Merge branch 'upstream-gallium-0.1' into nouveau-gallium-0.1

Conflicts:

	src/gallium/drivers/Makefile

17 files changed, 2645 insertions, 126 deletions

diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile
index fe9b150f30..c9980f0b83 100644
--- a/src/gallium/auxiliary/draw/Makefile
+++ b/src/gallium/auxiliary/draw/Makefile
@@ -1,2 +1,37 @@
-default:
-	cd ../../../mesa ; make
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = draw
+
+C_SOURCES = \
+	draw_aaline.c \
+	draw_aapoint.c \
+	draw_clip.c \
+	draw_vs_exec.c \
+	draw_vs_sse.c \
+	draw_vs_llvm.c \
+	draw_context.c\
+	draw_cull.c \
+	draw_debug.c \
+	draw_flatshade.c \
+	draw_offset.c \
+	draw_prim.c \
+	draw_pstipple.c \
+	draw_stipple.c \
+	draw_twoside.c \
+	draw_unfilled.c \
+	draw_validate.c \
+	draw_vbuf.c \
+	draw_vertex.c \
+	draw_vertex_cache.c \
+	draw_vertex_fetch.c \
+	draw_vertex_shader.c \
+	draw_vf.c \
+	draw_vf_generic.c \
+	draw_vf_sse.c \
+	draw_wide_prims.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript
new file mode 100644
index 0000000000..8e3a8caa74
--- /dev/null
+++ b/src/gallium/auxiliary/draw/SConscript
@@ -0,0 +1,31 @@
+Import('*')
+
+draw = env.ConvenienceLibrary(
+	target = 'draw',
+	source = [
+		'draw_clip.c',
+		'draw_vs_exec.c',
+		'draw_vs_sse.c',
+		'draw_vs_llvm.c',
+		'draw_context.c',
+		'draw_cull.c',
+		'draw_debug.c',
+		'draw_flatshade.c',
+		'draw_offset.c',
+		'draw_prim.c',
+		'draw_stipple.c',
+		'draw_twoside.c',
+		'draw_unfilled.c',
+		'draw_validate.c',
+		'draw_vbuf.c',
+		'draw_vertex.c',
+		'draw_vertex_cache.c',
+		'draw_vertex_fetch.c',
+		'draw_vertex_shader.c',
+		'draw_vf.c',
+		'draw_vf_generic.c',
+		'draw_vf_sse.c',
+		'draw_wide_prims.c',
+	])
+
+auxiliaries.insert(0, draw)
diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c
new file mode 100644
index 0000000000..73a02a32e4
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_aaline.c
@@ -0,0 +1,832 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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.
+ * 
+ **************************************************************************/
+
+/**
+ * AA line stage:  AA lines are converted to texture mapped triangles.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/util/tgsi_transform.h"
+#include "tgsi/util/tgsi_dump.h"
+
+#include "draw_context.h"
+#include "draw_private.h"
+
+
+/**
+ * Max texture level for the alpha texture used for antialiasing
+ */
+#define MAX_TEXTURE_LEVEL  5   /* 32 x 32 */
+
+
+/**
+ * Subclass of pipe_shader_state to carry extra fragment shader info.
+ */
+struct aaline_fragment_shader
+{
+   struct pipe_shader_state state;
+   void *driver_fs;
+   void *aaline_fs;
+   void *aapoint_fs; /* not yet */
+   void *sprite_fs; /* not yet */
+};
+
+
+/**
+ * Subclass of draw_stage
+ */
+struct aaline_stage
+{
+   struct draw_stage stage;
+
+   float half_line_width;
+
+   /** For AA lines, this is the vertex attrib slot for the new texcoords */
+   uint tex_slot;
+
+   void *sampler_cso;
+   struct pipe_texture *texture;
+   uint sampler_unit;
+
+
+   /*
+    * Currently bound state
+    */
+   struct aaline_fragment_shader *fs;
+   struct {
+      void *sampler[PIPE_MAX_SAMPLERS];
+      struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+   } state;
+
+   /*
+    * Driver interface/override functions
+    */
+   void * (*driver_create_fs_state)(struct pipe_context *,
+                                    const struct pipe_shader_state *);
+   void (*driver_bind_fs_state)(struct pipe_context *, void *);
+   void (*driver_delete_fs_state)(struct pipe_context *, void *);
+
+   void (*driver_bind_sampler_state)(struct pipe_context *, unsigned, void *);
+
+   void (*driver_set_sampler_texture)(struct pipe_context *,
+                                      unsigned sampler,
+                                      struct pipe_texture *);
+
+   struct pipe_context *pipe;
+};
+
+
+
+/**
+ * Subclass of tgsi_transform_context, used for transforming the
+ * user's fragment shader to add the special AA instructions.
+ */
+struct aa_transform_context {
+   struct tgsi_transform_context base;
+   uint tempsUsed;  /**< bitmask */
+   int colorOutput; /**< which output is the primary color */
+   int maxSampler;  /**< max sampler index found */
+   int maxInput, maxGeneric;  /**< max input index found */
+   int colorTemp, texTemp;  /**< temp registers */
+   boolean firstInstruction;
+};
+
+
+/**
+ * TGSI declaration transform callback.
+ * Look for a free sampler, a free input attrib, and two free temp regs.
+ */
+static void
+aa_transform_decl(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_declaration *decl)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+
+   if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+       decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR &&
+       decl->Semantic.SemanticIndex == 0) {
+      aactx->colorOutput = decl->u.DeclarationRange.First;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_SAMPLER) {
+      if ((int) decl->u.DeclarationRange.Last > aactx->maxSampler)
+         aactx->maxSampler = decl->u.DeclarationRange.Last + 1;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      if ((int) decl->u.DeclarationRange.Last > aactx->maxInput)
+         aactx->maxInput = decl->u.DeclarationRange.Last;
+      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC &&
+           (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) {
+         aactx->maxGeneric = decl->Semantic.SemanticIndex;
+      }
+   }
+   else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
+      uint i;
+      for (i = decl->u.DeclarationRange.First;
+           i <= decl->u.DeclarationRange.Last; i++) {
+         aactx->tempsUsed |= (1 << i);
+      }
+   }
+
+   ctx->emit_declaration(ctx, decl);
+}
+
+
+/**
+ * TGSI instruction transform callback.
+ * Replace writes to result.color w/ a temp reg.
+ * Upon END instruction, insert texture sampling code for antialiasing.
+ */
+static void
+aa_transform_inst(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_instruction *inst)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+
+   if (aactx->firstInstruction) {
+      /* emit our new declarations before the first instruction */
+
+      struct tgsi_full_declaration decl;
+      uint i;
+
+      /* find two free temp regs */
+      for (i = 0; i < 32; i++) {
+         if ((aactx->tempsUsed & (1 << i)) == 0) {
+            /* found a free temp */
+            if (aactx->colorTemp < 0)
+               aactx->colorTemp  = i;
+            else if (aactx->texTemp < 0)
+               aactx->texTemp  = i;
+            else
+               break;
+         }
+      }
+      assert(aactx->colorTemp >= 0);
+      assert(aactx->texTemp >= 0);
+
+      /* declare new generic input/texcoord */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_INPUT;
+      decl.Declaration.Semantic = 1;
+      decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
+      decl.Semantic.SemanticIndex = aactx->maxGeneric + 1;
+      decl.Declaration.Interpolate = 1;
+      /* XXX this could be linear... */
+      decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = aactx->maxInput + 1;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new sampler */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_SAMPLER;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = aactx->maxSampler + 1;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new temp regs */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = aactx->texTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = aactx->colorTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      aactx->firstInstruction = FALSE;
+   }
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_END &&
+       aactx->colorOutput != -1) {
+      struct tgsi_full_instruction newInst;
+
+      /* TEX */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_TEX;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->texTemp;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->maxSampler + 1;
+
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MOV rgb */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MUL alpha */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->texTemp;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* END */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_END;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 0;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+   else {
+      /* Not an END instruction.
+       * Look for writes to result.color and replace with colorTemp reg.
+       */
+      uint i;
+
+      for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+         struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+         if (dst->DstRegister.File == TGSI_FILE_OUTPUT &&
+             dst->DstRegister.Index == aactx->colorOutput) {
+            dst->DstRegister.File = TGSI_FILE_TEMPORARY;
+            dst->DstRegister.Index = aactx->colorTemp;
+         }
+      }
+
+      ctx->emit_instruction(ctx, inst);
+   }
+}
+
+
+/**
+ * Generate the frag shader we'll use for drawing AA lines.
+ * This will be the user's shader plus some texture/modulate instructions.
+ */
+static void
+generate_aaline_fs(struct aaline_stage *aaline)
+{
+   const struct pipe_shader_state *orig_fs = &aaline->fs->state;
+   struct draw_context *draw = aaline->stage.draw;
+   struct pipe_shader_state aaline_fs;
+   struct aa_transform_context transform;
+
+#define MAX 1000
+
+   aaline_fs = *orig_fs; /* copy to init */
+   aaline_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX);
+
+   memset(&transform, 0, sizeof(transform));
+   transform.colorOutput = -1;
+   transform.maxSampler = -1;
+   transform.maxInput = -1;
+   transform.maxGeneric = -1;
+   transform.colorTemp = -1;
+   transform.texTemp = -1;
+   transform.firstInstruction = TRUE;
+   transform.base.transform_instruction = aa_transform_inst;
+   transform.base.transform_declaration = aa_transform_decl;
+
+   tgsi_transform_shader(orig_fs->tokens,
+                         (struct tgsi_token *) aaline_fs.tokens,
+                         MAX, &transform.base);
+
+#if 0 /* DEBUG */
+   tgsi_dump(orig_fs->tokens, 0);
+   tgsi_dump(aaline_fs.tokens, 0);
+#endif
+
+   aaline_fs.input_semantic_name[aaline_fs.num_inputs] = TGSI_SEMANTIC_GENERIC;
+   aaline_fs.input_semantic_index[aaline_fs.num_inputs] = transform.maxGeneric + 1;
+   aaline_fs.num_inputs++;
+
+   aaline->fs->aaline_fs
+      = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs);
+
+   /* advertise the extra post-transform vertex attributes which will have
+    * the texcoords.
+    */
+   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+   draw->extra_vp_outputs.semantic_index = transform.maxGeneric + 1;
+}
+
+
+/**
+ * Create the texture map we'll use for antialiasing the lines.
+ */
+static void
+aaline_create_texture(struct aaline_stage *aaline)
+{
+   struct pipe_context *pipe = aaline->pipe;
+   struct pipe_texture texTemp;
+   uint level;
+
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */
+   texTemp.last_level = MAX_TEXTURE_LEVEL;
+   texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL;
+   texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL;
+   texTemp.depth[0] = 1;
+   texTemp.cpp = 1;
+
+   aaline->texture = pipe->texture_create(pipe, &texTemp);
+
+   /* Fill in mipmap images.
+    * Basically each level is solid opaque, except for the outermost
+    * texels which are zero.  Special case the 1x1 and 2x2 levels.
+    */
+   for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) {
+      struct pipe_surface *surface;
+      const uint size = aaline->texture->width[level];
