER_VERTEX],
                                    (i915->current.num_user_constants[PIPE_SHADER_VERTEX] *
                                       4 * sizeof(float)));
diff --git a/src/gallium/drivers/i915/i915_state_derived.c b/src/gallium/drivers/i915/i915_state_derived.c
index 178d4e8..03dd509 100644
--- a/src/gallium/drivers/i915/i915_state_derived.c
+++ b/src/gallium/drivers/i915/i915_state_derived.c
@@ -84,7 +84,7 @@ static void calculate_vertex_layout( struct i915_context *i915 )
 
   
    /* pos */
-   src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_POSITION, 0);
+   src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_POSITION, 0);
    if (needW) {
       draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src);
       vinfo.hwfmt[0] |= S4_VFMT_XYZW;
@@ -101,21 +101,21 @@ static void calculate_vertex_layout( struct i915_context *i915 )
 
    /* primary color */
    if (colors[0]) {
-      src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_COLOR, 0);
+      src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_COLOR, 0);
       draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src);
       vinfo.hwfmt[0] |= S4_VFMT_COLOR;
    }
 
    /* secondary color */
    if (colors[1]) {
-      src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_COLOR, 1);
+      src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_COLOR, 1);
       draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src);
       vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG;
    }
 
    /* fog coord, not fog blend factor */
    if (fog) {
-      src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_FOG, 0);
+      src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_FOG, 0);
       draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
       vinfo.hwfmt[0] |= S4_VFMT_FOG_PARAM;
    }
@@ -125,7 +125,7 @@ static void calculate_vertex_layout( struct i915_context *i915 )
       uint hwtc;
       if (texCoords[i]) {
          hwtc = TEXCOORDFMT_4D;
-         src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_GENERIC, i);
+         src = draw_find_shader_output(i915->draw, TGSI_SEMANTIC_GENERIC, i);
          draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
       }
       else {
diff --git a/src/gallium/drivers/llvmpipe/lp_state_derived.c b/src/gallium/drivers/llvmpipe/lp_state_derived.c
index e703964..fdc6a38 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_derived.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_derived.c
@@ -66,7 +66,7 @@ llvmpipe_get_vertex_info(struct llvmpipe_context *llvmpipe)
       /* compute vertex layout now */
       const struct lp_fragment_shader *lpfs = llvmpipe->fs;
       struct vertex_info *vinfo_vbuf = &llvmpipe->vertex_info_vbuf;
-      const uint num = draw_num_vs_outputs(llvmpipe->draw);
+      const uint num = draw_current_shader_outputs(llvmpipe->draw);
       uint i;
 
       /* Tell draw_vbuf to simply emit the whole post-xform vertex
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 22683ff..af053f1 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -734,7 +734,8 @@ llvmpipe_set_constant_buffer(struct pipe_context *pipe,
    }
 
    if(shader == PIPE_SHADER_VERTEX) {
-      draw_set_mapped_constant_buffer(llvmpipe->draw, data, size);
+      draw_set_mapped_constant_buffer(llvmpipe->draw, PIPE_SHADER_VERTEX,
+                                      data, size);
    }
 
    llvmpipe->dirty |= LP_NEW_CONSTANTS;
diff --git a/src/gallium/drivers/nv04/nv04_vbo.c b/src/gallium/drivers/nv04/nv04_vbo.c
index e316781..099ab10 100644
--- a/src/gallium/drivers/nv04/nv04_vbo.c
+++ b/src/gallium/drivers/nv04/nv04_vbo.c
@@ -45,7 +45,7 @@ boolean nv04_draw_elements( struct pipe_context *pipe,
  draw_set_mapped_element_buffer(draw, 0, NULL);
  }
 
- draw_set_mapped_constant_buffer(draw,
+ draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX,
  nv04->constbuf[PIPE_SHADER_VERTEX],
  nv04->constbuf_nr[PIPE_SHADER_VERTEX]);
 
diff --git a/src/gallium/drivers/nv10/nv10_vbo.c b/src/gallium/drivers/nv10/nv10_vbo.c
index 441a4f7..0d26141 100644
--- a/src/gallium/drivers/nv10/nv10_vbo.c
+++ b/src/gallium/drivers/nv10/nv10_vbo.c
@@ -45,6 +45,7 @@ boolean nv10_draw_elements( struct pipe_context *pipe,
  }
 
  draw_set_mapped_constant_buffer(draw,
+                                        PIPE_SHADER_VERTEX,
  nv10->constbuf[PIPE_SHADER_VERTEX],
  nv10->constbuf_nr[PIPE_SHADER_VERTEX]);
 
diff --git a/src/gallium/drivers/nv20/nv20_state_emit.c b/src/gallium/drivers/nv20/nv20_state_emit.c
index 0122b1c..63cba1f 100644
--- a/src/gallium/drivers/nv20/nv20_state_emit.c
+++ b/src/gallium/drivers/nv20/nv20_state_emit.c
@@ -228,7 +228,7 @@ static void nv20_vertex_layout(struct nv20_context *nv20)
  }
 
  /* always do position */ {
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_POSITION, 0);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_POSITION, 0);
  draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src);
  vinfo->hwfmt[0] |= (1 << 0);
  }
@@ -237,19 +237,19 @@ static void nv20_vertex_layout(struct nv20_context *nv20)
  for (i = 4; i < 6; i++) {
  if (!generics[i])
  continue;
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_GENERIC, i);
  draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
  vinfo->hwfmt[0] |= (1 << (i - 3));
  }
 
  if (colors[0]) {
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_COLOR, 0);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_COLOR, 0);
  draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
  vinfo->hwfmt[0] |= (1 << 3);
  }
 
  if (colors[1]) {
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_COLOR, 1);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_COLOR, 1);
  draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src);
  vinfo->hwfmt[0] |= (1 << 4);
  }
@@ -258,7 +258,7 @@ static void nv20_vertex_layout(struct nv20_context *nv20)
  for (i = 6; i < 10; i++) {
  if (!generics[i])
  continue;
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_GENERIC, i);
  draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
  vinfo->hwfmt[0] |= (1 << (i - 1));
  }
@@ -267,7 +267,7 @@ static void nv20_vertex_layout(struct nv20_context *nv20)
  for (i = 0; i < 4; i++) {
  if (!generics[i])
  continue;
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_GENERIC, i);
  draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
  vinfo->hwfmt[0] |= (1 << (i + 9));
  }
@@ -276,13 +276,13 @@ static void nv20_vertex_layout(struct nv20_context *nv20)
  for (i = 10; i < 12; i++) {
  if (!generics[i])
  continue;
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_GENERIC, i);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_GENERIC, i);
  draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src);
  vinfo->hwfmt[0] |= (1 << (i + 3));
  }
 
  if (fog) {
- src = draw_find_vs_output(dc, TGSI_SEMANTIC_FOG, 0);
+ src = draw_find_shader_output(dc, TGSI_SEMANTIC_FOG, 0);
  draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src);
  vinfo->hwfmt[0] |= (1 << 15);
  }
diff --git a/src/gallium/drivers/nv20/nv20_vbo.c b/src/gallium/drivers/nv20/nv20_vbo.c
index 84d7db6..4bf461e 100644
--- a/src/gallium/drivers/nv20/nv20_vbo.c
+++ b/src/gallium/drivers/nv20/nv20_vbo.c
@@ -45,7 +45,7 @@ boolean nv20_draw_elements( struct pipe_context *pipe,
  draw_set_mapped_element_buffer(draw, 0, NULL);
  }
 
- draw_set_mapped_constant_buffer(draw,
+ draw_set_mapped_constant_buffer(draw, PIPE_SHADER_VERTEX,
  nv20->constbuf[PIPE_SHADER_VERTEX],
  nv20->constbuf_nr[PIPE_SHADER_VERTEX]);
 
diff --git a/src/gallium/drivers/nv40/nv40_draw.c b/src/gallium/drivers/nv40/nv40_draw.c
index b2f19ec..3875bc3 100644
--- a/src/gallium/drivers/nv40/nv40_draw.c
+++ b/src/gallium/drivers/nv40/nv40_draw.c
@@ -261,7 +261,8 @@ nv40_draw_elements_swtnl(struct pipe_context *pipe,
  map = pipe_buffer_map(pscreen,
       nv40->constbuf[PIPE_SHADER_VERTEX],
       PIPE_BUFFER_USAGE_CPU_READ);
- draw_set_mapped_constant_buffer(nv40->draw, map, nr);
+ draw_set_mapped_constant_buffer(nv40->draw, PIPE_SHADER_VERTEX,
+                                                map, nr);
  }
 
  draw_arrays(nv40->draw, mode, start, count);
@@ -285,7 +286,7 @@ static INLINE void
 emit_attrib(struct nv40_context *nv40, unsigned hw, unsigned emit,
     unsigned semantic, unsigned index)
 {
- unsigned draw_out = draw_find_vs_output(nv40->draw, semantic, index);
+ unsigned draw_out = draw_find_shader_output(nv40->draw, semantic, index);
  unsigned a = nv40->swtnl.nr_attribs++;
 
  nv40->swtnl.hw[a] = hw;
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index 2a33587..0c890cb 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -222,6 +222,10 @@ softpipe_create( struct pipe_screen *screen )
    softpipe->pipe.bind_vs_state   = softpipe_bind_vs_state;
    softpipe->pipe.delete_vs_state = softpipe_delete_vs_state;
 
+   softpipe->pipe.create_gs_state = softpipe_create_gs_state;
+   softpipe->pipe.bind_gs_state   = softpipe_bind_gs_state;
+   softpipe->pipe.delete_gs_state = softpipe_delete_gs_state;
+
    softpipe->pipe.set_blend_color = softpipe_set_blend_color;
    softpipe->pipe.set_clip_state = softpipe_set_clip_state;
    softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer;
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index 8ce20c5..159547e 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -58,6 +58,7 @@ struct softpipe_context {
    struct pipe_rasterizer_state *rasterizer;
    struct sp_fragment_shader *fs;
    struct sp_vertex_shader *vs;
+   struct sp_geometry_shader *gs;
 
    /** Other rendering state */
    struct pipe_blend_color blend_color;
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index 518ef88..9ea5d6f 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -48,7 +48,7 @@ static void
 softpipe_map_constant_buffers(struct softpipe_context *sp)
 {
    struct pipe_winsys *ws = sp->pipe.winsys;
-   uint i, size;
+   uint i, vssize, gssize;
 
    for (i = 0; i < PIPE_SHADER_TYPES; i++) {
       if (sp->constants[i].buffer && sp->constants[i].buffer->size)
@@ -57,13 +57,21 @@ softpipe_map_constant_buffers(struct softpipe_context *sp)
    }
 
    if (sp->constants[PIPE_SHADER_VERTEX].buffer)
-      size = sp->constants[PIPE_SHADER_VERTEX].buffer->size;
+      vssize = sp->constants[PIPE_SHADER_VERTEX].buffer->size;
    else
-      size = 0;
+      vssize = 0;
 
-   draw_set_mapped_constant_buffer(sp->draw,
+   if (sp->constants[PIPE_SHADER_GEOMETRY].buffer)
+      gssize = sp->constants[PIPE_SHADER_GEOMETRY].buffer->size;
+   else
+      gssize = 0;
+
+   draw_set_mapped_constant_buffer(sp->draw, PIPE_SHADER_VERTEX,
                                    sp->mapped_constants[PIPE_SHADER_VERTEX],
-                                   size);
+                                   vssize);
+   draw_set_mapped_constant_buffer(sp->draw, PIPE_SHADER_GEOMETRY,
+                                   sp->mapped_constants[PIPE_SHADER_GEOMETRY],
+                                   gssize);
 }
 
 
@@ -78,9 +86,10 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp)
     */
    draw_flush(sp->draw);
 
-   draw_set_mapped_constant_buffer(sp->draw, NULL, 0);
+   draw_set_mapped_constant_buffer(sp->draw, PIPE_SHADER_VERTEX, NULL, 0);
+   draw_set_mapped_constant_buffer(sp->draw, PIPE_SHADER_GEOMETRY, NULL, 0);
 
-   for (i = 0; i < 2; i++) {
+   for (i = 0; i < PIPE_SHADER_TYPES; i++) {
       if (sp->constants[i].buffer && sp->constants[i].buffer->size)
          ws->buffer_unmap(ws, sp->constants[i].buffer);
       sp->mapped_constants[i] = NULL;
diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index bd3532d..a32312d 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -89,6 +89,8 @@ softpipe_get_param(struct pipe_screen *screen, int param)
       return 13; /* max 4Kx4K */
    case PIPE_CAP_TGSI_CONT_SUPPORTED:
       return 1;
+   case PIPE_CAP_GEOMETRY_SHADER4:
+      return 1;
    case PIPE_CAP_BLEND_EQUATION_SEPARATE:
       return 1;
    default:
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
index 615581b..3da7536 100644
--- a/src/gallium/drivers/softpipe/sp_setup.c
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -1268,7 +1268,7 @@ void sp_setup_prepare( struct setup_context *setup )
    }
 
    /* Note: nr_attrs is only used for debugging (vertex printing) */
-   setup->nr_vertex_attrs = draw_num_vs_outputs(sp->draw);
+   setup->nr_vertex_attrs = draw_num_shader_outputs(sp->draw);
 
    sp->quad.first->begin( sp->quad.first );
 
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index 26d5c3f..5a32d21 100644
--- a/src/gallium/drivers/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -50,6 +50,7 @@
 #define SP_NEW_VERTEX        0x1000
 #define SP_NEW_VS            0x2000
 #define SP_NEW_QUERY         0x4000
+#define SP_NEW_GS            0x8000
 
 
 struct tgsi_sampler;
@@ -90,6 +91,11 @@ struct sp_vertex_shader {
    int max_sampler;             /* -1 if no samplers */
 };
 