+      ubyte *data;
+      uint i, j;
+
+      assert(aaline->texture->width[level] == aaline->texture->height[level]);
+
+      surface = pipe->get_tex_surface(pipe, aaline->texture, 0, level, 0);
+      data = pipe_surface_map(surface);
+
+      for (i = 0; i < size; i++) {
+         for (j = 0; j < size; j++) {
+            ubyte d;
+            if (size == 1) {
+               d = 255;
+            }
+            else if (size == 2) {
+               d = 200; /* tuneable */
+            }
+            else if (i == 0 || j == 0 || i == size - 1 || j == size - 1) {
+               d = 0;
+            }
+            else {
+               d = 255;
+            }
+            data[i * surface->pitch + j] = d;
+         }
+      }
+
+      /* unmap */
+      pipe_surface_unmap(surface);
+      pipe_surface_reference(&surface, NULL);
+   }
+}
+
+
+/**
+ * Create the sampler CSO that'll be used for antialiasing.
+ * By using a mipmapped texture, we don't have to generate a different
+ * texture image for each line size.
+ */
+static void
+aaline_create_sampler(struct aaline_stage *aaline)
+{
+   struct pipe_sampler_state sampler;
+   struct pipe_context *pipe = aaline->pipe;
+
+   memset(&sampler, 0, sizeof(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_LINEAR;
+   sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR;
+   sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
+   sampler.normalized_coords = 1;
+   sampler.min_lod = 0.0f;
+   sampler.max_lod = MAX_TEXTURE_LEVEL;
+
+   aaline->sampler_cso = pipe->create_sampler_state(pipe, &sampler);
+}
+
+
+/**
+ * When we're about to draw our first AA line in a batch, this function is
+ * called to tell the driver to bind our modified fragment shader.
+ */
+static void
+bind_aaline_fragment_shader(struct aaline_stage *aaline)
+{
+   if (!aaline->fs->aaline_fs) {
+      generate_aaline_fs(aaline);
+   }
+   aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs);
+}
+
+
+
+static INLINE struct aaline_stage *
+aaline_stage( struct draw_stage *stage )
+{
+   return (struct aaline_stage *) stage;
+}
+
+
+static void
+passthrough_point(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->point(stage->next, header);
+}
+
+
+static void
+passthrough_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->tri(stage->next, header);
+}
+
+
+/**
+ * Draw a wide line by drawing a quad, using geometry which will
+ * fullfill GL's antialiased line requirements.
+ */
+static void
+aaline_line(struct draw_stage *stage, struct prim_header *header)
+{
+   const struct aaline_stage *aaline = aaline_stage(stage);
+   const float half_width = aaline->half_line_width;
+   struct prim_header tri;
+   struct vertex_header *v[8];
+   uint texPos = aaline->tex_slot;
+   float *pos, *tex;
+   float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0];
+   float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1];
+   double a = atan2(dy, dx);
+   float c_a = (float) cos(a), s_a = (float) sin(a);
+   uint i;
+
+   /* XXX the ends of lines aren't quite perfect yet, but probably passable */
+   dx = 0.5F * half_width;
+   dy = half_width;
+
+   /* allocate/dup new verts */
+   for (i = 0; i < 8; i++) {
+      v[i] = dup_vert(stage, header->v[i/4], i);
+   }
+
+   /*
+    * Quad strip for line from v0 to v1 (*=endpoints):
+    *
+    *  1   3                     5   7
+    *  +---+---------------------+---+
+    *  |                             |
+    *  | *v0                     v1* |
+    *  |                             |
+    *  +---+---------------------+---+
+    *  0   2                     4   6
+    */
+
+   /* new verts */
+   pos = v[0]->data[0];
+   pos[0] += (-dx * c_a -  dy * s_a);
+   pos[1] += (-dx * s_a +  dy * c_a);
+
+   pos = v[1]->data[0];
+   pos[0] += (-dx * c_a - -dy * s_a);
+   pos[1] += (-dx * s_a + -dy * c_a);
+
+   pos = v[2]->data[0];
+   pos[0] += ( dx * c_a -  dy * s_a);
+   pos[1] += ( dx * s_a +  dy * c_a);
+
+   pos = v[3]->data[0];
+   pos[0] += ( dx * c_a - -dy * s_a);
+   pos[1] += ( dx * s_a + -dy * c_a);
+
+   pos = v[4]->data[0];
+   pos[0] += (-dx * c_a -  dy * s_a);
+   pos[1] += (-dx * s_a +  dy * c_a);
+
+   pos = v[5]->data[0];
+   pos[0] += (-dx * c_a - -dy * s_a);
+   pos[1] += (-dx * s_a + -dy * c_a);
+
+   pos = v[6]->data[0];
+   pos[0] += ( dx * c_a -  dy * s_a);
+   pos[1] += ( dx * s_a +  dy * c_a);
+
+   pos = v[7]->data[0];
+   pos[0] += ( dx * c_a - -dy * s_a);
+   pos[1] += ( dx * s_a + -dy * c_a);
+
+   /* new texcoords */
+   tex = v[0]->data[texPos];
+   ASSIGN_4V(tex, 0, 0, 0, 1);
+
+   tex = v[1]->data[texPos];
+   ASSIGN_4V(tex, 0, 1, 0, 1);
+
+   tex = v[2]->data[texPos];
+   ASSIGN_4V(tex, .5, 0, 0, 1);
+
+   tex = v[3]->data[texPos];
+   ASSIGN_4V(tex, .5, 1, 0, 1);
+
+   tex = v[4]->data[texPos];
+   ASSIGN_4V(tex, .5, 0, 0, 1);
+
+   tex = v[5]->data[texPos];
+   ASSIGN_4V(tex, .5, 1, 0, 1);
+
+   tex = v[6]->data[texPos];
+   ASSIGN_4V(tex, 1, 0, 0, 1);
+
+   tex = v[7]->data[texPos];
+   ASSIGN_4V(tex, 1, 1, 0, 1);
+
+   /* emit 6 tris for the quad strip */
+   tri.v[0] = v[2];  tri.v[1] = v[1];  tri.v[2] = v[0];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[3];  tri.v[1] = v[1];  tri.v[2] = v[2];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[4];  tri.v[1] = v[3];  tri.v[2] = v[2];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[5];  tri.v[1] = v[3];  tri.v[2] = v[4];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[6];  tri.v[1] = v[5];  tri.v[2] = v[4];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[7];  tri.v[1] = v[5];  tri.v[2] = v[6];
+   stage->next->tri( stage->next, &tri );
+}
+
+
+static void
+aaline_first_line(struct draw_stage *stage, struct prim_header *header)
+{
+   auto struct aaline_stage *aaline = aaline_stage(stage);
+   struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = aaline->pipe;
+
+   assert(draw->rasterizer->line_smooth);
+
+   if (draw->rasterizer->line_width <= 3.0)
+      aaline->half_line_width = 1.5f;
+   else
+      aaline->half_line_width = 0.5f * draw->rasterizer->line_width;
+
+   aaline->tex_slot = draw->num_vs_outputs;
+   assert(aaline->tex_slot > 0); /* output[0] is vertex pos */
+   draw->extra_vp_outputs.slot = aaline->tex_slot;
+
+   /*
+    * Bind our fragprog, sampler and texture
+    */
+   bind_aaline_fragment_shader(aaline);
+
+   aaline->driver_bind_sampler_state(pipe, aaline->sampler_unit, aaline->sampler_cso);
+   aaline->driver_set_sampler_texture(pipe, aaline->sampler_unit, aaline->texture);
+
+   /* now really draw first line */
+   stage->line = aaline_line;
+   stage->line(stage, header);
+}
+
+
+static void
+aaline_flush(struct draw_stage *stage, unsigned flags)
+{
+   struct draw_context *draw = stage->draw;
+   struct aaline_stage *aaline = aaline_stage(stage);
+   struct pipe_context *pipe = aaline->pipe;
+
+   stage->line = aaline_first_line;
+   stage->next->flush( stage->next, flags );
+
+   /* restore original frag shader */
+   aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs);
+
+   /* XXX restore original texture, sampler state */
+   aaline->driver_bind_sampler_state(pipe, aaline->sampler_unit,
+                                 aaline->state.sampler[aaline->sampler_unit]);
+   aaline->driver_set_sampler_texture(pipe, aaline->sampler_unit,
+                                 aaline->state.texture[aaline->sampler_unit]);
+
+   draw->extra_vp_outputs.slot = 0;
+}
+
+
+static void
+aaline_reset_stipple_counter(struct draw_stage *stage)
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void
+aaline_destroy(struct draw_stage *stage)
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+static struct aaline_stage *
+draw_aaline_stage(struct draw_context *draw)
+{
+   struct aaline_stage *aaline = CALLOC_STRUCT(aaline_stage);
+
+   draw_alloc_temp_verts( &aaline->stage, 8 );
+
+   aaline->stage.draw = draw;
+   aaline->stage.next = NULL;
+   aaline->stage.point = passthrough_point;
+   aaline->stage.line = aaline_first_line;
+   aaline->stage.tri = passthrough_tri;
+   aaline->stage.flush = aaline_flush;
+   aaline->stage.reset_stipple_counter = aaline_reset_stipple_counter;
+   aaline->stage.destroy = aaline_destroy;
+
+   return aaline;
+}
+
+
+/*
+ * XXX temporary? solution to mapping a pipe_context to a aaline_stage.
+ */
+
+#define MAX_CONTEXTS 10
+
+static struct pipe_context *Pipe[MAX_CONTEXTS];
+static struct aaline_stage *Stage[MAX_CONTEXTS];
+static uint NumContexts;
+
+static void
+add_aa_pipe_context(struct pipe_context *pipe, struct aaline_stage *aa)
+{
+   assert(NumContexts < MAX_CONTEXTS);
+   Pipe[NumContexts] = pipe;
+   Stage[NumContexts] = aa;
+   NumContexts++;
+}
+
+static struct aaline_stage *
+aaline_stage_from_pipe(struct pipe_context *pipe)
+{
+   uint i;
+   for (i = 0; i < NumContexts; i++) {
+      if (Pipe[i] == pipe)
+         return Stage[i];
+   }
+   assert(0);
+   return NULL;
+}
+
+
+/**
+ * This function overrides the driver's create_fs_state() function and
+ * will typically be called by the state tracker.
+ */
+static void *
+aaline_create_fs_state(struct pipe_context *pipe,
+                       const struct pipe_shader_state *fs)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader);
+
+   if (aafs) {
+      aafs->state = *fs;
+
+      /* pass-through */
+      aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs);
+   }
+
+   return aafs;
+}
+
+
+static void
+aaline_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs;
+   /* save current */
+   aaline->fs = aafs;
+   /* pass-through */
+   aaline->driver_bind_fs_state(aaline->pipe, aafs->driver_fs);
+}
+
+
+static void
+aaline_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs;
+   /* pass-through */
+   aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs);
+   FREE(aafs);
+}
+
+
+static void
+aaline_bind_sampler_state(struct pipe_context *pipe,
+                          unsigned unit, void *sampler)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   /* save current */
+   aaline->state.sampler[unit] = sampler;
+   /* pass-through */
+   aaline->driver_bind_sampler_state(aaline->pipe, unit, sampler);
+}
+
+
+static void
+aaline_set_sampler_texture(struct pipe_context *pipe,
+                           unsigned sampler, struct pipe_texture *texture)
+{
+   struct aaline_stage *aaline = aaline_stage_from_pipe(pipe);
+   /* save current */
+   aaline->state.texture[sampler] = texture;
+   /* pass-through */
+   aaline->driver_set_sampler_texture(aaline->pipe, sampler, texture);
+}
+
+
+/**
+ * Called by drivers that want to install this AA line prim stage
+ * into the draw module's pipeline.  This will not be used if the
+ * hardware has native support for AA lines.
+ */
+void
+draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe)
+{
+   struct aaline_stage *aaline;
+
+   /*
+    * Create / install AA line drawing / prim stage
+    */
+   aaline = draw_aaline_stage( draw );
+   assert(aaline);
+   draw->pipeline.aaline = &aaline->stage;
+
+   aaline->pipe = pipe;
+
+   /* create special texture, sampler state */
+   aaline_create_texture(aaline);
+   aaline_create_sampler(aaline);
+
+   /* save original driver functions */
+   aaline->driver_create_fs_state = pipe->create_fs_state;
+   aaline->driver_bind_fs_state = pipe->bind_fs_state;
+   aaline->driver_delete_fs_state = pipe->delete_fs_state;
+
+   aaline->driver_bind_sampler_state = pipe->bind_sampler_state;
+   aaline->driver_set_sampler_texture = pipe->set_sampler_texture;
+
+   /* override the driver's functions */
+   pipe->create_fs_state = aaline_create_fs_state;
+   pipe->bind_fs_state = aaline_bind_fs_state;
+   pipe->delete_fs_state = aaline_delete_fs_state;
+
+   pipe->bind_sampler_state = aaline_bind_sampler_state;
+   pipe->set_sampler_texture = aaline_set_sampler_texture;
+
+   add_aa_pipe_context(pipe, aaline);
+}
diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c
new file mode 100644
index 0000000000..43119cc70b
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_aapoint.c
@@ -0,0 +1,875 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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.
+ * 
+ **************************************************************************/
+
+/**
+ * AA point stage:  AA points are converted to quads and rendered with a
+ * special fragment shader.  Another approach would be to use a texture
+ * map image of a point, but experiments indicate the quality isn't nearly
+ * as good as this approach.
+ *
+ * Note: this looks a lot like draw_aaline.c but there's actually little
+ * if any code that can be shared.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/util/tgsi_transform.h"
+#include "tgsi/util/tgsi_dump.h"
+
+#include "draw_context.h"
+#include "draw_private.h"
+
+
+/*
+ * Enabling NORMALIZE might give _slightly_ better results.
+ * Basically, it controls whether we compute distance as d=sqrt(x*x+y*y) or
+ * d=x*x+y*y.  Since we're working with a unit circle, the later seems
+ * close enough and saves some costly instructions.
+ */
+#define NORMALIZE 0
+
+
+/**
+ * Subclass of pipe_shader_state to carry extra fragment shader info.
+ */
+struct aapoint_fragment_shader
+{
+   struct pipe_shader_state state;
+   void *driver_fs;   /**< the regular shader */
+   void *aapoint_fs;  /**< the aa point-augmented shader */
+};
+
+
+/**
+ * Subclass of draw_stage
+ */
+struct aapoint_stage
+{
+   struct draw_stage stage;
+
+   int psize_slot;
+   float radius;
+
+   /** this is the vertex attrib slot for the new texcoords */
+   uint tex_slot;
+
+   /*
+    * Currently bound state
+    */
+   struct aapoint_fragment_shader *fs;
+
+   /*
+    * Driver interface/override functions
+    */
+   void * (*driver_create_fs_state)(struct pipe_context *,
+                                    const struct pipe_shader_state *);
+   void (*driver_bind_fs_state)(struct pipe_context *, void *);
+   void (*driver_delete_fs_state)(struct pipe_context *, void *);
+
+   struct pipe_context *pipe;
+};
+
+
+
+/**
+ * Subclass of tgsi_transform_context, used for transforming the
+ * user's fragment shader to add the special AA instructions.
+ */
+struct aa_transform_context {
+   struct tgsi_transform_context base;
+   uint tempsUsed;  /**< bitmask */
+   int colorOutput; /**< which output is the primary color */
+   int maxInput, maxGeneric;  /**< max input index found */
+   int tmp0, colorTemp;  /**< temp registers */
+   boolean firstInstruction;
+};
+
+
+/**
+ * TGSI declaration transform callback.
+ * Look for two free temp regs and available input reg for new texcoords.
+ */
+static void
+aa_transform_decl(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_declaration *decl)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+
+   if (decl->Declaration.File == TGSI_FILE_OUTPUT &&
+       decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR &&
+       decl->Semantic.SemanticIndex == 0) {
+      aactx->colorOutput = decl->u.DeclarationRange.First;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      if ((int) decl->u.DeclarationRange.Last > aactx->maxInput)
+         aactx->maxInput = decl->u.DeclarationRange.Last;
+      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC &&
+           (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) {
+         aactx->maxGeneric = decl->Semantic.SemanticIndex;
+      }
+   }
+   else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
+      uint i;
+      for (i = decl->u.DeclarationRange.First;
+           i <= decl->u.DeclarationRange.Last; i++) {
+         aactx->tempsUsed |= (1 << i);
+      }
+   }
+
+   ctx->emit_declaration(ctx, decl);
+}
+
+
+/**
+ * TGSI instruction transform callback.
+ * Replace writes to result.color w/ a temp reg.
+ * Upon END instruction, insert texture sampling code for antialiasing.
+ */
+static void
+aa_transform_inst(struct tgsi_transform_context *ctx,
+                  struct tgsi_full_instruction *inst)
+{
+   struct aa_transform_context *aactx = (struct aa_transform_context *) ctx;
+   struct tgsi_full_instruction newInst;
+
+   if (aactx->firstInstruction) {
+      /* emit our new declarations before the first instruction */
+
+      struct tgsi_full_declaration decl;
+      const int texInput = aactx->maxInput + 1;
+      int tmp0;
+      uint i;
+
+      /* find two free temp regs */
+      for (i = 0; i < 32; i++) {
+         if ((aactx->tempsUsed & (1 << i)) == 0) {
+            /* found a free temp */
+            if (aactx->tmp0 < 0)
+               aactx->tmp0 = i;
+            else if (aactx->colorTemp < 0)
+               aactx->colorTemp = i;
+            else
+               break;
+         }
+      }
+
+      assert(aactx->colorTemp != aactx->tmp0);
+
+      tmp0 = aactx->tmp0;
+
+      /* declare new generic input/texcoord */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_INPUT;
+      decl.Declaration.Semantic = 1;
+      decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
+      decl.Semantic.SemanticIndex = aactx->maxGeneric + 1;
+      decl.Declaration.Interpolate = 1;
+      /* XXX this could be linear... */
+      decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = texInput;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new temp regs */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = tmp0;
+      ctx->emit_declaration(ctx, &decl);
+
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = aactx->colorTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      aactx->firstInstruction = FALSE;
+
+
+      /*
+       * Emit code to compute fragment coverage, kill if outside point radius
+       *
+       * Temp reg0 usage:
+       *  t0.x = distance of fragment from center point
+       *  t0.y = boolean, is t0.x > 1 ?
+       *  t0.z = temporary for computing 1/(1-k) value
+       *  t0.w = final coverage value
+       */
+
+      /* MUL t0.xy, tex, tex;  # compute x^2, y^2 */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XY;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* ADD t0.x, t0.x, t0.y;  # x^2 + y^2 */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_ADD;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      ctx->emit_instruction(ctx, &newInst);
+
+#if NORMALIZE  /* OPTIONAL normalization of length */
+      /* RSQ t0.x, t0.x; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_RSQ;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* RCP t0.x, t0.x; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_RCP;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      ctx->emit_instruction(ctx, &newInst);
+#endif
+
+      /* SGT t0.y, t0.xxxx, t0.wwww;  # bool b = d > 1 (NOTE t0.w == 1) */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_SGT;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* KILP -t0.yyyy;   # if b, KILL */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_KILP;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* SGT t0.y, t0.x, tex.z;  # bool b = distance > k */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_SGT;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+      newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* IF t0.y   # if b then */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_IF;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+      newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+      ctx->emit_instruction(ctx, &newInst);
+
+      {
+         /* compute coverage factor = (1-d)/(1-k) */
+
+         /* SUB t0.z, tex.w, tex.z;  # m = 1 - k */
+         newInst = tgsi_default_full_instruction();
+         newInst.Instruction.Opcode = TGSI_OPCODE_SUB;
+         newInst.Instruction.NumDstRegs = 1;
+         newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+         newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z;
+         newInst.Instruction.NumSrcRegs = 2;
+         newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+         newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+         newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
+         newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT;
+         newInst.FullSrcRegisters[1].SrcRegister.Index = texInput;
+         newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z;
+         ctx->emit_instruction(ctx, &newInst);
+
+         /* RCP t0.z, t0.z;  # t0.z = 1 / m */
+         newInst = tgsi_default_full_instruction();
+         newInst.Instruction.Opcode = TGSI_OPCODE_RCP;
+         newInst.Instruction.NumDstRegs = 1;
+         newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+         newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z;
+         newInst.Instruction.NumSrcRegs = 1;
+         newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+         newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z;
+         ctx->emit_instruction(ctx, &newInst);
+
+         /* SUB t0.x, 1, t0.x;  # d = 1 - d */
+         newInst = tgsi_default_full_instruction();
+         newInst.Instruction.Opcode = TGSI_OPCODE_SUB;
+         newInst.Instruction.NumDstRegs = 1;
+         newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+         newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X;
+         newInst.Instruction.NumSrcRegs = 2;
+         newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+         newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+         newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
+         newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
+         newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+         ctx->emit_instruction(ctx, &newInst);
+
+         /* MUL t0.w, t0.x, t0.z;   # coverage = d * m */
+         newInst = tgsi_default_full_instruction();
+         newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+         newInst.Instruction.NumDstRegs = 1;
+         newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+         newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+         newInst.Instruction.NumSrcRegs = 2;
+         newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0;
+         newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+         newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0;
+         newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z;
+         ctx->emit_instruction(ctx, &newInst);
+      }
+
+      /* ELSE */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_ELSE;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 0;
+      ctx->emit_instruction(ctx, &newInst);
+
+      {
+         /* MOV t0.w, tex.w;  # coverage = 1.0 */
+         newInst = tgsi_default_full_instruction();
+         newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
+         newInst.Instruction.NumDstRegs = 1;
+         newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+         newInst.FullDstRegisters[0].DstRegister.Index = tmp0;
+         newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+         newInst.Instruction.NumSrcRegs = 1;
+         newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+         newInst.FullSrcRegisters[0].SrcRegister.Index = texInput;
+         ctx->emit_instruction(ctx, &newInst);
+      }
+
+      /* ENDIF */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_ENDIF;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 0;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+