+/** Subclass of pipe_shader_state */
+struct sp_geometry_shader {
+   struct pipe_shader_state shader;
+   struct draw_geometry_shader *draw_data;
+};
 
 
 void *
@@ -143,6 +149,10 @@ void *softpipe_create_vs_state(struct pipe_context *,
                                const struct pipe_shader_state *);
 void softpipe_bind_vs_state(struct pipe_context *, void *);
 void softpipe_delete_vs_state(struct pipe_context *, void *);
+void *softpipe_create_gs_state(struct pipe_context *,
+                               const struct pipe_shader_state *);
+void softpipe_bind_gs_state(struct pipe_context *, void *);
+void softpipe_delete_gs_state(struct pipe_context *, void *);
 
 void softpipe_set_polygon_stipple( struct pipe_context *,
   const struct pipe_poly_stipple * );
diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c
index c24a737..f6856a5 100644
--- a/src/gallium/drivers/softpipe/sp_state_derived.c
+++ b/src/gallium/drivers/softpipe/sp_state_derived.c
@@ -67,7 +67,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe)
       /* compute vertex layout now */
       const struct sp_fragment_shader *spfs = softpipe->fs;
       struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf;
-      const uint num = draw_num_vs_outputs(softpipe->draw);
+      const uint num = draw_current_shader_outputs(softpipe->draw);
       uint i;
 
       /* Tell draw_vbuf to simply emit the whole post-xform vertex
@@ -117,13 +117,13 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe)
          }
 
          /* this includes texcoords and varying vars */
-         src = draw_find_vs_output(softpipe->draw,
-                                   spfs->info.input_semantic_name[i],
-                                   spfs->info.input_semantic_index[i]);
+         src = draw_find_shader_output(softpipe->draw,
+                                       spfs->info.input_semantic_name[i],
+                                       spfs->info.input_semantic_index[i]);
          draw_emit_vertex_attr(vinfo, EMIT_4F, interp, src);
       }
 
-      softpipe->psize_slot = draw_find_vs_output(softpipe->draw,
+      softpipe->psize_slot = draw_find_shader_output(softpipe->draw,
                                                  TGSI_SEMANTIC_PSIZE, 0);
       if (softpipe->psize_slot > 0) {
          draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT,
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index b41f7e8..22f82b1 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -165,3 +165,62 @@ softpipe_set_constant_buffer(struct pipe_context *pipe,
 
    softpipe->dirty |= SP_NEW_CONSTANTS;
 }
+
+void *
+softpipe_create_gs_state(struct pipe_context *pipe,
+                         const struct pipe_shader_state *templ)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+   struct sp_geometry_shader *state;
+
+   state = CALLOC_STRUCT(sp_geometry_shader);
+   if (state == NULL )
+      goto fail;
+
+   /* copy shader tokens, the ones passed in will go away.
+    */
+   state->shader.tokens = tgsi_dup_tokens(templ->tokens);
+   if (state->shader.tokens == NULL)
+      goto fail;
+
+   state->draw_data = draw_create_geometry_shader(softpipe->draw, templ);
+   if (state->draw_data == NULL)
+      goto fail;
+
+   return state;
+
+fail:
+   if (state) {
+      FREE( (void *)state->shader.tokens );
+      FREE( state->draw_data );
+      FREE( state );
+   }
+   return NULL;
+}
+
+
+void
+softpipe_bind_gs_state(struct pipe_context *pipe, void *gs)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   softpipe->gs = (struct sp_geometry_shader *)gs;
+
+   draw_bind_geometry_shader(softpipe->draw,
+                             (softpipe->gs ? softpipe->gs->draw_data : NULL));
+
+   softpipe->dirty |= SP_NEW_GS;
+}
+
+
+void
+softpipe_delete_gs_state(struct pipe_context *pipe, void *gs)
+{
+   struct softpipe_context *softpipe = softpipe_context(pipe);
+
+   struct sp_geometry_shader *state =
+      (struct sp_geometry_shader *)gs;
+
+   draw_delete_geometry_shader(softpipe->draw, state->draw_data);
+   FREE(state);
+}
diff --git a/src/gallium/drivers/svga/svga_swtnl_draw.c b/src/gallium/drivers/svga/svga_swtnl_draw.c
index 8b14c91..7655121 100644
--- a/src/gallium/drivers/svga/svga_swtnl_draw.c
+++ b/src/gallium/drivers/svga/svga_swtnl_draw.c
@@ -90,7 +90,7 @@ svga_swtnl_draw_range_elements(struct svga_context *svga,
                             PIPE_BUFFER_USAGE_CPU_READ);
       assert(map);
       draw_set_mapped_constant_buffer(
-         draw,
+         draw, PIPE_SHADER_VERTEX,
          map,
          svga->curr.cb[PIPE_SHADER_VERTEX]->size);
    }
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index 11bcdc0..6c06fb9 100644
--- a/src/gallium/include/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -142,6 +142,12 @@ struct pipe_context {
                              const struct pipe_shader_state *);
    void   (*bind_vs_state)(struct pipe_context *, void *);
    void   (*delete_vs_state)(struct pipe_context *, void *);
+
+   void * (*create_gs_state)(struct pipe_context *,
+                             const struct pipe_shader_state *);
+   void   (*bind_gs_state)(struct pipe_context *, void *);
+   void   (*delete_gs_state)(struct pipe_context *, void *);
+
    /*@}*/
 
    /**
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index fe1390d..90fc333 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -321,23 +321,28 @@ enum pipe_transfer_usage {
  */
 #define PIPE_SHADER_VERTEX   0
 #define PIPE_SHADER_FRAGMENT 1
-#define PIPE_SHADER_TYPES    2
+#define PIPE_SHADER_GEOMETRY 2
+#define PIPE_SHADER_TYPES    3
 
 
 /**
  * Primitive types:
  */
-#define PIPE_PRIM_POINTS          0
-#define PIPE_PRIM_LINES           1
-#define PIPE_PRIM_LINE_LOOP       2
-#define PIPE_PRIM_LINE_STRIP      3
-#define PIPE_PRIM_TRIANGLES       4
-#define PIPE_PRIM_TRIANGLE_STRIP  5
-#define PIPE_PRIM_TRIANGLE_FAN    6
-#define PIPE_PRIM_QUADS           7
-#define PIPE_PRIM_QUAD_STRIP      8
-#define PIPE_PRIM_POLYGON         9
-#define PIPE_PRIM_MAX             10
+#define PIPE_PRIM_POINTS               0
+#define PIPE_PRIM_LINES                1
+#define PIPE_PRIM_LINE_LOOP            2
+#define PIPE_PRIM_LINE_STRIP           3
+#define PIPE_PRIM_TRIANGLES            4
+#define PIPE_PRIM_TRIANGLE_STRIP       5
+#define PIPE_PRIM_TRIANGLE_FAN         6
+#define PIPE_PRIM_QUADS                7
+#define PIPE_PRIM_QUAD_STRIP           8
+#define PIPE_PRIM_POLYGON              9
+#define PIPE_PRIM_LINES_ADJACENCY          10
+#define PIPE_PRIM_LINE_STRIP_ADJACENCY    11
+#define PIPE_PRIM_TRIANGLES_ADJACENCY      12
+#define PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY 13
+#define PIPE_PRIM_MAX                      14
 
 
 /**
@@ -393,6 +398,7 @@ enum pipe_transfer_usage {
 #define PIPE_CAP_MAX_PREDICATE_REGISTERS 30
 #define PIPE_CAP_MAX_COMBINED_SAMPLERS   31  /*< Maximum texture image units accessible from vertex
                                                  and fragment shaders combined */
+#define PIPE_CAP_GEOMETRY_SHADER4        32
 
 
 /**
diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h
index 5da85bb..3e7335b 100644
--- a/src/gallium/include/pipe/p_shader_tokens.h
+++ b/src/gallium/include/pipe/p_shader_tokens.h
@@ -121,16 +121,18 @@ struct tgsi_declaration_range
    unsigned Last    : 16; /**< UINT */
 };
 
-#define TGSI_SEMANTIC_POSITION 0
-#define TGSI_SEMANTIC_COLOR    1
-#define TGSI_SEMANTIC_BCOLOR   2 /**< back-face color */
-#define TGSI_SEMANTIC_FOG      3
-#define TGSI_SEMANTIC_PSIZE    4
-#define TGSI_SEMANTIC_GENERIC  5
-#define TGSI_SEMANTIC_NORMAL   6
-#define TGSI_SEMANTIC_FACE     7
-#define TGSI_SEMANTIC_EDGEFLAG 8
-#define TGSI_SEMANTIC_COUNT    9 /**< number of semantic values */
+#define TGSI_SEMANTIC_POSITION  0
+#define TGSI_SEMANTIC_COLOR     1
+#define TGSI_SEMANTIC_BCOLOR    2 /**< back-face color */
+#define TGSI_SEMANTIC_FOG       3
+#define TGSI_SEMANTIC_PSIZE     4
+#define TGSI_SEMANTIC_GENERIC   5
+#define TGSI_SEMANTIC_NORMAL    6
+#define TGSI_SEMANTIC_FACE      7
+#define TGSI_SEMANTIC_EDGEFLAG  8
+#define TGSI_SEMANTIC_VERTICES  9
+#define TGSI_SEMANTIC_PRIMID   10
+#define TGSI_SEMANTIC_COUNT    11 /**< number of semantic values */
 
 struct tgsi_declaration_semantic
 {
diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i
index 9728207..3c35e6f 100644
--- a/src/gallium/state_trackers/python/p_context.i
+++ b/src/gallium/state_trackers/python/p_context.i
@@ -103,6 +103,25 @@ struct st_context {
       $self->vs = vs;
    }
 