+   if (inst->Instruction.Opcode == TGSI_OPCODE_END) {
+      /* add alpha modulation code at tail of program */
+
+      /* MOV result.color.xyz, colorTemp; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MOV;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* MUL result.color.w, colorTemp, tmp0.w; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+      newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput;
+      newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->tmp0;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+   else {
+      /* Not an END instruction.
+       * Look for writes to result.color and replace with colorTemp reg.
+       */
+      uint i;
+
+      for (i = 0; i < inst->Instruction.NumDstRegs; i++) {
+         struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i];
+         if (dst->DstRegister.File == TGSI_FILE_OUTPUT &&
+             dst->DstRegister.Index == aactx->colorOutput) {
+            dst->DstRegister.File = TGSI_FILE_TEMPORARY;
+            dst->DstRegister.Index = aactx->colorTemp;
+         }
+      }
+   }
+
+   ctx->emit_instruction(ctx, inst);
+}
+
+
+/**
+ * Generate the frag shader we'll use for drawing AA lines.
+ * This will be the user's shader plus some texture/modulate instructions.
+ */
+static void
+generate_aapoint_fs(struct aapoint_stage *aapoint)
+{
+   const struct pipe_shader_state *orig_fs = &aapoint->fs->state;
+   struct draw_context *draw = aapoint->stage.draw;
+   struct pipe_shader_state aapoint_fs;
+   struct aa_transform_context transform;
+
+#define MAX 1000
+
+   aapoint_fs = *orig_fs; /* copy to init */
+   aapoint_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX);
+
+   memset(&transform, 0, sizeof(transform));
+   transform.colorOutput = -1;
+   transform.maxInput = -1;
+   transform.maxGeneric = -1;
+   transform.colorTemp = -1;
+   transform.tmp0 = -1;
+   transform.firstInstruction = TRUE;
+   transform.base.transform_instruction = aa_transform_inst;
+   transform.base.transform_declaration = aa_transform_decl;
+
+   tgsi_transform_shader(orig_fs->tokens,
+                         (struct tgsi_token *) aapoint_fs.tokens,
+                         MAX, &transform.base);
+
+#if 0 /* DEBUG */
+   tgsi_dump(orig_fs->tokens, 0);
+   tgsi_dump(aapoint_fs.tokens, 0);
+#endif
+
+   aapoint_fs.input_semantic_name[aapoint_fs.num_inputs] = TGSI_SEMANTIC_GENERIC;
+   aapoint_fs.input_semantic_index[aapoint_fs.num_inputs] = transform.maxGeneric + 1;
+   aapoint_fs.num_inputs++;
+
+   aapoint->fs->aapoint_fs
+      = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs);
+
+   /* advertise the extra post-transform vertex attributes which will have
+    * the texcoords.
+    */
+   draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC;
+   draw->extra_vp_outputs.semantic_index = transform.maxGeneric + 1;
+}
+
+
+/**
+ * When we're about to draw our first AA line in a batch, this function is
+ * called to tell the driver to bind our modified fragment shader.
+ */
+static void
+bind_aapoint_fragment_shader(struct aapoint_stage *aapoint)
+{
+   if (!aapoint->fs->aapoint_fs) {
+      generate_aapoint_fs(aapoint);
+   }
+   aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs);
+}
+
+
+
+static INLINE struct aapoint_stage *
+aapoint_stage( struct draw_stage *stage )
+{
+   return (struct aapoint_stage *) stage;
+}
+
+
+static void
+passthrough_line(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->line(stage->next, header);
+}
+
+
+static void
+passthrough_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->tri(stage->next, header);
+}
+
+
+/**
+ * Draw an AA point by drawing a quad.
+ */
+static void
+aapoint_point(struct draw_stage *stage, struct prim_header *header)
+{
+   const struct aapoint_stage *aapoint = aapoint_stage(stage);
+   struct prim_header tri;
+   struct vertex_header *v[4];
+   uint texPos = aapoint->tex_slot;
+   float radius, *pos, *tex;
+   uint i;
+   float k;
+
+   if (aapoint->psize_slot >= 0) {
+      radius = 0.5f * header->v[0]->data[aapoint->psize_slot][0];
+   }
+   else {
+      radius = aapoint->radius;
+   }
+
+   /*
+    * Note: the texcoords (generic attrib, really) we use are special:
+    * The S and T components simply vary from -1 to +1.
+    * The R component is k, below.
+    * The Q component is 1.0 and will used as a handy constant in the
+    * fragment shader.
+    */
+
+   /*
+    * k is the threshold distance from the point's center at which
+    * we begin alpha attenuation (the coverage value).
+    * Operating within a unit circle, we'll compute the fragment's
+    * distance 'd' from the center point using the texcoords.
+    * IF d > 1.0 THEN
+    *    KILL fragment
+    * ELSE IF d > k THEN
+    *    compute coverage in [0,1] proportional to d in [k, 1].
+    * ELSE
+    *    coverage = 1.0;  // full coverage
+    * ENDIF
+    */
+
+#if !NORMALIZE
+   k = 1.0 / radius;
+   k = 1.0 - 2.0 * k + k * k;
+#else
+   k = 1.0 - 1.0 / radius;
+#endif
+
+   /* allocate/dup new verts */
+   for (i = 0; i < 4; i++) {
+      v[i] = dup_vert(stage, header->v[0], i);
+   }
+
+   /* new verts */
+   pos = v[0]->data[0];
+   pos[0] -= radius;
+   pos[1] -= radius;
+
+   pos = v[1]->data[0];
+   pos[0] += radius;
+   pos[1] -= radius;
+
+   pos = v[2]->data[0];
+   pos[0] += radius;
+   pos[1] += radius;
+
+   pos = v[3]->data[0];
+   pos[0] -= radius;
+   pos[1] += radius;
+
+   /* new texcoords */
+   tex = v[0]->data[texPos];
+   ASSIGN_4V(tex, -1, -1, k, 1);
+
+   tex = v[1]->data[texPos];
+   ASSIGN_4V(tex,  1, -1, k, 1);
+
+   tex = v[2]->data[texPos];
+   ASSIGN_4V(tex,  1,  1, k, 1);
+
+   tex = v[3]->data[texPos];
+   ASSIGN_4V(tex, -1,  1, k, 1);
+
+   /* emit 2 tris for the quad strip */
+   tri.v[0] = v[0];
+   tri.v[1] = v[1];
+   tri.v[2] = v[2];
+   stage->next->tri( stage->next, &tri );
+
+   tri.v[0] = v[0];
+   tri.v[1] = v[2];
+   tri.v[2] = v[3];
+   stage->next->tri( stage->next, &tri );
+}
+
+
+static void
+aapoint_first_point(struct draw_stage *stage, struct prim_header *header)
+{
+   auto struct aapoint_stage *aapoint = aapoint_stage(stage);
+   struct draw_context *draw = stage->draw;
+
+   assert(draw->rasterizer->point_smooth);
+
+   if (draw->rasterizer->point_size <= 2.0)
+      aapoint->radius = 1.0;
+   else
+      aapoint->radius = 0.5f * draw->rasterizer->point_size;
+
+   aapoint->tex_slot = draw->num_vs_outputs;
+   assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */
+   draw->extra_vp_outputs.slot = aapoint->tex_slot;
+
+   /*
+    * Bind our fragprog.
+    */
+   bind_aapoint_fragment_shader(aapoint);
+
+   /* find psize slot in post-transform vertex */
+   aapoint->psize_slot = -1;
+   if (draw->rasterizer->point_size_per_vertex) {
+      /* find PSIZ vertex output */
+      const struct draw_vertex_shader *vs = draw->vertex_shader;
+      uint i;
+      for (i = 0; i < vs->state->num_outputs; i++) {
+         if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) {
+            aapoint->psize_slot = i;
+            break;
+         }
+      }
+   }
+
+   /* now really draw first line */
+   stage->point = aapoint_point;
+   stage->point(stage, header);
+}
+
+
+static void
+aapoint_flush(struct draw_stage *stage, unsigned flags)
+{
+   struct draw_context *draw = stage->draw;
+   struct aapoint_stage *aapoint = aapoint_stage(stage);
+   struct pipe_context *pipe = aapoint->pipe;
+
+   stage->point = aapoint_first_point;
+   stage->next->flush( stage->next, flags );
+
+   /* restore original frag shader */
+   aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs);
+
+   draw->extra_vp_outputs.slot = 0;
+}
+
+
+static void
+aapoint_reset_stipple_counter(struct draw_stage *stage)
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void
+aapoint_destroy(struct draw_stage *stage)
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+static struct aapoint_stage *
+draw_aapoint_stage(struct draw_context *draw)
+{
+   struct aapoint_stage *aapoint = CALLOC_STRUCT(aapoint_stage);
+
+   draw_alloc_temp_verts( &aapoint->stage, 4 );
+
+   aapoint->stage.draw = draw;
+   aapoint->stage.next = NULL;
+   aapoint->stage.point = aapoint_first_point;
+   aapoint->stage.line = passthrough_line;
+   aapoint->stage.tri = passthrough_tri;
+   aapoint->stage.flush = aapoint_flush;
+   aapoint->stage.reset_stipple_counter = aapoint_reset_stipple_counter;
+   aapoint->stage.destroy = aapoint_destroy;
+
+   return aapoint;
+}
+
+
+/*
+ * XXX temporary? solution to mapping a pipe_context to a aapoint_stage.
+ */
+
+#define MAX_CONTEXTS 10
+
+static struct pipe_context *Pipe[MAX_CONTEXTS];
+static struct aapoint_stage *Stage[MAX_CONTEXTS];
+static uint NumContexts;
+
+static void
+add_aa_pipe_context(struct pipe_context *pipe, struct aapoint_stage *aa)
+{
+   assert(NumContexts < MAX_CONTEXTS);
+   Pipe[NumContexts] = pipe;
+   Stage[NumContexts] = aa;
+   NumContexts++;
+}
+
+static struct aapoint_stage *
+aapoint_stage_from_pipe(struct pipe_context *pipe)
+{
+   uint i;
+   for (i = 0; i < NumContexts; i++) {
+      if (Pipe[i] == pipe)
+         return Stage[i];
+   }
+   assert(0);
+   return NULL;
+}
+
+
+/**
+ * This function overrides the driver's create_fs_state() function and
+ * will typically be called by the state tracker.
+ */
+static void *
+aapoint_create_fs_state(struct pipe_context *pipe,
+                       const struct pipe_shader_state *fs)
+{
+   struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe);
+   struct aapoint_fragment_shader *aafs = CALLOC_STRUCT(aapoint_fragment_shader);
+
+   if (aafs) {
+      aafs->state = *fs;
+
+      /* pass-through */
+      aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs);
+   }
+
+   return aafs;
+}
+
+
+static void
+aapoint_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe);
+   struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs;
+   /* save current */
+   aapoint->fs = aafs;
+   /* pass-through */
+   aapoint->driver_bind_fs_state(aapoint->pipe, aafs->driver_fs);
+}
+
+
+static void
+aapoint_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe);
+   struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs;
+   /* pass-through */
+   aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs);
+   FREE(aafs);
+}
+
+
+/**
+ * Called by drivers that want to install this AA point prim stage
+ * into the draw module's pipeline.  This will not be used if the
+ * hardware has native support for AA points.
+ */
+void
+draw_install_aapoint_stage(struct draw_context *draw,
+                           struct pipe_context *pipe)
+{
+   struct aapoint_stage *aapoint;
+
+   /*
+    * Create / install AA point drawing / prim stage
+    */
+   aapoint = draw_aapoint_stage( draw );
+   assert(aapoint);
+   draw->pipeline.aapoint = &aapoint->stage;
+
+   aapoint->pipe = pipe;
+
+   /* save original driver functions */
+   aapoint->driver_create_fs_state = pipe->create_fs_state;
+   aapoint->driver_bind_fs_state = pipe->bind_fs_state;
+   aapoint->driver_delete_fs_state = pipe->delete_fs_state;
+
+   /* override the driver's functions */
+   pipe->create_fs_state = aapoint_create_fs_state;
+   pipe->bind_fs_state = aapoint_bind_fs_state;
+   pipe->delete_fs_state = aapoint_delete_fs_state;
+
+   add_aa_pipe_context(pipe, aapoint);
+}
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index 4be3830316..c28e78d33a 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -72,10 +72,10 @@ struct draw_context *draw_create( void )
    {
       uint i;
       const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f;
-      char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16);
+      char *tmp = align_malloc(Elements(draw->vs.queue) * size, 16);
 