+   void set_geometry_shader( const struct pipe_shader_state *state ) {
+      void *gs;
+
+      if(!state) {
+         cso_set_geometry_shader_handle($self->cso, NULL);
+         return;
+      }
+
+      gs = $self->pipe->create_gs_state($self->pipe, state);
+      if(!gs)
+         return;
+
+      if(cso_set_geometry_shader_handle($self->cso, gs) != PIPE_OK)
+         return;
+
+      cso_delete_geometry_shader($self->cso, $self->gs);
+      $self->gs = gs;
+   }
+
    /*
     * Parameter-like state (or properties)
     */
diff --git a/src/gallium/state_trackers/python/samples/gs.py b/src/gallium/state_trackers/python/samples/gs.py
new file mode 100644
index 0000000..f4f083e
--- /dev/null
+++ b/src/gallium/state_trackers/python/samples/gs.py
@@ -0,0 +1,254 @@
+#!/usr/bin/env python
+##########################################################################
+#
+# Copyright 2009 VMware
+# All Rights Reserved.
+#
+# Permission is hereby granted, free of charge, to any person obtaining a
+# copy of this software and associated documentation files (the
+# "Software"), to deal in the Software without restriction, including
+# without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sub license, and/or sell copies of the Software, and to
+# permit persons to whom the Software is furnished to do so, subject to
+# the following conditions:
+#
+# The above copyright notice and this permission notice (including the
+# next paragraph) shall be included in all copies or substantial portions
+# of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+#
+##########################################################################
+
+
+from gallium import *
+
+
+def make_image(surface):
+    data = surface.get_tile_rgba8(0, 0, surface.width, surface.height)
+
+    import Image
+    outimage = Image.fromstring('RGBA', (surface.width, surface.height), data, "raw", 'RGBA', 0, 1)
+    return outimage
+
+def save_image(filename, surface):
+    outimage = make_image(surface)
+    outimage.save(filename, "PNG")
+
+def show_image(surface):
+    outimage = make_image(surface)
+
+    import Tkinter as tk
+    from PIL import Image, ImageTk
+    root = tk.Tk()
+
+    root.title('background image')
+
+    image1 = ImageTk.PhotoImage(outimage)
+    w = image1.width()
+    h = image1.height()
+    x = 100
+    y = 100
+    root.geometry("%dx%d+%d+%d" % (w, h, x, y))
+    panel1 = tk.Label(root, image=image1)
+    panel1.pack(side='top', fill='both', expand='yes')
+    panel1.image = image1
+    root.mainloop()
+
+
+def test(dev):
+    ctx = dev.context_create()
+
+    width = 255
+    height = 255
+    minz = 0.0
+    maxz = 1.0
+
+    # disabled blending/masking
+    blend = Blend()
+    blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE
+    blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE
+    blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO
+    blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO
+    blend.colormask = PIPE_MASK_RGBA
+    ctx.set_blend(blend)
+
+    # depth/stencil/alpha
+    depth_stencil_alpha = DepthStencilAlpha()
+    depth_stencil_alpha.depth.enabled = 1
+    depth_stencil_alpha.depth.writemask = 1
+    depth_stencil_alpha.depth.func = PIPE_FUNC_LESS
+    ctx.set_depth_stencil_alpha(depth_stencil_alpha)
+
+    # rasterizer
+    rasterizer = Rasterizer()
+    rasterizer.front_winding = PIPE_WINDING_CW
+    rasterizer.cull_mode = PIPE_WINDING_NONE
+    rasterizer.scissor = 1
+    ctx.set_rasterizer(rasterizer)
+
+    # viewport
+    viewport = Viewport()
+    scale = FloatArray(4)
+    scale[0] = width / 2.0
+    scale[1] = -height / 2.0
+    scale[2] = (maxz - minz) / 2.0
+    scale[3] = 1.0
+    viewport.scale = scale
+    translate = FloatArray(4)
+    translate[0] = width / 2.0
+    translate[1] = height / 2.0
+    translate[2] = (maxz - minz) / 2.0
+    translate[3] = 0.0
+    viewport.translate = translate
+    ctx.set_viewport(viewport)
+
+    # samplers
+    sampler = Sampler()
+    sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+    sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+    sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE
+    sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE
+    sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST
+    sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST
+    sampler.normalized_coords = 1
+    ctx.set_sampler(0, sampler)
+
+    # scissor
+    scissor = Scissor()
+    scissor.minx = 0
+    scissor.miny = 0
+    scissor.maxx = width
+    scissor.maxy = height
+    ctx.set_scissor(scissor)
+
+    clip = Clip()
+    clip.nr = 0
+    ctx.set_clip(clip)
+
+    # framebuffer
+    cbuf = dev.texture_create(
+        PIPE_FORMAT_X8R8G8B8_UNORM,
+        width, height,
+        tex_usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET,
+    ).get_surface()
+    zbuf = dev.texture_create(
+        PIPE_FORMAT_Z16_UNORM,
+        width, height,
+        tex_usage=PIPE_TEXTURE_USAGE_DEPTH_STENCIL,
+    ).get_surface()
+    fb = Framebuffer()
+    fb.width = width
+    fb.height = height
+    fb.nr_cbufs = 1
+    fb.set_cbuf(0, cbuf)
+    fb.set_zsbuf(zbuf)
+    ctx.set_framebuffer(fb)
+    rgba = FloatArray(4);
+    rgba[0] = 0.0
+    rgba[1] = 0.0
+    rgba[2] = 0.0
+    rgba[3] = 0.0
+    ctx.clear(PIPE_CLEAR_COLOR | PIPE_CLEAR_DEPTHSTENCIL, rgba, 1.0, 0xff)
+
+    # vertex shader
+    vs = Shader('''
+        VERT
+        DCL IN[0], POSITION, CONSTANT
+        DCL IN[1], COLOR, CONSTANT
+        DCL OUT[0], POSITION, CONSTANT
+        DCL OUT[1], COLOR, CONSTANT
+        0:MOV OUT[0], IN[0]
+        1:MOV OUT[1], IN[1]
+        2:END
+    ''')
+    ctx.set_vertex_shader(vs)
+
+    gs = Shader('''
+        GEOM
+        PROPERTY GS_INPUT_PRIMITIVE TRIANGLES
+        PROPERTY GS_OUTPUT_PRIMITIVE TRIANGLE_STRIP
+        DCL IN[][0], POSITION, CONSTANT
+        DCL IN[][1], COLOR, CONSTANT
+        DCL OUT[0], POSITION, CONSTANT
+        DCL OUT[1], COLOR, CONSTANT
+        0:MOV OUT[0], IN[0][0]
+        1:MOV OUT[1], IN[0][1]
+        2:EMIT_VERTEX
+        3:MOV OUT[0], IN[1][0]
+        4:MOV OUT[1], IN[1][1]
+        5:EMIT_VERTEX
+        6:MOV OUT[0], IN[2][0]
+        7:MOV OUT[1], IN[2][1]
+        8:EMIT_VERTEX
+        9:END_PRIMITIVE
+        10:END
+    ''')
+    ctx.set_geometry_shader(gs)
+
+    # fragment shader
+    fs = Shader('''
+        FRAG
+        DCL IN[0], COLOR, LINEAR
+        DCL OUT[0], COLOR, CONSTANT
+        0:MOV OUT[0], IN[0]
+        1:END
+    ''')
+    ctx.set_fragment_shader(fs)
+
+    nverts = 3
+    nattrs = 2
+    verts = FloatArray(nverts * nattrs * 4)
+
+    verts[ 0] =   0.0 # x1
+    verts[ 1] =   0.8 # y1
+    verts[ 2] =   0.2 # z1
+    verts[ 3] =   1.0 # w1
+    verts[ 4] =   1.0 # r1
+    verts[ 5] =   0.0 # g1
+    verts[ 6] =   0.0 # b1
+    verts[ 7] =   1.0 # a1
+    verts[ 8] =  -0.8 # x2
+    verts[ 9] =  -0.8 # y2
+    verts[10] =   0.5 # z2
+    verts[11] =   1.0 # w2
+    verts[12] =   0.0 # r2
+    verts[13] =   1.0 # g2
+    verts[14] =   0.0 # b2
+    verts[15] =   1.0 # a2
+    verts[16] =   0.8 # x3
+    verts[17] =  -0.8 # y3
+    verts[18] =   0.8 # z3
+    verts[19] =   1.0 # w3
+    verts[20] =   0.0 # r3
+    verts[21] =   0.0 # g3
+    verts[22] =   1.0 # b3
+    verts[23] =   1.0 # a3
+
+    ctx.draw_vertices(PIPE_PRIM_TRIANGLES,
+                      nverts,
+                      nattrs,
+                      verts)
+
+    ctx.flush()
+
+    show_image(cbuf)
+    #show_image(zbuf)
+    #save_image('cbuf.png', cbuf)
+    #save_image('zbuf.png', zbuf)
+
+
+
+def main():
+    dev = Device()
+    test(dev)
+
+
+if __name__ == '__main__':
+    main()
diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h
index a246b6a..2a7a323 100644
--- a/src/gallium/state_trackers/python/st_device.h
+++ b/src/gallium/state_trackers/python/st_device.h
@@ -57,6 +57,7 @@ struct st_context {
   
    void *vs;
    void *fs;
+   void *gs;
 
    struct pipe_texture *default_texture;
    struct pipe_texture *sampler_textures[PIPE_MAX_SAMPLERS];
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index d793f82..cfc0caa 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -241,7 +241,8 @@ st_feedback_draw_vbo(GLcontext *ctx,
    mapped_constants = pipe_buffer_map(pipe->screen,
                                       st->state.constants[PIPE_SHADER_VERTEX].buffer,
                                       PIPE_BUFFER_USAGE_CPU_READ);
-   draw_set_mapped_constant_buffer(st->draw, mapped_constants,
+   draw_set_mapped_constant_buffer(st->draw, PIPE_SHADER_VERTEX,
+                                   mapped_constants,
                                    st->state.constants[PIPE_SHADER_VERTEX].buffer->size);
 
 
--
1.6.5.7

---

[0002-tgsi-make-the-tgsi-assembly-parser-report-line-colum.patch]

---

From da5a9010381c9f064d01a10cc88ae0d274bc300a Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Tue, 15 Dec 2009 09:26:51 -0500
Subject: [PATCH 02/12] tgsi: make the tgsi assembly parser report line/column on error

---
 src/gallium/auxiliary/tgsi/tgsi_text.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index 84e863b..9673686 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -197,7 +197,20 @@ struct translate_ctx
 
 static void report_error( struct translate_ctx *ctx, const char *msg )
 {
-   debug_printf( "\nError: %s", msg );
+   int line = 1;
+   int column = 1;
+   const char *itr = ctx->text;
+
+   while (itr != ctx->cur) {
+      if (*itr == '\n') {
+         column = 1;
+         ++line;
+      }
+      ++column;
+      ++itr;
+   }
+
+   debug_printf( "\nTGSI asm error: %s [%d : %d] \n", msg, line, column );
 }
 
 /* Parse shader header.
--
1.6.5.7

---

[0004-softpipe-gs-don-t-crash-with-null-shader.patch]

---

From 9a9b712a7489d91f06d1d82f2b53e9013ba69aee Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 12:34:34 -0500
Subject: [PATCH 04/12] softpipe/gs: don't crash with null shader

---
 src/gallium/drivers/softpipe/sp_state_fs.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index 22f82b1..b904bde 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -221,6 +221,7 @@ softpipe_delete_gs_state(struct pipe_context *pipe, void *gs)
    struct sp_geometry_shader *state =
       (struct sp_geometry_shader *)gs;
 
-   draw_delete_geometry_shader(softpipe->draw, state->draw_data);
+   draw_delete_geometry_shader(softpipe->draw,
+                               (state) ? state->draw_data : 0);
    FREE(state);
 }
--
1.6.5.7

---

[0006-python-gs-fix-the-example.patch]

---

From 9f6fdd790281a8d24f247ce011cf60837b1f4f18 Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 12:40:19 -0500
Subject: [PATCH 06/12] python/gs: fix the example

---
 src/gallium/state_trackers/python/samples/gs.py |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/src/gallium/state_trackers/python/samples/gs.py b/src/gallium/state_trackers/python/samples/gs.py
index f4f083e..1ceead5 100644
--- a/src/gallium/state_trackers/python/samples/gs.py
+++ b/src/gallium/state_trackers/python/samples/gs.py
@@ -180,14 +180,14 @@ def test(dev):
         DCL OUT[1], COLOR, CONSTANT
         0:MOV OUT[0], IN[0][0]
         1:MOV OUT[1], IN[0][1]
-        2:EMIT_VERTEX
+        2:EMIT
         3:MOV OUT[0], IN[1][0]
         4:MOV OUT[1], IN[1][1]
-        5:EMIT_VERTEX
+        5:EMIT
         6:MOV OUT[0], IN[2][0]
         7:MOV OUT[1], IN[2][1]
-        8:EMIT_VERTEX
-        9:END_PRIMITIVE
+        8:EMIT
+        9:ENDPRIM
         10:END
     ''')
     ctx.set_geometry_shader(gs)
--
1.6.5.7

---

[0007-tgsi-add-missing-support-for-two-dimensional-arrays-.patch]

---

From ab800dd005e8f4a0f372462d055c5a526d2d1379 Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 12:40:39 -0500
Subject: [PATCH 07/12] tgsi: add missing support for two dimensional arrays in various places

in particular asm text parsing and sanity checking were missing code
to handle multi-dimensional arrays/geometry shaders
---
 src/gallium/auxiliary/tgsi/tgsi_dump.c   |    6 +
 src/gallium/auxiliary/tgsi/tgsi_sanity.c |  284 +++++++++++++++++++------
 src/gallium/auxiliary/tgsi/tgsi_text.c   |  351 ++++++++++++++++++++----------
 3 files changed, 465 insertions(+), 176 deletions(-)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index bb4f564..1f6b1d8 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -183,6 +183,12 @@ _dump_register(
    int last )
 {
    ENM( file, file_names );
+
+   /* all geometry shader inputs are two dimensional */
+   if (file == TGSI_FILE_INPUT &&
+       ctx->iter.processor.Processor == TGSI_PROCESSOR_GEOMETRY)
+      TXT("[]");
+
    CHR( '[' );
    SID( first );
    if (first != last) {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index c27579e..5d11c19 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -26,32 +26,112 @@
  **************************************************************************/
 
 #include "util/u_debug.h"
+#include "util/u_memory.h"
+#include "pipe/p_inlines.h"
+#include "cso_cache/cso_hash.h"
 #include "tgsi_sanity.h"
 #include "tgsi_info.h"
 #include "tgsi_iterate.h"
 
-typedef uint reg_flag;
-
-#define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8)
-
-#define MAX_REGISTERS 1024
-#define MAX_REG_FLAGS ((MAX_REGISTERS + BITS_IN_REG_FLAG - 1) / BITS_IN_REG_FLAG)
+typedef struct {
+   uint file : 28;
+   /* max 2 dimensions */
+   uint dimensions : 4;
+   uint indices[2];
+} scan_register;
 
 struct sanity_check_ctx
 {
    struct tgsi_iterate_context iter;
+   struct cso_hash *regs_decl;
+   struct cso_hash *regs_used;
+   struct cso_hash *regs_ind_used;
 
-   reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REG_FLAGS];
-   reg_flag regs_used[TGSI_FILE_COUNT][MAX_REG_FLAGS];
-   boolean regs_ind_used[TGSI_FILE_COUNT];
    uint num_imms;
    uint num_instructions;
    uint index_of_END;
 
    uint errors;
    uint warnings;
+   uint implied_array_size;
 };
 
+static INLINE unsigned
+scan_register_key(const scan_register *reg)
+{
+   unsigned key = reg->file;
+   key |= (reg->indices[0] << 4);
+   key |= (reg->indices[1] << 18);
+
+   return key;
+}
+
+static void
+fill_scan_register1d(scan_register *reg,
+                     uint file, uint index)
+{
+   reg->file = file;
+   reg->dimensions = 1;
+   reg->indices[0] = index;
+   reg->indices[1] = 0;
+}
+
+static void
+fill_scan_register2d(scan_register *reg,
+                     uint file, uint index1, uint index2)
+{
+   reg->file = file;
+   reg->dimensions = 2;
+   reg->indices[0] = index1;
+   reg->indices[1] = index2;
+}
+
+static void
+scan_register_dst(scan_register *reg,
+                  struct tgsi_full_dst_register *dst)
+{
+   fill_scan_register1d(reg,
+                        dst->Register.File,
+                        dst->Register.Index);
+}
+
+static void
+scan_register_src(scan_register *reg,
+                  struct tgsi_full_src_register *src)
+{
+   if (src->Register.Dimension) {
+      /*FIXME: right now we don't support indirect
+       * multidimensional addressing */
+      debug_assert(!src->Dimension.Indirect);
+      fill_scan_register2d(reg,
+                           src->Register.File,
+                           src->Register.Index,
+                           src->Dimension.Index);
+   } else {
+      fill_scan_register1d(reg,
+                           src->Register.File,
+                           src->Register.Index);
+   }
+}
+
+static scan_register *
+create_scan_register_src(struct tgsi_full_src_register *src)
+{
+   scan_register *reg = MALLOC(sizeof(scan_register));
+   scan_register_src(reg, src);
+
+   return reg;
+}
+
+static scan_register *
+create_scan_register_dst(struct tgsi_full_dst_register *dst)
+{
+   scan_register *reg = MALLOC(sizeof(scan_register));
+   scan_register_dst(reg, dst);
+
+   return reg;
+}
+
 static void
 report_error(
    struct sanity_check_ctx *ctx,
@@ -99,12 +179,12 @@ check_file_name(
 static boolean
 is_register_declared(
    struct sanity_check_ctx *ctx,
-   uint file,
-   int index )
+   const scan_register *reg)
 {
-   assert( index >= 0 && index < MAX_REGISTERS );
-
-   return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
+   void *data = cso_hash_find_data_from_template(
+      ctx->regs_decl, scan_register_key(reg),
+      (void*)reg, sizeof(scan_register));
+   return  data ? TRUE : FALSE;
 }
 
 static boolean
@@ -112,23 +192,37 @@ is_any_register_declared(
    struct sanity_check_ctx *ctx,
    uint file )
 {
-   uint i;
+   struct cso_hash_iter iter =
+      cso_hash_first_node(ctx->regs_decl);
 
-   for (i = 0; i < MAX_REG_FLAGS; i++)
-      if (ctx->regs_decl[file][i])
+   while (cso_hash_iter_is_null(iter)) {
+      scan_register *reg = (scan_register *)cso_hash_iter_data(iter);
+      if (reg->file == file)
          return TRUE;
+      iter = cso_hash_iter_next(iter);
+   }
+
    return FALSE;
 }
 
 static boolean
 is_register_used(
    struct sanity_check_ctx *ctx,
-   uint file,
-   int index )
+   scan_register *reg)
 {
-   assert( index < MAX_REGISTERS );
+   void *data = cso_hash_find_data_from_template(
+      ctx->regs_used, scan_register_key(reg),
+      reg, sizeof(scan_register));
+   return  data ? TRUE : FALSE;
+}
 
-   return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE;
+
+static boolean
+is_ind_register_used(
+   struct sanity_check_ctx *ctx,
+   scan_register *reg)
+{
+   return cso_hash_contains(ctx->regs_ind_used, reg->file);
 }
 
 static const char *file_names[TGSI_FILE_COUNT] =
@@ -148,31 +242,40 @@ static const char *file_names[TGSI_FILE_COUNT] =
 static boolean
 check_register_usage(
    struct sanity_check_ctx *ctx,
-   uint file,
-   int index,
+   scan_register *reg,
    const char *name,
    boolean indirect_access )
 {
-   if (!check_file_name( ctx, file ))
+   if (!check_file_name( ctx, reg->file )) {
+      free(reg);
       return FALSE;
+   }
 
    if (indirect_access) {
       /* Note that 'index' is an offset relative to the value of the
-       * address register.  No range checking done here.
-       */
-      if (!is_any_register_declared( ctx, file ))
-         report_error( ctx, "%s: Undeclared %s register", file_names[file], name );
-      ctx->regs_ind_used[file] = TRUE;
+       * address register.  No range checking done here.*/
+      reg->indices[0] = 0;
+      reg->indices[1] = 0;
+      if (!is_any_register_declared( ctx, reg->file ))
+         report_error( ctx, "%s: Undeclared %s register", file_names[reg->file], name );
+      if (!is_ind_register_used(ctx, reg))
+         cso_hash_insert(ctx->regs_ind_used, reg->file, reg);
+      else
+         free(reg);
    }
    else {
-      if (index < 0 || index >= MAX_REGISTERS) {
-         report_error( ctx, "%s[%d]: Invalid %s index", file_names[file], index, name );
-         return FALSE;
-      }
-
-      if (!is_register_declared( ctx, file, index ))
-         report_error( ctx, "%s[%d]: Undeclared %s register", file_names[file], index, name );
-      ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG));
+      if (!is_register_declared( ctx, reg )) {
+         if (reg->dimensions == 2)
+            report_error( ctx, "%s[%d][%d]: Undeclared %s register", file_names[reg->file],
+                          reg->indices[0], reg->indices[1], name );
+         else
+            report_error( ctx, "%s[%d]: Undeclared %s register", file_names[reg->file],
+                          reg->indices[0], name );
+         }
+      if (!is_register_used( ctx, reg ))
+         cso_hash_insert(ctx->regs_used, scan_register_key(reg), reg);
+      else
+         free(reg);
    }
    return TRUE;
 }
@@ -210,33 +313,33 @@ iter_instruction(
     * Mark the registers as used.
     */
    for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+      scan_register *reg = create_scan_register_dst(&inst->Dst[i]);
       check_register_usage(
          ctx,
-         inst->Dst[i].Register.File,
-         inst->Dst[i].Register.Index,
+         reg,
          "destination",
          FALSE );
    }
    for (i = 0; i < inst->Instruction.NumSrcRegs; i++) {
+      scan_register *reg = create_scan_register_src(&inst->Src[i]);
       check_register_usage(
          ctx,
-         inst->Src[i].Register.File,
-         inst->Src[i].Register.Index,
+         reg,
          "source",
          (boolean)inst->Src[i].Register.Indirect );
       if (inst->Src[i].Register.Indirect) {
-         uint file;
-         int index;
+         scan_register *ind_reg = MALLOC(sizeof(scan_register));
 
-         file = inst->Src[i].Indirect.File;
-         index = inst->Src[i].Indirect.Index;
+         fill_scan_register1d(ind_reg,
+                              inst->Src[i].Indirect.File,
+                              inst->Src[i].Indirect.Index);
          check_register_usage(
             ctx,
-            file,
-            index,
+            reg,
             "indirect",
             FALSE );
-         if (!(file == TGSI_FILE_ADDRESS || file == TGSI_FILE_LOOP) || index != 0) {
+         if (!(reg->file == TGSI_FILE_ADDRESS || reg->file == TGSI_FILE_LOOP) ||
+             reg->indices[0] != 0) {
             report_warning(ctx, "Indirect register neither ADDR[0] nor LOOP[0]");
          }
       }
@@ -266,6 +369,19 @@ iter_instruction(
    return TRUE;
 }
 
+static void
+check_and_declare(struct sanity_check_ctx *ctx,
+                  scan_register *reg)
+{
+   if (is_register_declared( ctx, reg))
+      report_error( ctx, "%s[%u]: The same register declared more than once",
+                    file_names[reg->file], reg->indices[0] );
+   cso_hash_insert(ctx->regs_decl,
+                   scan_register_key(reg),
+                   reg);
+}
+
+
 static boolean
 iter_declaration(
    struct tgsi_iterate_context *iter,
@@ -287,9 +403,21 @@ iter_declaration(
    if (!check_file_name( ctx, file ))
       return TRUE;
    for (i = decl->Range.First; i <= decl->Range.Last; i++) {
-      if (is_register_declared( ctx, file, i ))
-         report_error( ctx, "%s[%u]: The same register declared more than once", file_names[file], i );
-      ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG));
+      /* declared TGSI_FILE_INPUT's for geometry processor
+       * have an implied second dimension */
+      if (file == TGSI_FILE_INPUT &&
+          ctx->iter.processor.Processor == TGSI_PROCESSOR_GEOMETRY) {
+         uint vert;
+         for (vert = 0; vert < ctx->implied_array_size; ++vert) {
+            scan_register *reg = MALLOC(sizeof(scan_register));
+            fill_scan_register2d(reg, file, vert, i);
+            check_and_declare(ctx, reg);
+         }
+      } else {
+         scan_register *reg = MALLOC(sizeof(scan_register));
+         fill_scan_register1d(reg, file, i);
+         check_and_declare(ctx, reg);
+      }
    }
 
    return TRUE;
@@ -301,8 +429,7 @@ iter_immediate(
    struct tgsi_full_immediate *imm )
 {
    struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
-
-   assert( ctx->num_imms < MAX_REGISTERS );
+   scan_register *reg;
 
    /* No immediates allowed after the first instruction.
     */
@@ -311,7 +438,9 @@ iter_immediate(
 
    /* Mark the register as declared.
     */
-   ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG));
+   reg = MALLOC(sizeof(scan_register));
+   fill_scan_register1d(reg, TGSI_FILE_IMMEDIATE, ctx->num_imms);
+   cso_hash_insert(ctx->regs_decl, scan_register_key(reg), reg);
    ctx->num_imms++;
 
    /* Check data type validity.
@@ -330,8 +459,13 @@ iter_property(
    struct tgsi_iterate_context *iter,
    struct tgsi_full_property *prop )
 {
-   /*struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;*/
+   struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
 
+   if (iter->processor.Processor == TGSI_PROCESSOR_GEOMETRY &&
+       prop->Property.PropertyName == TGSI_PROPERTY_GS_INPUT_PRIM) {
+      ctx->implied_array_size =
+         pipe_vertices_per_primitive(prop->u[0].Data);
+   }
    return TRUE;
 }
 
@@ -340,7 +474,6 @@ epilog(
    struct tgsi_iterate_context *iter )
 {
    struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter;
-   uint file;
 
    /* There must be an END instruction somewhere.
     */
@@ -350,13 +483,17 @@ epilog(
 
    /* Check if all declared registers were used.
     */
-   for (file = TGSI_FILE_NULL; file < TGSI_FILE_COUNT; file++) {
-      uint i;
-
-      for (i = 0; i < MAX_REGISTERS; i++) {
-         if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) {
-            report_warning( ctx, "%s[%u]: Register never used", file_names[file], i );
+   {
+      struct cso_hash_iter iter =
+         cso_hash_first_node(ctx->regs_decl);
+
+      while (cso_hash_iter_is_null(iter)) {
+         scan_register *reg = (scan_register *)cso_hash_iter_data(iter);
+         if (!is_register_used(ctx, reg) && !is_ind_register_used(ctx, reg)) {
+            report_warning( ctx, "%s[%u]: Register never used",
+                            file_names[reg->file], reg->indices[0] );
          }
+         iter = cso_hash_iter_next(iter);
       }
    }
 
@@ -368,6 +505,18 @@ epilog(
    return TRUE;
 }
 
+static void
+regs_hash_destroy(struct cso_hash *hash)
+{
+   struct cso_hash_iter iter = cso_hash_first_node(hash);
+   while (!cso_hash_iter_is_null(iter)) {
+      scan_register *reg = (scan_register *)cso_hash_iter_data(iter);
+      iter = cso_hash_erase(hash, iter);
+      free(reg);
+   }
+   cso_hash_delete(hash);
+}
+
 boolean
 tgsi_sanity_check(
    const struct tgsi_token *tokens )
@@ -381,18 +530,23 @@ tgsi_sanity_check(
    ctx.iter.iterate_property = iter_property;
    ctx.iter.epilog = epilog;
 
-   memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) );
-   memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) );
-   memset( ctx.regs_ind_used, 0, sizeof( ctx.regs_ind_used ) );
+   ctx.regs_decl = cso_hash_create();
+   ctx.regs_used = cso_hash_create();
+   ctx.regs_ind_used = cso_hash_create();
+
    ctx.num_imms = 0;
    ctx.num_instructions = 0;
    ctx.index_of_END = ~0;
 
    ctx.errors = 0;
    ctx.warnings = 0;
+   ctx.implied_array_size = 0;
 
    if (!tgsi_iterate_shader( tokens, &ctx.iter ))
       return FALSE;
 
+   regs_hash_destroy(ctx.regs_decl);
+   regs_hash_destroy(ctx.regs_used);
+   regs_hash_destroy(ctx.regs_ind_used);
    return ctx.errors == 0;
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index 9673686..80d3a1e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -28,6 +28,7 @@
 #include "util/u_debug.h"
 #include "util/u_memory.h"
 #include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
 #include "tgsi_text.h"
 #include "tgsi_build.h"
 #include "tgsi_info.h"
@@ -193,6 +194,8 @@ struct translate_ctx
    struct tgsi_token *tokens_cur;
    struct tgsi_token *tokens_end;
    struct tgsi_header *header;
+   unsigned processor : 4;
+   int implied_array_size : 5;
 };
 
 static void report_error( struct translate_ctx *ctx, const char *msg )
@@ -242,6 +245,7 @@ static boolean parse_header( struct translate_ctx *ctx )
    if (ctx->tokens_cur >= ctx->tokens_end)
       return FALSE;
    *(struct tgsi_processor *) ctx->tokens_cur++ = tgsi_build_processor( processor, ctx->header );
+   ctx->processor = processor;
 
    return TRUE;
 }
@@ -338,92 +342,36 @@ parse_opt_writemask(
    return TRUE;
 }
 
-/* <register_file_bracket> ::= <file> `['
- */
 static boolean
-parse_register_file_bracket(
-   struct translate_ctx *ctx,
-   uint *file )
-{
-   if (!parse_file( &ctx->cur, file )) {
-      report_error( ctx, "Unknown register file" );
-      return FALSE;
-   }
-   eat_opt_white( &ctx->cur );
-   if (*ctx->cur != '[') {
-      report_error( ctx, "Expected `['" );
-      return FALSE;
-   }
-   ctx->cur++;
-   return TRUE;
-}
+parse_register_dst( struct translate_ctx *ctx,
+                    uint *file,
+                    int *index );
 
-/* <register_file_bracket_index> ::= <register_file_bracket> <uint>
- */
-static boolean
-parse_register_file_bracket_index(
-   struct translate_ctx *ctx,
-   uint *file,
-   int *index )
-{
-   uint uindex;
+struct parsed_src_bracket {
+   int index;
 