-      for (i = 0; i < Elements(draw->vcache.vertex); i++)
-	 draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size);
+      for (i = 0; i < Elements(draw->vs.queue); i++)
+	 draw->vs.queue[i].vertex = (struct vertex_header *)(tmp + i * size);
    }
 
    draw->shader_queue_flush = draw_vertex_shader_queue_flush;
@@ -103,10 +103,14 @@ void draw_destroy( struct draw_context *draw )
    draw->pipeline.flatshade->destroy( draw->pipeline.flatshade );
    draw->pipeline.cull->destroy( draw->pipeline.cull );
    draw->pipeline.validate->destroy( draw->pipeline.validate );
+   if (draw->pipeline.aaline)
+      draw->pipeline.aaline->destroy( draw->pipeline.aaline );
+   if (draw->pipeline.aapoint)
+      draw->pipeline.aapoint->destroy( draw->pipeline.aapoint );
    if (draw->pipeline.rasterize)
       draw->pipeline.rasterize->destroy( draw->pipeline.rasterize );
    tgsi_exec_machine_free_data(&draw->machine);
-   align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */
+   align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */
    FREE( draw );
 }
 
@@ -240,6 +244,26 @@ draw_convert_wide_lines(struct draw_context *draw, boolean enable)
 
 
 /**
+ * The draw module may sometimes generate vertices with extra attributes
+ * (such as texcoords for AA lines).  The driver can call this function
+ * to find those attributes.
+ */
+int
+draw_find_vs_output(struct draw_context *draw,
+                    uint semantic_name, uint semantic_index)
+{
+   /* 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;
+   }
+   return 0;
+}
+
+
+/**
  * Allocate space for temporary post-transform vertices, such as for clipping.
  */
 void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr )
diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h
index ddeb184497..c25301f71d 100644
--- a/src/gallium/auxiliary/draw/draw_context.h
+++ b/src/gallium/auxiliary/draw/draw_context.h
@@ -41,6 +41,7 @@
 #include "pipe/p_state.h"
 