-   if (!parse_register_file_bracket( ctx, file ))
-      return FALSE;
-   eat_opt_white( &ctx->cur );
-   if (!parse_uint( &ctx->cur, &uindex )) {
-      report_error( ctx, "Expected literal unsigned integer" );
-      return FALSE;
-   }
-   *index = (int) uindex;
-   return TRUE;
-}
+   uint ind_file;
+   int ind_index;
+   uint ind_comp;
+};
 
-/* Parse destination register operand.
- *    <register_dst> ::= <register_file_bracket_index> `]'
- */
-static boolean
-parse_register_dst(
-   struct translate_ctx *ctx,
-   uint *file,
-   int *index )
-{
-   if (!parse_register_file_bracket_index( ctx, file, index ))
-      return FALSE;
-   eat_opt_white( &ctx->cur );
-   if (*ctx->cur != ']') {
-      report_error( ctx, "Expected `]'" );
-      return FALSE;
-   }
-   ctx->cur++;
-   return TRUE;
-}
 
-/* Parse source register operand.
- *    <register_src> ::= <register_file_bracket_index> `]' |
- *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `]' |
- *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `+' <uint> `]' |
- *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `-' <uint> `]'
- */
 static boolean
-parse_register_src(
+parse_register_src_bracket(
    struct translate_ctx *ctx,
-   uint *file,
-   int *index,
-   uint *ind_file,
-   int *ind_index,
-   uint *ind_comp)
+   struct parsed_src_bracket *brackets)
 {
    const char *cur;
    uint uindex;
 
-   *ind_comp = TGSI_SWIZZLE_X;
-   if (!parse_register_file_bracket( ctx, file ))
-      return FALSE;
+   memset(brackets, 0, sizeof(struct parsed_src_bracket));
+
    eat_opt_white( &ctx->cur );
+
    cur = ctx->cur;
-   if (parse_file( &cur, ind_file )) {
-      if (!parse_register_dst( ctx, ind_file, ind_index ))
+   if (parse_file( &cur, &brackets->ind_file )) {
+      if (!parse_register_dst( ctx, &brackets->ind_file,
+                               &brackets->ind_index ))
          return FALSE;
       eat_opt_white( &ctx->cur );
 
@@ -433,16 +381,16 @@ parse_register_src(
 
          switch (uprcase(*ctx->cur)) {
          case 'X':
-            *ind_comp = TGSI_SWIZZLE_X;
+            brackets->ind_comp = TGSI_SWIZZLE_X;
             break;
          case 'Y':
-            *ind_comp = TGSI_SWIZZLE_Y;
+            brackets->ind_comp = TGSI_SWIZZLE_Y;
             break;
          case 'Z':
-            *ind_comp = TGSI_SWIZZLE_Z;
+            brackets->ind_comp = TGSI_SWIZZLE_Z;
             break;
          case 'W':
-            *ind_comp = TGSI_SWIZZLE_W;
+            brackets->ind_comp = TGSI_SWIZZLE_W;
             break;
          default:
             report_error(ctx, "Expected indirect register swizzle component `x', `y', `z' or `w'");
@@ -463,12 +411,12 @@ parse_register_src(
             return FALSE;
          }
          if (negate)
-            *index = -(int) uindex;
+            brackets->index = -(int) uindex;
          else
-            *index = (int) uindex;
+            brackets->index = (int) uindex;
       }
       else {
-         *index = 0;
+         brackets->index = 0;
       }
    }
    else {
@@ -476,9 +424,9 @@ parse_register_src(
          report_error( ctx, "Expected literal unsigned integer" );
          return FALSE;
       }
-      *index = (int) uindex;
-      *ind_file = TGSI_FILE_NULL;
-      *ind_index = 0;
+      brackets->index = (int) uindex;
+      brackets->ind_file = TGSI_FILE_NULL;
+      brackets->ind_index = 0;
    }
    eat_opt_white( &ctx->cur );
    if (*ctx->cur != ']') {
@@ -489,20 +437,123 @@ parse_register_src(
    return TRUE;
 }
 
-/* Parse register declaration.
- *    <register_dcl> ::= <register_file_bracket_index> `]' |
- *                       <register_file_bracket_index> `..' <index> `]'
+static boolean
+parse_opt_register_src_bracket(
+   struct translate_ctx *ctx,
+   struct parsed_src_bracket *brackets,
+   int *parsed_brackets)
+{
+   const char *cur = ctx->cur;
+
+   *parsed_brackets = 0;
+
+   eat_opt_white( &cur );
+   if (cur[0] == '[') {
+      ++cur;
+      ctx->cur = cur;
+
+      if (!parse_register_src_bracket(ctx, brackets))
+         return FALSE;
+
+      *parsed_brackets = 1;
+   }
+
+   return TRUE;
+}
+
+/* <register_file_bracket> ::= <file> `['
  */
 static boolean
-parse_register_dcl(
+parse_register_file_bracket(
+   struct translate_ctx *ctx,
+   uint *file )
+{
+   if (!parse_file( &ctx->cur, file )) {
+      report_error( ctx, "Unknown register file" );
+      return FALSE;
+   }
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != '[') {
+      report_error( ctx, "Expected `['" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
+
+/* <register_file_bracket_index> ::= <register_file_bracket> <uint>
+ */
+static boolean
+parse_register_file_bracket_index(
    struct translate_ctx *ctx,
    uint *file,
-   int *first,
-   int *last )
+   int *index )
 {
-   if (!parse_register_file_bracket_index( ctx, file, first ))
+   uint uindex;
+
+   if (!parse_register_file_bracket( ctx, file ))
       return FALSE;
    eat_opt_white( &ctx->cur );
+   if (!parse_uint( &ctx->cur, &uindex )) {
+      report_error( ctx, "Expected literal unsigned integer" );
+      return FALSE;
+   }
+   *index = (int) uindex;
+   return TRUE;
+}
+
+/* Parse source register operand.
+ *    <register_src> ::= <register_file_bracket_index> `]' |
+ *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `]' |
+ *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `+' <uint> `]' |
+ *                       <register_file_bracket> <register_dst> [`.' (`x' | `y' | `z' | `w')] `-' <uint> `]'
+ */
+static boolean
+parse_register_src(
+   struct translate_ctx *ctx,
+   uint *file,
+   struct parsed_src_bracket *brackets)
+{
+
+   brackets->ind_comp = TGSI_SWIZZLE_X;
+   if (!parse_register_file_bracket( ctx, file ))
+      return FALSE;
+   if (!parse_register_src_bracket( ctx, brackets ))
+       return FALSE;
+
+   return TRUE;
+}
+
+struct parsed_dcl_bracket {
+   uint first;
+   uint last;
+};
+
+static boolean
+parse_register_dcl_bracket(
+   struct translate_ctx *ctx,
+   struct parsed_dcl_bracket *bracket)
+{
+   uint uindex;
+   memset(bracket, 0, sizeof(struct parsed_dcl_bracket));
+
+   eat_opt_white( &ctx->cur );
+
+   if (!parse_uint( &ctx->cur, &uindex )) {
+      /* it can be an empty bracket [] which means its range
+       * is from 0 to some implied size */
+      if (ctx->cur[0] == ']' && ctx->implied_array_size != 0) {
+         bracket->first = 0;
+         bracket->last = ctx->implied_array_size - 1;
+         goto cleanup;
+      }
+      report_error( ctx, "Expected literal unsigned integer" );
+      return FALSE;
+   }
+   bracket->first = (int) uindex;
+
+   eat_opt_white( &ctx->cur );
+
    if (ctx->cur[0] == '.' && ctx->cur[1] == '.') {
       uint uindex;
 
@@ -512,12 +563,14 @@ parse_register_dcl(
          report_error( ctx, "Expected literal integer" );
          return FALSE;
       }
-      *last = (int) uindex;
+      bracket->last = (int) uindex;
       eat_opt_white( &ctx->cur );
    }
    else {
-      *last = *first;
+      bracket->last = bracket->first;
    }
+
+cleanup:
    if (*ctx->cur != ']') {
       report_error( ctx, "Expected `]' or `..'" );
       return FALSE;
@@ -526,6 +579,70 @@ parse_register_dcl(
    return TRUE;
 }
 
+/* Parse register declaration.
+ *    <register_dcl> ::= <register_file_bracket_index> `]' |
+ *                       <register_file_bracket_index> `..' <index> `]'
+ */
+static boolean
+parse_register_dcl(
+   struct translate_ctx *ctx,
+   uint *file,
+   struct parsed_dcl_bracket *brackets,
+   int *num_brackets)
+{
+   const char *cur;
+
+   *num_brackets = 0;
+
+   if (!parse_register_file_bracket( ctx, file ))
+      return FALSE;
+   if (!parse_register_dcl_bracket( ctx, &brackets[0] ))
+      return FALSE;
+
+   *num_brackets = 1;
+
+   cur = ctx->cur;
+   eat_opt_white( &cur );
+
+   if (cur[0] == '[') {
+      ++cur;
+      ctx->cur = cur;
+      if (!parse_register_dcl_bracket( ctx, &brackets[1] ))
+         return FALSE;
+      /* for geometry shader we don't really care about
+       * the first brackets it's always the size of the
+       * input primitive. so we want to declare just
+       * the index relevant to the semantics which is in
+       * the second bracket */
+      if (ctx->processor == TGSI_PROCESSOR_GEOMETRY) {
+         brackets[0] = brackets[1];
+      }
+      *num_brackets = 2;
+   }
+
+   return TRUE;
+}
+
+
+/* Parse destination register operand.
+ *    <register_dst> ::= <register_file_bracket_index> `]'
+ */
+static boolean
+parse_register_dst(
+   struct translate_ctx *ctx,
+   uint *file,
+   int *index )
+{
+   if (!parse_register_file_bracket_index( ctx, file, index ))
+      return FALSE;
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur != ']') {
+      report_error( ctx, "Expected `]'" );
+      return FALSE;
+   }
+   ctx->cur++;
+   return TRUE;
+}
 
 static boolean
 parse_dst_operand(
@@ -595,37 +712,44 @@ parse_src_operand(
    struct tgsi_full_src_register *src )
 {
    uint file;
-   int index;
-   uint ind_file;
-   int ind_index;
-   uint ind_comp;
    uint swizzle[4];
    boolean parsed_swizzle;
+   struct parsed_src_bracket bracket[2];
+   int parsed_opt_brackets;
 
    if (*ctx->cur == '-') {
       ctx->cur++;
       eat_opt_white( &ctx->cur );
       src->Register.Negate = 1;
    }
-  
+
    if (*ctx->cur == '|') {
       ctx->cur++;
       eat_opt_white( &ctx->cur );
       src->Register.Absolute = 1;
    }
 
-   if (!parse_register_src(ctx, &file, &index, &ind_file, &ind_index, &ind_comp))
+   if (!parse_register_src(ctx, &file, &bracket[0]))
+      return FALSE;
+   if (!parse_opt_register_src_bracket(ctx, &bracket[1], &parsed_opt_brackets))
       return FALSE;
+
    src->Register.File = file;
-   src->Register.Index = index;
-   if (ind_file != TGSI_FILE_NULL) {
+   src->Register.Index = bracket[0].index;
+   if (bracket[0].ind_file != TGSI_FILE_NULL) {
       src->Register.Indirect = 1;
-      src->Indirect.File = ind_file;
-      src->Indirect.Index = ind_index;
-      src->Indirect.SwizzleX = ind_comp;
-      src->Indirect.SwizzleY = ind_comp;
-      src->Indirect.SwizzleZ = ind_comp;
-      src->Indirect.SwizzleW = ind_comp;
+      src->Indirect.File = bracket[0].ind_file;
+      src->Indirect.Index = bracket[0].ind_index;
+      src->Indirect.SwizzleX = bracket[0].ind_comp;
+      src->Indirect.SwizzleY = bracket[0].ind_comp;
+      src->Indirect.SwizzleZ = bracket[0].ind_comp;
+      src->Indirect.SwizzleW = bracket[0].ind_comp;
+   }
+   if (parsed_opt_brackets) {
+      src->Register.Dimension = 1;
+      src->Dimension.Indirect = 0;
+      src->Dimension.Dimension = 0;
+      src->Dimension.Index = bracket[1].index;
    }
 
    /* Parse optional swizzle.
@@ -820,8 +944,8 @@ static boolean parse_declaration( struct translate_ctx *ctx )
 {
    struct tgsi_full_declaration decl;
    uint file;
-   int first;
-   int last;
+   struct parsed_dcl_bracket brackets[2];
+   int num_brackets;
    uint writemask;
    const char *cur;
    uint advance;
@@ -833,7 +957,7 @@ static boolean parse_declaration( struct translate_ctx *ctx )
       report_error( ctx, "Syntax error" );
       return FALSE;
    }
-   if (!parse_register_dcl( ctx, &file, &first, &last ))
+   if (!parse_register_dcl( ctx, &file, brackets, &num_brackets))
       return FALSE;
    if (!parse_opt_writemask( ctx, &writemask ))
       return FALSE;
@@ -841,8 +965,8 @@ static boolean parse_declaration( struct translate_ctx *ctx )
    decl = tgsi_default_full_declaration();
    decl.Declaration.File = file;
    decl.Declaration.UsageMask = writemask;
-   decl.Range.First = first;
-   decl.Range.Last = last;
+   decl.Range.First = brackets[0].first;
+   decl.Range.Last = brackets[0].last;
 
    cur = ctx->cur;
    eat_opt_white( &cur );
@@ -1059,6 +1183,11 @@ static boolean parse_property( struct translate_ctx *ctx )
          report_error( ctx, "Unknown primitive name as property!" );
          return FALSE;
       }
+      if (property_name == TGSI_PROPERTY_GS_INPUT_PRIM &&
+          ctx->processor == TGSI_PROCESSOR_GEOMETRY) {
+         ctx->implied_array_size =
+            pipe_vertices_per_primitive(values[0]);
+      }
       break;
    default:
       if (!parse_uint(&ctx->cur, &values[0] )) {
--
1.6.5.7

---

[0005-gallium-add-an-inline-that-returns-number-of-vertice.patch]

---

From 2b83aff8e805ac2bd8ba50660b54091288971e5c Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 12:39:21 -0500
Subject: [PATCH 05/12] gallium: add an inline that returns number of vertices per primitive

---
 src/gallium/include/pipe/p_inlines.h |   32 ++++++++++++++++++++++++++++++++
 1 files changed, 32 insertions(+), 0 deletions(-)

diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h
index 5fbd62a..95ec55d 100644
--- a/src/gallium/include/pipe/p_inlines.h
+++ b/src/gallium/include/pipe/p_inlines.h
@@ -192,6 +192,38 @@ pipe_transfer_buffer_flags( struct pipe_transfer *transf )
    }
 }
 
+static INLINE unsigned
+pipe_vertices_per_primitive(int primitive)
+{
+   switch(primitive) {
+   case PIPE_PRIM_POINTS:
+      return 1;
+   case PIPE_PRIM_LINES:
+   case PIPE_PRIM_LINE_LOOP:
+   case PIPE_PRIM_LINE_STRIP:
+      return 2;
+   case PIPE_PRIM_TRIANGLES:
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+      return 3;
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      return 4;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      return 6;
+
+   /* following primitives should never be used
+    * with geometry shaders */
+   case PIPE_PRIM_POLYGON:
+   case PIPE_PRIM_QUADS:
+   case PIPE_PRIM_QUAD_STRIP:
+   default:
+      debug_printf("Unrecognized geometry shader primitive");
+      return 3;
+   }
+}
+
 #ifdef __cplusplus
 }
 #endif
--
1.6.5.7

---

[0008-softpipe-make-it-possible-to-dump-geometry-shader-fr.patch]

---

From b545d1eded8ba05a0bd058a40081950b20cc28d1 Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 13:30:12 -0500
Subject: [PATCH 08/12] softpipe: make it possible to dump geometry shader from the softpipe

---
 src/gallium/drivers/softpipe/sp_context.c  |    1 +
 src/gallium/drivers/softpipe/sp_context.h  |    1 +
 src/gallium/drivers/softpipe/sp_state_fs.c |    4 ++++
 3 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index 0c890cb..82173a3 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -191,6 +191,7 @@ softpipe_create( struct pipe_screen *screen )
 #endif
 
    softpipe->dump_fs = debug_get_bool_option( "GALLIUM_DUMP_FS", FALSE );
+   softpipe->dump_gs = debug_get_bool_option( "SOFTPIPE_DUMP_GS", FALSE );
 
    softpipe->pipe.winsys = screen->winsys;
    softpipe->pipe.screen = screen;
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index 159547e..6a89bd4 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -148,6 +148,7 @@ struct softpipe_context {
 
    unsigned use_sse : 1;
    unsigned dump_fs : 1;
+   unsigned dump_gs : 1;
    unsigned no_rast : 1;
 };
 
diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c
index b904bde..acec62d 100644
--- a/src/gallium/drivers/softpipe/sp_state_fs.c
+++ b/src/gallium/drivers/softpipe/sp_state_fs.c
@@ -177,6 +177,10 @@ softpipe_create_gs_state(struct pipe_context *pipe,
    if (state == NULL )
       goto fail;
 
+   /* debug */
+   if (softpipe->dump_gs)
+      tgsi_dump(templ->tokens, 0);
+
    /* copy shader tokens, the ones passed in will go away.
     */
    state->shader.tokens = tgsi_dup_tokens(templ->tokens);
--
1.6.5.7

---

[0009-python-fix-python-state-tracker-after-some-latest-ch.patch]

---

From 76b3f69fd5d25d22483e8461e79b0f40796fe6fa Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 13:43:30 -0500
Subject: [PATCH 09/12] python: fix python state tracker after some latest changes

---
 src/gallium/state_trackers/python/gallium.i   |    1 +
 src/gallium/state_trackers/python/p_texture.i |    8 ++++----
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i
index 8e323f4..96b13c2 100644
--- a/src/gallium/state_trackers/python/gallium.i
+++ b/src/gallium/state_trackers/python/gallium.i
@@ -46,6 +46,7 @@
 #include "util/u_draw_quad.h"
 #include "util/u_tile.h"
 #include "util/u_math.h"
+#include "util/u_format.h"
 #include "util/u_memory.h"
 #include "tgsi/tgsi_text.h"
 #include "tgsi/tgsi_dump.h"
diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i
index 1de7f86..761587d 100644
--- a/src/gallium/state_trackers/python/p_texture.i
+++ b/src/gallium/state_trackers/python/p_texture.i
@@ -132,8 +132,8 @@ struct st_surface
       struct pipe_transfer *transfer;
       unsigned stride;
 
-      stride = pf_get_stride(texture->format, w);
-      *LENGTH = pf_get_nblocksy(texture->format, h) * stride;
+      stride = util_format_get_stride(texture->format, w);
+      *LENGTH = util_format_get_nblocksy(texture->format, h) * stride;
       *STRING = (char *) malloc(*LENGTH);
       if(!*STRING)
          return;
@@ -159,9 +159,9 @@ struct st_surface
       struct pipe_transfer *transfer;
     
       if(stride == 0)
-         stride = pf_get_stride(texture->format, w);
+         stride = util_format_get_stride(texture->format, w);
       
-      if(LENGTH < pf_get_nblocksy(texture->format, h) * stride)
+      if(LENGTH < util_format_get_nblocksy(texture->format, h) * stride)
          SWIG_exception(SWIG_ValueError, "offset must be smaller than buffer size");
         
       transfer = screen->get_tex_transfer(screen,
--
1.6.5.7

---

[0010-tgsi-fix-property-parsing-building.patch]

---

From 6df70e18ff045062dac25ffaa1f8788365ba117d Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 14:39:16 -0500
Subject: [PATCH 10/12] tgsi: fix property parsing/building

---
 src/gallium/auxiliary/tgsi/tgsi_build.c |    2 +-
 src/gallium/auxiliary/tgsi/tgsi_text.c  |   10 ++++++----
 2 files changed, 7 insertions(+), 5 deletions(-)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c
index 92903fe..de9cbc8 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_build.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_build.c
@@ -1025,7 +1025,7 @@ tgsi_build_full_property(
    size++;
 
    *property = tgsi_build_property(
-      TGSI_PROPERTY_GS_INPUT_PRIM,
+      full_prop->Property.PropertyName,
       header );
 
    assert( full_prop->Property.NrTokens <= 8 + 1 );
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index 80d3a1e..ca247a1 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -62,18 +62,20 @@ static boolean uprcase( char c )
 }
 
 /*
- * Ignore case of str1 and assume str2 is already uppercase.
+ * Ignore case of str1 and assume str1 is already uppercase.
  * Return TRUE iff str1 and str2 are equal.
  */
 static int
 streq_nocase_uprcase(const char *str1,
                      const char *str2)
 {
-   while (*str1 && uprcase(*str1) == *str2) {
+   while (*str1 && *str2) {
+      if (*str1 != uprcase(*str2))
+         return FALSE;
       str1++;
       str2++;
    }
-   return *str1 == *str2;
+   return TRUE;
 }
 
 static boolean str_match_no_case( const char **pcur, const char *str )
@@ -1166,7 +1168,7 @@ static boolean parse_property( struct translate_ctx *ctx )
    }
    for (property_name = 0; property_name < TGSI_PROPERTY_COUNT;
         ++property_name) {
-      if (streq_nocase_uprcase(id, property_names[property_name])) {
+      if (streq_nocase_uprcase(property_names[property_name], id)) {
          break;
       }
    }
--
1.6.5.7

---

[0011-tgsi-dump-the-indices-correctly-when-dealing-with-2d.patch]

---

From 44b94c491c62ab868df5748bc3a2df9f3f04f81f Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Wed, 23 Dec 2009 15:41:47 -0500
Subject: [PATCH 11/12] tgsi: dump the indices correctly when dealing with 2d arrays

---
 src/gallium/auxiliary/tgsi/tgsi_dump.c |   70 +++++++++++++++++++++++---------
 1 files changed, 50 insertions(+), 20 deletions(-)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index 1f6b1d8..5bfe019 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -176,7 +176,7 @@ static const char *primitive_names[] =
 
 
 static void
-_dump_register(
+_dump_register_decl(
    struct dump_ctx *ctx,
    uint file,
    int first,
@@ -199,6 +199,52 @@ _dump_register(
 }
 
 static void
+_dump_register_dst(
+   struct dump_ctx *ctx,
+   uint file,
+   int index)
+{
+   ENM( file, file_names );
+
+   CHR( '[' );
+   SID( index );
+   CHR( ']' );
+}
+
+
+static void
+_dump_register_src(
+   struct dump_ctx *ctx,
+   const struct tgsi_full_src_register *src )
+{
+   if (src->Register.Indirect) {
+      ENM( src->Register.File, file_names );
+      CHR( '[' );
+      ENM( src->Indirect.File, file_names );
+      CHR( '[' );
+      SID( src->Indirect.Index );
+      TXT( "]." );
+      ENM( src->Indirect.SwizzleX, swizzle_names );
+      if (src->Register.Index != 0) {
+         if (src->Register.Index > 0)
+            CHR( '+' );
+         SID( src->Register.Index );
+      }
+      CHR( ']' );
+   } else {
+      ENM( src->Register.File, file_names );
+      CHR( '[' );
+      SID( src->Register.Index );
+      CHR( ']' );
+   }
+   if (src->Register.Dimension) {
+      CHR( '[' );
+      SID( src->Dimension.Index );
+      CHR( ']' );
+   }
+}
+
+static void
 _dump_register_ind(
    struct dump_ctx *ctx,
    uint file,
@@ -252,7 +298,7 @@ iter_declaration(
 
    TXT( "DCL " );
 
-   _dump_register(
+   _dump_register_decl(
       ctx,
       decl->Declaration.File,
       decl->Range.First,
@@ -443,10 +489,9 @@ iter_instruction(
             dst->Indirect.SwizzleX );
       }
       else {
-         _dump_register(
+         _dump_register_dst(
             ctx,
             dst->Register.File,
-            dst->Register.Index,
             dst->Register.Index );
       }
       _dump_writemask( ctx, dst->Register.WriteMask );
@@ -466,22 +511,7 @@ iter_instruction(
       if (src->Register.Absolute)
          CHR( '|' );
 
-      if (src->Register.Indirect) {
-         _dump_register_ind(
-            ctx,
-            src->Register.File,
-            src->Register.Index,
-            src->Indirect.File,
-            src->Indirect.Index,
-            src->Indirect.SwizzleX );
-      }
-      else {
-         _dump_register(
-            ctx,
-            src->Register.File,
-            src->Register.Index,
-            src->Register.Index );
-      }
+      _dump_register_src(ctx, src);
 
       if (src->Register.SwizzleX != TGSI_SWIZZLE_X ||
           src->Register.SwizzleY != TGSI_SWIZZLE_Y ||
--
1.6.5.7

---

[0012-draw-tgsi-fix-geometry-shader-input-output-swizzling.patch]

---

From 180ce736f28e4dca79b9cc96b725d0360dd77fc0 Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Thu, 24 Dec 2009 08:15:29 -0500
Subject: [PATCH 12/12] draw/tgsi: fix geometry shader input/output swizzling

---
 src/gallium/auxiliary/draw/draw_gs.c   |    1 +
 src/gallium/auxiliary/tgsi/tgsi_exec.c |   10 +++++++++-
 2 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
index 1154228..0a742cb 100644
--- a/src/gallium/auxiliary/draw/draw_gs.c
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -210,6 +210,7 @@ static void draw_fetch_geometry_input(struct draw_geometry_shader *shader,
                const float (*input)[4];
                input = (const float (*)[4])(
                   (const char *)prim_ptr + (k * input_vertex_stride));
+               vidx = k * TGSI_EXEC_MAX_INPUT_ATTRIBS + slot;
                debug_printf("\t%d)(%d) Input vert:\n", vidx, k);
 #if 1
                assert(!util_is_inf_or_nan(input[vs_slot][0]));
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index e2f2878..ba89f2f 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -292,6 +292,14 @@ tgsi_exec_machine_bind_shader(
                                    * sizeof(struct tgsi_full_declaration));
             maxDeclarations += 10;
          }
+         if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_OUTPUT) {
+            unsigned reg;
+            for (reg = parse.FullToken.FullDeclaration.Range.First;
+                 reg <= parse.FullToken.FullDeclaration.Range.Last;
+                 ++reg) {
+               ++mach->NumOutputs;
+            }
+         }
          memcpy(declarations + numDeclarations,
                 &parse.FullToken.FullDeclaration,
                 sizeof(declarations[0]));
@@ -1471,7 +1479,7 @@ store_dest(
       dst = &mach->Outputs[offset + index].xyzw[chan_index];
 #if 0
       if (TGSI_PROCESSOR_GEOMETRY == mach->Processor) {
-         fprintf(stderr, "STORING OUT[%d] mask(%d), = (", index, execmask);
+         fprintf(stderr, "STORING OUT[%d] mask(%d), = (", offset + index, execmask);
          for (i = 0; i < QUAD_SIZE; i++)
             if (execmask & (1 << i))
                fprintf(stderr, "%f, ", chan->f[i]);
--
1.6.5.7

---

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Keith Whitwell-3]

Zack -- Impressive!  

Only one comment on patch #5 -- this is probably a util function rather than something for include/pipe.  In particular we already have util/u_prim.h which has some similar functions in it -- I'd put this function there if possible...