 
+struct pipe_context;
 struct vertex_buffer;
 struct vertex_info;
 struct draw_context;
@@ -93,6 +94,26 @@ void draw_convert_wide_points(struct draw_context *draw, boolean enable);
 
 void draw_convert_wide_lines(struct draw_context *draw, boolean enable);
 
+boolean draw_use_sse(struct draw_context *draw);
+
+void
+draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe);
+
+void
+draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe);
+
+void
+draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe);
+
+
+int
+draw_find_vs_output(struct draw_context *draw,
+                    uint semantic_name, uint semantic_index);
+
+
+/*
+ * Vertex shader functions
+ */
 
 struct draw_vertex_shader *
 draw_create_vertex_shader(struct draw_context *draw,
@@ -102,7 +123,11 @@ void draw_bind_vertex_shader(struct draw_context *draw,
 void draw_delete_vertex_shader(struct draw_context *draw,
                                struct draw_vertex_shader *dvs);
 
-boolean draw_use_sse(struct draw_context *draw);
+
+
+/*
+ * Vertex data functions
+ */
 
 void draw_set_vertex_buffer(struct draw_context *draw,
 			    unsigned attr,
diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c
index 51e2242719..7d6cd43410 100644
--- a/src/gallium/auxiliary/draw/draw_prim.c
+++ b/src/gallium/auxiliary/draw/draw_prim.c
@@ -114,6 +114,7 @@ static void draw_prim_queue_flush( struct draw_context *draw )
    }
 
    draw->pq.queue_nr = 0;   
+   draw->vs.post_nr = 0;   
    draw_vertex_cache_unreference( draw );
 }
 
@@ -121,11 +122,15 @@ static void draw_prim_queue_flush( struct draw_context *draw )
 
 void draw_do_flush( struct draw_context *draw, unsigned flags )
 {
+   static boolean flushing = FALSE;
+
    if (0)
       debug_printf("Flushing with %d verts, %d prims\n",
                    draw->vs.queue_nr,
                    draw->pq.queue_nr );
 
+   if (!flushing) {
+      flushing = TRUE;
 
    if (flags >= DRAW_FLUSH_SHADER_QUEUE) {
       if (draw->vs.queue_nr)
@@ -146,6 +151,9 @@ void draw_do_flush( struct draw_context *draw, unsigned flags )
 	 }
       }    
    }
+
+      flushing = FALSE;
+   }
 }
 
 
@@ -219,6 +227,7 @@ static void do_triangle( struct draw_context *draw,
 {
    struct prim_header *prim = get_queued_prim( draw, 3 );
    
+//   _mesa_printf("tri %d %d %d\n", i0, i1, i2);
    prim->reset_line_stipple = 1;
    prim->edgeflags = ~0;
    prim->pad = 0;
diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h
index bc11259cb2..6abced139b 100644
--- a/src/gallium/auxiliary/draw/draw_private.h
+++ b/src/gallium/auxiliary/draw/draw_private.h
@@ -44,10 +44,11 @@
 #include "pipe/p_state.h"
 #include "pipe/p_defines.h"
 
-#include "x86/rtasm/x86sse.h"
+#include "rtasm/rtasm_x86sse.h"
 #include "tgsi/exec/tgsi_exec.h"
 
 
+struct pipe_context;
 struct gallivm_prog;
 struct gallivm_cpu_engine;
 
@@ -119,7 +120,7 @@ struct draw_stage
 };
 
 
-#define PRIM_QUEUE_LENGTH      16
+#define PRIM_QUEUE_LENGTH      32
 #define VCACHE_SIZE            32
 #define VCACHE_OVERFLOW        4
 #define VS_QUEUE_LENGTH        (VCACHE_SIZE + VCACHE_OVERFLOW + 1)	/* can never fill up */
@@ -179,6 +180,9 @@ struct draw_context
       struct draw_stage *offset;
       struct draw_stage *unfilled;
       struct draw_stage *stipple;
+      struct draw_stage *aapoint;
+      struct draw_stage *aaline;
+      struct draw_stage *pstipple;
       struct draw_stage *wide;
       struct draw_stage *rasterize;
    } pipeline;
@@ -212,9 +216,17 @@ struct draw_context
    unsigned nr_planes;
 
    boolean convert_wide_points; /**< convert wide points to tris? */
-   boolean convert_wide_lines;  /**< convert side lines to tris? */
+   boolean convert_wide_lines;  /**< convert wide lines to tris? */
    boolean use_sse;
 
+   /* If a prim stage introduces new vertex attributes, they'll be stored here
+    */
+   struct {
+      uint semantic_name;
+      uint semantic_index;
+      int slot;
+   } extra_vp_outputs;
+
    unsigned reduced_prim;
 
    /** TGSI program interpreter runtime state */
@@ -234,8 +246,12 @@ struct draw_context
     */
    struct {
       unsigned referenced;  /**< bitfield */
-      unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW];
-      struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW];
+
+      struct {
+	 unsigned in;		/* client array element */
+	 unsigned out;		/* index in vs queue/array */
+      } idx[VCACHE_SIZE + VCACHE_OVERFLOW];
+
       unsigned overflow;
 
       /** To find space in the vertex cache: */
@@ -248,9 +264,10 @@ struct draw_context
    struct {
       struct {
 	 unsigned elt;   /**< index into the user's vertex arrays */
-	 struct vertex_header *dest; /**< points into vcache.vertex[] array */
+	 struct vertex_header *vertex;
       } queue[VS_QUEUE_LENGTH];
       unsigned queue_nr;
+      unsigned post_nr;
    } vs;
 
    /**
diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c
new file mode 100644
index 0000000000..4048abf856
--- /dev/null
+++ b/src/gallium/auxiliary/draw/draw_pstipple.c
@@ -0,0 +1,717 @@
+/**************************************************************************
+ * 
+ * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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.
+ * 
+ **************************************************************************/
+
+/**
+ * Polygon stipple stage:  implement polygon stipple with texture map and
+ * fragment program.  The fragment program samples the texture and does
+ * a fragment kill for the stipple-failing fragments.
+ *
+ * Authors:  Brian Paul
+ */
+
+
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "tgsi/util/tgsi_transform.h"
+#include "tgsi/util/tgsi_dump.h"
+
+#include "draw_context.h"
+#include "draw_private.h"
+
+
+
+/**
+ * Subclass of pipe_shader_state to carry extra fragment shader info.
+ */
+struct pstip_fragment_shader
+{
+   struct pipe_shader_state state;
+   void *driver_fs;
+   void *pstip_fs;
+};
+
+
+/**
+ * Subclass of draw_stage
+ */
+struct pstip_stage
+{
+   struct draw_stage stage;
+
+   void *sampler_cso;
+   struct pipe_texture *texture;
+   uint sampler_unit;
+
+   /*
+    * Currently bound state
+    */
+   struct pstip_fragment_shader *fs;
+   struct {
+      void *sampler[PIPE_MAX_SAMPLERS];
+      struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
+      const struct pipe_poly_stipple *stipple;
+   } state;
+
+   /*
+    * Driver interface/override functions
+    */
+   void * (*driver_create_fs_state)(struct pipe_context *,
+                                    const struct pipe_shader_state *);
+   void (*driver_bind_fs_state)(struct pipe_context *, void *);
+   void (*driver_delete_fs_state)(struct pipe_context *, void *);
+
+   void (*driver_bind_sampler_state)(struct pipe_context *, unsigned, void *);
+
+   void (*driver_set_sampler_texture)(struct pipe_context *,
+                                      unsigned sampler,
+                                      struct pipe_texture *);
+
+   void (*driver_set_polygon_stipple)(struct pipe_context *,
+                                      const struct pipe_poly_stipple *);
+
+   struct pipe_context *pipe;
+};
+
+
+
+/**
+ * Subclass of tgsi_transform_context, used for transforming the
+ * user's fragment shader to add the special AA instructions.
+ */
+struct pstip_transform_context {
+   struct tgsi_transform_context base;
+   uint tempsUsed;  /**< bitmask */
+   int wincoordInput;
+   int maxInput;
+   int maxSampler;  /**< max sampler index found */
+   int texTemp;  /**< temp registers */
+   int numImmed;
+   boolean firstInstruction;
+};
+
+
+/**
+ * TGSI declaration transform callback.
+ * Look for a free sampler, a free input attrib, and two free temp regs.
+ */
+static void
+pstip_transform_decl(struct tgsi_transform_context *ctx,
+                     struct tgsi_full_declaration *decl)
+{
+   struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+
+   if (decl->Declaration.File == TGSI_FILE_SAMPLER) {
+      if ((int) decl->u.DeclarationRange.Last > pctx->maxSampler)
+         pctx->maxSampler = (int) decl->u.DeclarationRange.Last;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_INPUT) {
+      pctx->maxInput = MAX2(pctx->maxInput, decl->u.DeclarationRange.Last);
+      if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION)
+         pctx->wincoordInput = (int) decl->u.DeclarationRange.First;
+   }
+   else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) {
+      uint i;
+      for (i = decl->u.DeclarationRange.First;
+           i <= decl->u.DeclarationRange.Last; i++) {
+         pctx->tempsUsed |= (1 << i);
+      }
+   }
+
+   ctx->emit_declaration(ctx, decl);
+}
+
+
+static void
+pstip_transform_immed(struct tgsi_transform_context *ctx,
+                      struct tgsi_full_immediate *immed)
+{
+   struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+   pctx->numImmed++;
+}
+
+
+/**
+ * TGSI instruction transform callback.
+ * Replace writes to result.color w/ a temp reg.
+ * Upon END instruction, insert texture sampling code for antialiasing.
+ */
+static void
+pstip_transform_inst(struct tgsi_transform_context *ctx,
+                     struct tgsi_full_instruction *inst)
+{
+   struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx;
+
+   if (pctx->firstInstruction) {
+      /* emit our new declarations before the first instruction */
+
+      struct tgsi_full_declaration decl;
+      struct tgsi_full_instruction newInst;
+      uint i;
+      int wincoordInput;
+      const int sampler = pctx->maxSampler + 1;
+
+      if (pctx->wincoordInput < 0)
+         wincoordInput = pctx->maxInput + 1;
+      else
+         wincoordInput = pctx->wincoordInput;
+
+      /* find one free temp reg */
+      for (i = 0; i < 32; i++) {
+         if ((pctx->tempsUsed & (1 << i)) == 0) {
+            /* found a free temp */
+            if (pctx->texTemp < 0)
+               pctx->texTemp  = i;
+            else
+               break;
+         }
+      }
+      assert(pctx->texTemp >= 0);
+
+      if (pctx->wincoordInput < 0) {
+         /* declare new position input reg */
+         decl = tgsi_default_full_declaration();
+         decl.Declaration.File = TGSI_FILE_INPUT;
+         decl.Declaration.Semantic = 1;
+         decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION;
+         decl.Semantic.SemanticIndex = 0;
+         decl.Declaration.Interpolate = 1;
+         decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */
+         decl.u.DeclarationRange.First = 
+            decl.u.DeclarationRange.Last = wincoordInput;
+         ctx->emit_declaration(ctx, &decl);
+      }
+
+      /* declare new sampler */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_SAMPLER;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = sampler;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* declare new temp regs */
+      decl = tgsi_default_full_declaration();
+      decl.Declaration.File = TGSI_FILE_TEMPORARY;
+      decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = pctx->texTemp;
+      ctx->emit_declaration(ctx, &decl);
+
+      /* emit immediate = {1/32, 1/32, 1, 1}
+       * The index/position of this immediate will be pctx->numImmed
+       */
+      {
+         static const float value[4] = { 1.0/32, 1.0/32, 1.0, 1.0 };
+         struct tgsi_full_immediate immed;
+         uint size = 4;
+         immed = tgsi_default_full_immediate();
+         immed.Immediate.Size = 1 + size; /* one for the token itself */
+         immed.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value;
+         ctx->emit_immediate(ctx, &immed);
+      }
+
+      pctx->firstInstruction = FALSE;
+
+
+      /* 
+       * Insert new MUL/TEX/KILP instructions at start of program
+       * Take gl_FragCoord, divide by 32 (stipple size), sample the
+       * texture and kill fragment if needed.
+       *
+       * We'd like to use non-normalized texcoords to index into a RECT
+       * texture, but we can only use GL_REPEAT wrap mode with normalized
+       * texcoords.  Darn.
+       */
+
+      /* MUL texTemp, INPUT[wincoord], 1/32; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_MUL;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = wincoordInput;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_IMMEDIATE;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->numImmed;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* TEX texTemp, texTemp, sampler; */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_TEX;
+      newInst.Instruction.NumDstRegs = 1;
+      newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp;
+      newInst.Instruction.NumSrcRegs = 2;
+      newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp;
+      newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
+      newInst.FullSrcRegisters[1].SrcRegister.Index = sampler;
+      ctx->emit_instruction(ctx, &newInst);
+
+      /* KILP texTemp;   # if texTemp < 0, KILL fragment */
+      newInst = tgsi_default_full_instruction();
+      newInst.Instruction.Opcode = TGSI_OPCODE_KILP;
+      newInst.Instruction.NumDstRegs = 0;
+      newInst.Instruction.NumSrcRegs = 1;
+      newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
+      newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp;
+      newInst.FullSrcRegisters[0].SrcRegister.Negate = 1;
+      ctx->emit_instruction(ctx, &newInst);
+   }
+
+   /* emit this instruction */
+   ctx->emit_instruction(ctx, inst);
+}
+
+
+/**
+ * Generate the frag shader we'll use for doing polygon stipple.
+ * This will be the user's shader prefixed with a TEX and KIL instruction.
+ */
+static void
+generate_pstip_fs(struct pstip_stage *pstip)
+{
+   const struct pipe_shader_state *orig_fs = &pstip->fs->state;
+   /*struct draw_context *draw = pstip->stage.draw;*/
+   struct pipe_shader_state pstip_fs;
+   struct pstip_transform_context transform;
+
+#define MAX 1000
+
+   pstip_fs = *orig_fs; /* copy to init */
+   pstip_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX);
+
+   memset(&transform, 0, sizeof(transform));
+   transform.wincoordInput = -1;
+   transform.maxInput = -1;
+   transform.maxSampler = -1;
+   transform.texTemp = -1;
+   transform.firstInstruction = TRUE;
+   transform.base.transform_instruction = pstip_transform_inst;
+   transform.base.transform_declaration = pstip_transform_decl;
+   transform.base.transform_immediate = pstip_transform_immed;
+
+   tgsi_transform_shader(orig_fs->tokens,
+                         (struct tgsi_token *) pstip_fs.tokens,
+                         MAX, &transform.base);
+
+#if 1 /* DEBUG */
+   tgsi_dump(orig_fs->tokens, 0);
+   tgsi_dump(pstip_fs.tokens, 0);
+#endif
+
+   pstip->sampler_unit = transform.maxSampler + 1;
+
+   if (transform.wincoordInput < 0) {
+      pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION;
+      pstip_fs.input_semantic_index[pstip_fs.num_inputs] = transform.maxInput;
+      pstip_fs.num_inputs++;
+   }
+
+   pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs);
+}
+
+
+/**
+ * Load texture image with current stipple pattern.
+ */
+static void
+pstip_update_texture(struct pstip_stage *pstip)
+{
+   static const uint bit31 = 1 << 31;
+   struct pipe_context *pipe = pstip->pipe;
+   struct pipe_surface *surface;
+   const uint *stipple = pstip->state.stipple->stipple;
+   uint i, j;
+   ubyte *data;
+
+   surface = pipe->get_tex_surface(pipe, pstip->texture, 0, 0, 0);
+   data = pipe_surface_map(surface);
+
+   /*
+    * Load alpha texture.
+    * Note: 0 means keep the fragment, 255 means kill it.
+    * We'll negate the texel value and use KILP which kills if value
+    * is negative.
+    */
+   for (i = 0; i < 32; i++) {
+      for (j = 0; j < 32; j++) {
+         if (stipple[i] & (bit31 >> j)) {
+            /* fragment "on" */
+            data[i * surface->pitch + j] = 0;
+         }
+         else {
+            /* fragment "off" */
+            data[i * surface->pitch + j] = 255;
+         }
+      }
+   }
+
+   /* unmap */
+   pipe_surface_unmap(surface);
+   pipe_surface_reference(&surface, NULL);
+   pipe->texture_update(pipe, pstip->texture);
+}
+
+
+/**
+ * Create the texture map we'll use for stippling.
+ */
+static void
+pstip_create_texture(struct pstip_stage *pstip)
+{
+   struct pipe_context *pipe = pstip->pipe;
+   struct pipe_texture texTemp;
+
+   memset(&texTemp, 0, sizeof(texTemp));
+   texTemp.target = PIPE_TEXTURE_2D;
+   texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */
+   texTemp.last_level = 0;
+   texTemp.width[0] = 32;
+   texTemp.height[0] = 32;
+   texTemp.depth[0] = 1;
+   texTemp.cpp = 1;
+
+   pstip->texture = pipe->texture_create(pipe, &texTemp);
+
+   //pstip_update_texture(pstip);
+}
+
+
+/**
+ * Create the sampler CSO that'll be used for antialiasing.
+ * By using a mipmapped texture, we don't have to generate a different
+ * texture image for each line size.
+ */
+static void
+pstip_create_sampler(struct pstip_stage *pstip)
+{
+   struct pipe_sampler_state sampler;
+   struct pipe_context *pipe = pstip->pipe;
+
+   memset(&sampler, 0, sizeof(sampler));
+   sampler.wrap_s = PIPE_TEX_WRAP_REPEAT;
+   sampler.wrap_t = PIPE_TEX_WRAP_REPEAT;
+   sampler.wrap_r = PIPE_TEX_WRAP_REPEAT;
+   sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+   sampler.normalized_coords = 1;
+   sampler.min_lod = 0.0f;
+   sampler.max_lod = 0.0f;
+
+   pstip->sampler_cso = pipe->create_sampler_state(pipe, &sampler);
+}
+
+
+/**
+ * When we're about to draw our first AA line in a batch, this function is
+ * called to tell the driver to bind our modified fragment shader.
+ */
+static void
+bind_pstip_fragment_shader(struct pstip_stage *pstip)
+{
+   if (!pstip->fs->pstip_fs) {
+      generate_pstip_fs(pstip);
+   }
+   pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs);
+}
+
+
+
+static INLINE struct pstip_stage *
+pstip_stage( struct draw_stage *stage )
+{
+   return (struct pstip_stage *) stage;
+}
+
+
+static void
+passthrough_point(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->point(stage->next, header);
+}
+
+
+static void
+passthrough_line(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->line(stage->next, header);
+}
+
+
+static void
+passthrough_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   stage->next->tri(stage->next, header);
+}
+
+
+
+static void
+pstip_first_tri(struct draw_stage *stage, struct prim_header *header)
+{
+   struct pstip_stage *pstip = pstip_stage(stage);
+   struct draw_context *draw = stage->draw;
+   struct pipe_context *pipe = pstip->pipe;
+
+   assert(draw->rasterizer->poly_stipple_enable);
+
+   /*
+    * Bind our fragprog, sampler and texture
+    */
+   bind_pstip_fragment_shader(pstip);
+
+   pstip->driver_bind_sampler_state(pipe, pstip->sampler_unit, pstip->sampler_cso);
+   pstip->driver_set_sampler_texture(pipe, pstip->sampler_unit, pstip->texture);
+
+   /* now really draw first line */
+   stage->tri = passthrough_tri;
+   stage->tri(stage, header);
+}
+
+
+static void
+pstip_flush(struct draw_stage *stage, unsigned flags)
+{
+   /*struct draw_context *draw = stage->draw;*/
+   struct pstip_stage *pstip = pstip_stage(stage);
+   struct pipe_context *pipe = pstip->pipe;
+
+   stage->tri = pstip_first_tri;
+   stage->next->flush( stage->next, flags );
+
+   /* restore original frag shader */
+   pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs);
+
+   /* XXX restore original texture, sampler state */
+   pstip->driver_bind_sampler_state(pipe, pstip->sampler_unit,
+                                 pstip->state.sampler[pstip->sampler_unit]);
+   pstip->driver_set_sampler_texture(pipe, pstip->sampler_unit,
+                                 pstip->state.texture[pstip->sampler_unit]);
+}
+
+
+static void
+pstip_reset_stipple_counter(struct draw_stage *stage)
+{
+   stage->next->reset_stipple_counter( stage->next );
+}
+
+
+static void
+pstip_destroy(struct draw_stage *stage)
+{
+   draw_free_temp_verts( stage );
+   FREE( stage );
+}
+
+
+static struct pstip_stage *
+draw_pstip_stage(struct draw_context *draw)
+{
+   struct pstip_stage *pstip = CALLOC_STRUCT(pstip_stage);
+
+   draw_alloc_temp_verts( &pstip->stage, 8 );
+
+   pstip->stage.draw = draw;
+   pstip->stage.next = NULL;
+   pstip->stage.point = passthrough_point;
+   pstip->stage.line = passthrough_line;
+   pstip->stage.tri = pstip_first_tri;
+   pstip->stage.flush = pstip_flush;
+   pstip->stage.reset_stipple_counter = pstip_reset_stipple_counter;
+   pstip->stage.destroy = pstip_destroy;
+
+   return pstip;
+}
+
+
+/*
+ * XXX temporary? solution to mapping a pipe_context to a pstip_stage.
+ */
+
+#define MAX_CONTEXTS 10
+
+static struct pipe_context *Pipe[MAX_CONTEXTS];
+static struct pstip_stage *Stage[MAX_CONTEXTS];
+static uint NumContexts;
+
+static void
+add_pstip_pipe_context(struct pipe_context *pipe, struct pstip_stage *pstip)
+{
+   assert(NumContexts < MAX_CONTEXTS);
+   Pipe[NumContexts] = pipe;
+   Stage[NumContexts] = pstip;
+   NumContexts++;
+}
+
+static struct pstip_stage *
+pstip_stage_from_pipe(struct pipe_context *pipe)
+{
+   uint i;
+   for (i = 0; i < NumContexts; i++) {
+      if (Pipe[i] == pipe)
+         return Stage[i];
+   }
+   assert(0);
+   return NULL;
+}
+
+
+/**
+ * This function overrides the driver's create_fs_state() function and
+ * will typically be called by the state tracker.
+ */
+static void *
+pstip_create_fs_state(struct pipe_context *pipe,
+                       const struct pipe_shader_state *fs)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   struct pstip_fragment_shader *aafs = CALLOC_STRUCT(pstip_fragment_shader);
+
+   if (aafs) {
+      aafs->state = *fs;
+
+      /* pass-through */
+      aafs->driver_fs = pstip->driver_create_fs_state(pstip->pipe, fs);
+   }
+
+   return aafs;
+}
+
+
+static void
+pstip_bind_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs;
+   /* save current */
+   pstip->fs = aafs;
+   /* pass-through */
+   pstip->driver_bind_fs_state(pstip->pipe, aafs->driver_fs);
+}
+
+
+static void
+pstip_delete_fs_state(struct pipe_context *pipe, void *fs)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs;
+   /* pass-through */
+   pstip->driver_delete_fs_state(pstip->pipe, aafs->driver_fs);
+   FREE(aafs);
+}
+
+
+static void
+pstip_bind_sampler_state(struct pipe_context *pipe,
+                         unsigned unit, void *sampler)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   /* save current */
+   pstip->state.sampler[unit] = sampler;
+   /* pass-through */
+   pstip->driver_bind_sampler_state(pstip->pipe, unit, sampler);
+}
+
+
+static void
+pstip_set_sampler_texture(struct pipe_context *pipe,
+                          unsigned sampler, struct pipe_texture *texture)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   /* save current */
+   pstip->state.texture[sampler] = texture;
+   /* pass-through */
+   pstip->driver_set_sampler_texture(pstip->pipe, sampler, texture);
+}
+
+
+static void
+pstip_set_polygon_stipple(struct pipe_context *pipe,
+                          const struct pipe_poly_stipple *stipple)
+{
+   struct pstip_stage *pstip = pstip_stage_from_pipe(pipe);
+   /* save current */
+   pstip->state.stipple = stipple;
+   /* pass-through */
+   pstip->driver_set_polygon_stipple(pstip->pipe, stipple);
+
+   pstip_update_texture(pstip);
+}
+
+
+
+/**
+ * Called by drivers that want to install this AA line prim stage
+ * into the draw module's pipeline.  This will not be used if the
+ * hardware has native support for AA lines.
+ */
+void
+draw_install_pstipple_stage(struct draw_context *draw,
+                            struct pipe_context *pipe)
+{
+   struct pstip_stage *pstip;
+
+   /*
+    * Create / install AA line drawing / prim stage
+    */
+   pstip = draw_pstip_stage( draw );
+   assert(pstip);
+   draw->pipeline.pstipple = &pstip->stage;
+
+   pstip->pipe = pipe;
+
+   /* create special texture, sampler state */
+   pstip_create_texture(pstip);
+   pstip_create_sampler(pstip);
+
+   /* save original driver functions */
+   pstip->driver_create_fs_state = pipe->create_fs_state;
+   pstip->driver_bind_fs_state = pipe->bind_fs_state;
+   pstip->driver_delete_fs_state = pipe->delete_fs_state;
+
+   pstip->driver_bind_sampler_state = pipe->bind_sampler_state;
+   pstip->driver_set_sampler_texture = pipe->set_sampler_texture;
+   pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple;
+
+   /* override the driver's functions */
+   pipe->create_fs_state = pstip_create_fs_state;
+   pipe->bind_fs_state = pstip_bind_fs_state;
+   pipe->delete_fs_state = pstip_delete_fs_state;
+
+   pipe->bind_sampler_state = pstip_bind_sampler_state;
+   pipe->set_sampler_texture = pstip_set_sampler_texture;
+   pipe->set_polygon_stipple = pstip_set_polygon_stipple;
+
+   add_pstip_pipe_context(pipe, pstip);
+}
diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c
index 4375ebabbc..efd6793f2b 100644
--- a/src/gallium/auxiliary/draw/draw_validate.c
+++ b/src/gallium/auxiliary/draw/draw_validate.c
@@ -58,7 +58,18 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
     * shorter pipelines for lines & points.
     */
 
-   if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) ||
+   if (draw->rasterizer->line_smooth && draw->pipeline.aaline) {
+      draw->pipeline.aaline->next = next;
+      next = draw->pipeline.aaline;
+   }
+
+   if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) {
+      draw->pipeline.aapoint->next = next;
+      next = draw->pipeline.aapoint;
+   }
+
+   if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines
+        && !draw->rasterizer->line_smooth) ||
        (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) ||
        draw->rasterizer->point_sprite) {
       draw->pipeline.wide->next = next;
@@ -71,6 +82,12 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage )
       precalc_flat = 1;		/* only needed for lines really */
    }
 
+   if (draw->rasterizer->poly_stipple_enable
+       && draw->pipeline.pstipple) {
+      draw->pipeline.pstipple->next = next;
+      next = draw->pipeline.pstipple;
+   }
+
    if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL ||
        draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) {
       draw->pipeline.unfilled->next = next;
diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c
index 44427999cc..53f8bbec44 100644
--- a/src/gallium/auxiliary/draw/draw_vertex_cache.c
+++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c
@@ -41,7 +41,7 @@ void draw_vertex_cache_invalidate( struct draw_context *draw )
    assert(draw->vs.queue_nr == 0);
    assert(draw->vcache.referenced == 0);
 
-   memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx));
+//   memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx));
 }
 
 
@@ -62,43 +62,51 @@ static struct vertex_header *get_vertex( struct draw_context *draw,
    
    assert(slot < 32); /* so we don't exceed the bitfield size below */
 
-   /* Cache miss?
-    */
-   if (draw->vcache.idx[slot] != i) {
-
-      /* If slot is in use, use the overflow area:
+   if (draw->vcache.referenced & (1<<slot))
+   {
+      /* Cache hit?
        */
-      if (draw->vcache.referenced & (1 << slot)) {
-	 slot = VCACHE_SIZE + draw->vcache.overflow++;
+      if (draw->vcache.idx[slot].in == i) {
+//	 _mesa_printf("HIT %d %d\n", slot, i);
+	 assert(draw->vcache.idx[slot].out < draw->vs.queue_nr);
+	 return draw->vs.queue[draw->vcache.idx[slot].out].vertex;
       }
 
+      /* Otherwise a collision
+       */
+      slot = VCACHE_SIZE + draw->vcache.overflow++;
+//      _mesa_printf("XXX %d --> %d\n", i, slot);
+