Keith
________________________________________
From: Zack Rusin [zackr@...]
Sent: Thursday, December 24, 2009 5:24 AM
To: mesa3d-dev
Cc: Keith Whitwell; Michal Krol
Subject: geometry shading patches

yo,

after our discussions i hacked a new version of geometry shading support in
gallium. the new geometry shading syntax looks as follows:
        GEOM
        PROPERTY GS_INPUT_PRIMITIVE TRIANGLES
        PROPERTY GS_OUTPUT_PRIMITIVE TRIANGLE_STRIP
        DCL IN[][0], POSITION, CONSTANT
        DCL IN[][1], COLOR, CONSTANT
        DCL OUT[0], POSITION, CONSTANT
        DCL OUT[1], COLOR, CONSTANT
        0:MOV OUT[0], IN[0][0]
        1:MOV OUT[1], IN[0][1]
        2:EMIT
        3:MOV OUT[0], IN[1][0]
        4:MOV OUT[1], IN[1][1]
        5:EMIT
        6:MOV OUT[0], IN[2][0]
        7:MOV OUT[1], IN[2][1]
        8:EMIT
        9:ENDPRIM
        10:END
the attached patch series adds support for all of it throughout gallium
(including support for 2dimensional arrays in tgsi_sanity checking, tgsi_dump
and tgsi_text).
Right now don't support switching of primitives in the draw module (e.g. input
primitive being points and output primitive being a triangle_strip) but i'd
like to do that after committing the attached changes (also the semantic for
vertices which is a leftover from an old patch should be removed then if it's
ok with you).

Keith, most relevant patches for you are 0001 and 0005.
Michal the most relevant to you are 0007, 0010, 0011 and 0012.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Keith Whitwell-3]

Zack,

In terms of advertising support for this - I wonder if this isn't something we should just turn on for all drivers, given that in the worst case it is just a vertex path fallback, and a lot of drivers will be hitting those for various reasons anyway.

Keith
________________________________________
From: Zack Rusin [zackr@...]
Sent: Thursday, December 24, 2009 5:24 AM
To: mesa3d-dev
Cc: Keith Whitwell; Michal Krol
Subject: geometry shading patches

yo,

after our discussions i hacked a new version of geometry shading support in
gallium. the new geometry shading syntax looks as follows:
        GEOM
        PROPERTY GS_INPUT_PRIMITIVE TRIANGLES
        PROPERTY GS_OUTPUT_PRIMITIVE TRIANGLE_STRIP
        DCL IN[][0], POSITION, CONSTANT
        DCL IN[][1], COLOR, CONSTANT
        DCL OUT[0], POSITION, CONSTANT
        DCL OUT[1], COLOR, CONSTANT
        0:MOV OUT[0], IN[0][0]
        1:MOV OUT[1], IN[0][1]
        2:EMIT
        3:MOV OUT[0], IN[1][0]
        4:MOV OUT[1], IN[1][1]
        5:EMIT
        6:MOV OUT[0], IN[2][0]
        7:MOV OUT[1], IN[2][1]
        8:EMIT
        9:ENDPRIM
        10:END
the attached patch series adds support for all of it throughout gallium
(including support for 2dimensional arrays in tgsi_sanity checking, tgsi_dump
and tgsi_text).
Right now don't support switching of primitives in the draw module (e.g. input
primitive being points and output primitive being a triangle_strip) but i'd
like to do that after committing the attached changes (also the semantic for
vertices which is a leftover from an old patch should be removed then if it's
ok with you).

Keith, most relevant patches for you are 0001 and 0005.
Michal the most relevant to you are 0007, 0010, 0011 and 0012.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Zack Rusin-2]

On Thursday 24 December 2009 09:09:44 Keith Whitwell wrote:
> Zack,
>
> In terms of advertising support for this - I wonder if this isn't something
>  we should just turn on for all drivers, given that in the worst case it is
>  just a vertex path fallback, and a lot of drivers will be hitting those
>  for various reasons anyway.

Yes, I completely agree with both of your emails. The attached patches
implement that (plus the last one comments out some unnecessary debugging
output). They do it on top of the previous patches but if you'd like to have a
history clear of them I can try to redo the entire series.

z

[0013-util-put-vertices_per_primitive-function-in-its-prop.patch]

---

From 0483e3ed1c28982857da3292f8247388e8f9d0d9 Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Thu, 24 Dec 2009 09:20:45 -0500
Subject: [PATCH 13/15] util: put vertices_per_primitive function in its proper location

---
 src/gallium/auxiliary/tgsi/tgsi_sanity.c |    5 +--
 src/gallium/auxiliary/tgsi/tgsi_text.c   |    4 +-
 src/gallium/auxiliary/util/u_prim.h      |   33 ++++++++++++++++++++++++++++++
 src/gallium/include/pipe/p_inlines.h     |   32 -----------------------------
 4 files changed, 37 insertions(+), 37 deletions(-)

diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index 5d11c19..16b8ec6 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -27,7 +27,7 @@
 
 #include "util/u_debug.h"
 #include "util/u_memory.h"
-#include "pipe/p_inlines.h"
+#include "util/u_prim.h"
 #include "cso_cache/cso_hash.h"
 #include "tgsi_sanity.h"
 #include "tgsi_info.h"
@@ -463,8 +463,7 @@ iter_property(
 
    if (iter->processor.Processor == TGSI_PROCESSOR_GEOMETRY &&
        prop->Property.PropertyName == TGSI_PROPERTY_GS_INPUT_PRIM) {
-      ctx->implied_array_size =
-         pipe_vertices_per_primitive(prop->u[0].Data);
+      ctx->implied_array_size = u_vertices_per_prim(prop->u[0].Data);
    }
    return TRUE;
 }
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index ca247a1..825d17a 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -27,6 +27,7 @@
 
 #include "util/u_debug.h"
 #include "util/u_memory.h"
+#include "util/u_prim.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_inlines.h"
 #include "tgsi_text.h"
@@ -1187,8 +1188,7 @@ static boolean parse_property( struct translate_ctx *ctx )
       }
       if (property_name == TGSI_PROPERTY_GS_INPUT_PRIM &&
           ctx->processor == TGSI_PROCESSOR_GEOMETRY) {
-         ctx->implied_array_size =
-            pipe_vertices_per_primitive(values[0]);
+         ctx->implied_array_size = u_vertices_per_prim(values[0]);
       }
       break;
    default:
diff --git a/src/gallium/auxiliary/util/u_prim.h b/src/gallium/auxiliary/util/u_prim.h
index 7434329..10a874f 100644
--- a/src/gallium/auxiliary/util/u_prim.h
+++ b/src/gallium/auxiliary/util/u_prim.h
@@ -135,6 +135,39 @@ static INLINE unsigned u_reduced_prim( unsigned pipe_prim )
    }
 }
 
+static INLINE unsigned
+u_vertices_per_prim(int primitive)
+{
+   switch(primitive) {
+   case PIPE_PRIM_POINTS:
+      return 1;
+   case PIPE_PRIM_LINES:
+   case PIPE_PRIM_LINE_LOOP:
+   case PIPE_PRIM_LINE_STRIP:
+      return 2;
+   case PIPE_PRIM_TRIANGLES:
+   case PIPE_PRIM_TRIANGLE_STRIP:
+   case PIPE_PRIM_TRIANGLE_FAN:
+      return 3;
+   case PIPE_PRIM_LINES_ADJACENCY:
+   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
+      return 4;
+   case PIPE_PRIM_TRIANGLES_ADJACENCY:
+   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
+      return 6;
+
+   /* following primitives should never be used
+    * with geometry shaders abd their size is
+    * undefined */
+   case PIPE_PRIM_POLYGON:
+   case PIPE_PRIM_QUADS:
+   case PIPE_PRIM_QUAD_STRIP:
+   default:
+      debug_printf("Unrecognized geometry shader primitive");
+      return 3;
+   }
+}
+
 const char *u_prim_name( unsigned pipe_prim );
 
 #endif
diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h
index 95ec55d..5fbd62a 100644
--- a/src/gallium/include/pipe/p_inlines.h
+++ b/src/gallium/include/pipe/p_inlines.h
@@ -192,38 +192,6 @@ pipe_transfer_buffer_flags( struct pipe_transfer *transf )
    }
 }
 
-static INLINE unsigned
-pipe_vertices_per_primitive(int primitive)
-{
-   switch(primitive) {
-   case PIPE_PRIM_POINTS:
-      return 1;
-   case PIPE_PRIM_LINES:
-   case PIPE_PRIM_LINE_LOOP:
-   case PIPE_PRIM_LINE_STRIP:
-      return 2;
-   case PIPE_PRIM_TRIANGLES:
-   case PIPE_PRIM_TRIANGLE_STRIP:
-   case PIPE_PRIM_TRIANGLE_FAN:
-      return 3;
-   case PIPE_PRIM_LINES_ADJACENCY:
-   case PIPE_PRIM_LINE_STRIP_ADJACENCY:
-      return 4;
-   case PIPE_PRIM_TRIANGLES_ADJACENCY:
-   case PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY:
-      return 6;
-
-   /* following primitives should never be used
-    * with geometry shaders */
-   case PIPE_PRIM_POLYGON:
-   case PIPE_PRIM_QUADS:
-   case PIPE_PRIM_QUAD_STRIP:
-   default:
-      debug_printf("Unrecognized geometry shader primitive");
-      return 3;
-   }
-}
-
 #ifdef __cplusplus
 }
 #endif
--
1.6.5.7

---

[0014-gallium-geometry-shader-can-be-always-enabled-and-we.patch]

---

From 099fd63d2819716e2c162650c6dfc7196eade751 Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Thu, 24 Dec 2009 09:30:00 -0500
Subject: [PATCH 14/15] gallium: geometry shader can be always enabled and we don't need a cap for that

using the draw module allows us to enable geometry shading even on hardware
that doesn't support it.
---
 src/gallium/drivers/softpipe/sp_screen.c |    2 --
 src/gallium/include/pipe/p_defines.h     |    1 -
 2 files changed, 0 insertions(+), 3 deletions(-)

diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c
index a32312d..bd3532d 100644
--- a/src/gallium/drivers/softpipe/sp_screen.c
+++ b/src/gallium/drivers/softpipe/sp_screen.c
@@ -89,8 +89,6 @@ softpipe_get_param(struct pipe_screen *screen, int param)
       return 13; /* max 4Kx4K */
    case PIPE_CAP_TGSI_CONT_SUPPORTED:
       return 1;
-   case PIPE_CAP_GEOMETRY_SHADER4:
-      return 1;
    case PIPE_CAP_BLEND_EQUATION_SEPARATE:
       return 1;
    default:
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 90fc333..2cda408 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -398,7 +398,6 @@ enum pipe_transfer_usage {
 #define PIPE_CAP_MAX_PREDICATE_REGISTERS 30
 #define PIPE_CAP_MAX_COMBINED_SAMPLERS   31  /*< Maximum texture image units accessible from vertex
                                                  and fragment shaders combined */
-#define PIPE_CAP_GEOMETRY_SHADER4        32
 
 
 /**
--
1.6.5.7

---

[0015-draw-disable-some-debugging-output-coming-from-the-g.patch]

---

From 269df1b1f6fbe15f033cced6e7dc291b2c09683c Mon Sep 17 00:00:00 2001
From: Zack Rusin <zackr@...>
Date: Thu, 24 Dec 2009 09:34:07 -0500
Subject: [PATCH 15/15] draw: disable some debugging output coming from the geometry shaders

---
 src/gallium/auxiliary/draw/draw_gs.c |   10 +++++-----
 1 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
index 0a742cb..3edfb64 100644
--- a/src/gallium/auxiliary/draw/draw_gs.c
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -185,8 +185,8 @@ static void draw_fetch_geometry_input(struct draw_geometry_shader *shader,
    int idx = 0;
 
    for (slot = 0, vs_slot = 0; slot < shader->info.num_inputs; slot++) {
-      debug_printf("Slot = %d (semantic = %d)\n", slot,
-                   shader->info.input_semantic_name[slot]);
+      /*debug_printf("Slot = %d (semantic = %d)\n", slot,
+        shader->info.input_semantic_name[slot]);*/
       if (shader->info.input_semantic_name[slot] ==
           TGSI_SEMANTIC_VERTICES) {
          for (j = 0; j < num_primitives; ++j) {
@@ -200,8 +200,8 @@ static void draw_fetch_geometry_input(struct draw_geometry_shader *shader,
          for (j = 0; j < num_primitives; ++j) {
             int vidx = idx;
             const float (*prim_ptr)[4];
-            debug_printf("    %d) Prim (num_verts = %d)\n", start_primitive + j,
-                         num_vertices);
+            /*debug_printf("    %d) Prim (num_verts = %d)\n", start_primitive + j,
+              num_vertices);*/
             prim_ptr = (const float (*)[4])(
                (const char *)input_ptr +
                (j * num_vertices * input_vertex_stride));
@@ -211,7 +211,7 @@ static void draw_fetch_geometry_input(struct draw_geometry_shader *shader,
                input = (const float (*)[4])(
                   (const char *)prim_ptr + (k * input_vertex_stride));
                vidx = k * TGSI_EXEC_MAX_INPUT_ATTRIBS + slot;
-               debug_printf("\t%d)(%d) Input vert:\n", vidx, k);
+               /*debug_printf("\t%d)(%d) Input vert:\n", vidx, k);*/
 #if 1
                assert(!util_is_inf_or_nan(input[vs_slot][0]));
                assert(!util_is_inf_or_nan(input[vs_slot][1]));
--
1.6.5.7

---

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Keith Whitwell-3]

Thanks Zack.  I'm fine with doing it on top of the others...

Keith
________________________________________
From: Zack Rusin [zackr@...]
Sent: Thursday, December 24, 2009 6:39 AM
To: Keith Whitwell
Cc: mesa3d-dev; Michal Krol
Subject: Re: geometry shading patches

On Thursday 24 December 2009 09:09:44 Keith Whitwell wrote:
> Zack,
>
> In terms of advertising support for this - I wonder if this isn't something
>  we should just turn on for all drivers, given that in the worst case it is
>  just a vertex path fallback, and a lot of drivers will be hitting those
>  for various reasons anyway.

Yes, I completely agree with both of your emails. The attached patches
implement that (plus the last one comments out some unnecessary debugging
output). They do it on top of the previous patches but if you'd like to have a
history clear of them I can try to redo the entire series.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Zack Rusin-2]

On Thursday 24 December 2009 10:03:25 Keith Whitwell wrote:
> Thanks Zack.  I'm fine with doing it on top of the others...

ok, great, thanks Keith. In that case I'll wait for any objections until
tomorrow and if nothing will show up commit in the morning.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[michal-9]

Zack Rusin wrote on 2009-12-24 14:24:
> yo,
>
> after our discussions i hacked a new version of geometry shading support in
> gallium. the new geometry shading syntax looks as follows:
>  
Zack,

That looks nice. Once you commit I will take a closer look at patch #10
and see what's the issue there without bothering you.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Corbin Simpson]

Isn't this incredibly at odds with our previous discussion, in which
we generally agreed to not advertise support for unaccelerated things?
Geometry shaders are simply not going to have any kind of speed on
i915, nv30, nv40, or r300.

I can't speak for Jakob or the Nouveau guys, but at least r300g is
trying to keep Draw usage to a minimum. I know i915 has to use Draw,
so that won't be a problem, but nv30 and nv40 are Draw-free IIRC.

~ C.

On Thu, Dec 24, 2009 at 6:09 AM, Keith Whitwell <keithw@...> wrote:

> Zack,
>
> In terms of advertising support for this - I wonder if this isn't something we should just turn on for all drivers, given that in the worst case it is just a vertex path fallback, and a lot of drivers will be hitting those for various reasons anyway.
>
> Keith
> ________________________________________
> From: Zack Rusin [zackr@...]
> Sent: Thursday, December 24, 2009 5:24 AM
> To: mesa3d-dev
> Cc: Keith Whitwell; Michal Krol
> Subject: geometry shading patches
>
> yo,
>
> after our discussions i hacked a new version of geometry shading support in
> gallium. the new geometry shading syntax looks as follows:
>        GEOM
>        PROPERTY GS_INPUT_PRIMITIVE TRIANGLES
>        PROPERTY GS_OUTPUT_PRIMITIVE TRIANGLE_STRIP
>        DCL IN[][0], POSITION, CONSTANT
>        DCL IN[][1], COLOR, CONSTANT
>        DCL OUT[0], POSITION, CONSTANT
>        DCL OUT[1], COLOR, CONSTANT
>        0:MOV OUT[0], IN[0][0]
>        1:MOV OUT[1], IN[0][1]
>        2:EMIT
>        3:MOV OUT[0], IN[1][0]
>        4:MOV OUT[1], IN[1][1]
>        5:EMIT
>        6:MOV OUT[0], IN[2][0]
>        7:MOV OUT[1], IN[2][1]
>        8:EMIT
>        9:ENDPRIM
>        10:END
> the attached patch series adds support for all of it throughout gallium
> (including support for 2dimensional arrays in tgsi_sanity checking, tgsi_dump
> and tgsi_text).
> Right now don't support switching of primitives in the draw module (e.g. input
> primitive being points and output primitive being a triangle_strip) but i'd
> like to do that after committing the attached changes (also the semantic for
> vertices which is a leftover from an old patch should be removed then if it's
> ok with you).
>
> Keith, most relevant patches for you are 0001 and 0005.
> Michal the most relevant to you are 0007, 0010, 0011 and 0012.
>
> z
>
> ------------------------------------------------------------------------------
> This SF.Net email is sponsored by the Verizon Developer Community
> Take advantage of Verizon's best-in-class app development support
> A streamlined, 14 day to market process makes app distribution fast and easy
> Join now and get one step closer to millions of Verizon customers
> http://p.sf.net/sfu/verizon-dev2dev
> _______________________________________________
> Mesa3d-dev mailing list
> Mesa3d-dev@...
> https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
>

--
Only fools are easily impressed by what is only
barely beyond their reach. ~ Unknown

Corbin Simpson
<MostAwesomeDude@...>

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Zack Rusin-2]

On Friday 25 December 2009 07:03:02 Corbin Simpson wrote:
> Isn't this incredibly at odds with our previous discussion, in which
> we generally agreed to not advertise support for unaccelerated things?

No, it's really not. We don't have caps for core features, e.g we don't have
caps for vertex shaders and this goes hand in hand with that. Geometry shaders
are optional in the pipeline meaning that unlike fragment shaders they can be
absent in which case the pipeline behaves just like it would if the api didn't
have geometry shaders exposed at all i.e. vertex shader outputs go directly do
the fragment shader. So for games/apps that don't use geometry shaders this
won't matter at all. And games/app that are so new that they actually check
for geometry shaders will already be slow on i915 and r300 not because of
geometry shaders, but because they're running on it on i915 or r300 =)

Not to mention that this is not a fringe feature that will be present only in
super high-end and futuristic hardware.
 
All in all it's a bit like fixed-point hardware - programmable hardware is not
a cap because it's what Gallium models. We can't just keep the Gallium
interface at i915 level and mark everything above that as a cap, it'd be silly
given that we're generations past that now.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Corbin Simpson]

Half of the in-tree HW-accelerated drivers lack geo shaders, not to mention out-of-tree drivers. I politely request that drivers not be required to have the geo shader hooks filled out; mesa st can detect this and enable extensions appropriately.

Posting from a mobile, pardon my terseness. ~ C.

On Dec 25, 2009 8:16 AM, "Zack Rusin" <zackr@...> wrote:

On Friday 25 December 2009 07:03:02 Corbin Simpson wrote: > Isn't this incredibly at odds with our p...

No, it's really not. We don't have caps for core features, e.g we don't have
caps for vertex shaders and this goes hand in hand with that. Geometry shaders
are optional in the pipeline meaning that unlike fragment shaders they can be
absent in which case the pipeline behaves just like it would if the api didn't
have geometry shaders exposed at all i.e. vertex shader outputs go directly do
the fragment shader. So for games/apps that don't use geometry shaders this
won't matter at all. And games/app that are so new that they actually check
for geometry shaders will already be slow on i915 and r300 not because of
geometry shaders, but because they're running on it on i915 or r300 =)

Not to mention that this is not a fringe feature that will be present only in
super high-end and futuristic hardware.

All in all it's a bit like fixed-point hardware - programmable hardware is not
a cap because it's what Gallium models. We can't just keep the Gallium
interface at i915 level and mark everything above that as a cap, it'd be silly
given that we're generations past that now.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Marek Olšák]

Zack,

to be honest, Direct3D 11 can report geometry shaders are not supported through so-called feature levels. There are six levels to my knowledge (9_1, 9_2, 9_3, 10_0, 10_1, 11_0). i915 is 9_1, R300 is 9_2, R500 is 9_3, and so on. Direct3D 11 is indeed accelerated on those pieces of hardware and, though the feature set is a little limited, the hardware support is covered well. Is Direct3D 11 generations past because of that? No, it isn't.

Let's say I have R500 and I want to use geometry shaders in Direct3D 11. What are my options? I can't use my R500 and I must manually switch to the device called WARP (Windows Advanced Rasterization Platform), which reports the 10_1 feature level. This kind of device is very similar to llvmpipe in Gallium.

In the past you said we should do it the same way as Direct3D, so why should Gallium be different now? Moreover, if applications decide to use geometry shaders to emulate point sprites or wide lines, we'll be screwed. If they decide to do texture fetches in geometry shaders, we'll be screwed even more because we'll have to move textures out of VRAM and that will be a total performance killer. So I agree with Corbin that the CAP for geometry shaders should be added and we should let drivers decide what's best for them.

Marek

On Fri, Dec 25, 2009 at 5:11 PM, Zack Rusin <zackr@...> wrote:

On Friday 25 December 2009 07:03:02 Corbin Simpson wrote:
> Isn't this incredibly at odds with our previous discussion, in which
> we generally agreed to not advertise support for unaccelerated things?

No, it's really not. We don't have caps for core features, e.g we don't have
caps for vertex shaders and this goes hand in hand with that. Geometry shaders
are optional in the pipeline meaning that unlike fragment shaders they can be
absent in which case the pipeline behaves just like it would if the api didn't
have geometry shaders exposed at all i.e. vertex shader outputs go directly do
the fragment shader. So for games/apps that don't use geometry shaders this
won't matter at all. And games/app that are so new that they actually check
for geometry shaders will already be slow on i915 and r300 not because of
geometry shaders, but because they're running on it on i915 or r300 =)

Not to mention that this is not a fringe feature that will be present only in
super high-end and futuristic hardware.

All in all it's a bit like fixed-point hardware - programmable hardware is not
a cap because it's what Gallium models. We can't just keep the Gallium
interface at i915 level and mark everything above that as a cap, it'd be silly
given that we're generations past that now.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Keith Whitwell-3]

I don't think this is a big deal either way.  My feeling was that since most or all drivers will have some use of the draw module, this wouldn't be any worse.  But if there really are pre-geometry shader drivers that can avoid the draw module in all other situations, that's probably something we want to preserve.  Corbin,  do these drivers actually implement all of the GL stuff that the draw module provides fallbacks for?  Or is there missing functionality that will eventually need help to implement?

Basically my desire is to avoid an explosion of capability bits and advertise just a small number of broad classes of hardware.  Probably the way to answer the question of whether this should be a cap or not is to do the work of figuring out what these mythical harware classes are and whether that can be made to work.

Keith


________________________________________
From: Corbin Simpson [mostawesomedude@...]
Sent: Friday, December 25, 2009 4:03 AM
To: Keith Whitwell
Cc: Zack Rusin; mesa3d-dev
Subject: Re: [Mesa3d-dev] geometry shading patches

I can't speak for Jakob or the Nouveau guys, but at least r300g is
trying to keep Draw usage to a minimum. I know i915 has to use Draw,
so that won't be a problem, but nv30 and nv40 are Draw-free IIRC.

~ C.

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

[Zack Rusin-2]

On Saturday 26 December 2009 02:19:40 Marek Olšák wrote:

> Zack,
>
> to be honest, Direct3D 11 can report geometry shaders are not supported
>  through so-called feature levels. There are six levels to my knowledge
>  (9_1, 9_2, 9_3, 10_0, 10_1, 11_0). i915 is 9_1, R300 is 9_2, R500 is 9_3,
>  and so on. Direct3D 11 is indeed accelerated on those pieces of hardware
>  and, though the feature set is a little limited, the hardware support is
>  covered well. Is Direct3D 11 generations past because of that? No, it
>  isn't.
>
> Let's say I have R500 and I want to use geometry shaders in Direct3D 11.
>  What are my options? I can't use my R500 and I must manually switch to the
>  device called WARP (Windows Advanced Rasterization Platform), which
>  reports the 10_1 feature level. This kind of device is very similar to
>  llvmpipe in Gallium.
>
> In the past you said we should do it the same way as Direct3D, so why
>  should Gallium be different now?

I think you're using "it" a bit broadly here because we never had a discussion
about caps. First of all Gallium3D is already different when it comes to
capabilities reporting. We have a buttload of caps, realistically speaking
most of them are likely tight together. What we do right now is what d3d9 used
to do which is what everyone agreed is awful. Not to mention that we don't
have an option of actually selecting llvmpipe vs whatever hardware driver by
hand.
The argument that would certainly make sense is one for moving Gallium3D caps
model towards a shader-model reporting. e.g. shader-model 2.x, 3.x, 4.x, 5.x
versus every single feature that they bring forward, e.g. 4.x implies geometry
shader, 5.x implies tessellation/compute.
I absolutely abhor the idea of reporting as a cap everything above of what
i915 or r300 can do, for the lack of better wording it's just ridicules. I do
think though that we should look at our caps bits and come up with something
better.

>  Moreover, if applications decide to use geometry shaders to emulate point
>  sprites or wide lines, we'll be screwed.

If the hardware doesn't implement those features they'll be in the draw module
anyway. So it's really draw module vs draw module.

>  If they decide to do texture fetches in geometry shaders, we'll be screwed
>  even more because we'll have to move textures out of VRAM and that will be
>  a total performance killer. So I agree with Corbin that the CAP for
>  geometry shaders should be added and we should let drivers decide what's
>  best for them.

How is that different from the same problem applied to a vertex shader on i915
and the ways that works right now?
I agree that we need to solve that problems, but I just refuse that the best
we can is "everything above i915 is a feature cap". We need to come up with a
scheme that actually works or assume it's ok for draw module to handle some of
those features.
For us it likely should be some combination of API and shader-model support
(shader-models don't tell us anything about gl specific features like shadow
samplers or aa lines/points), if we can figure out we can reasonable handle
that we'll be fine.

z

------------------------------------------------------------------------------
This SF.Net email is sponsored by the Verizon Developer Community
Take advantage of Verizon's best-in-class app development support
A streamlined, 14 day to market process makes app distribution fast and easy
Join now and get one step closer to millions of Verizon customers
http://p.sf.net/sfu/verizon-dev2dev 
_______________________________________________
Mesa3d-dev mailing list
Mesa3d-dev@...
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev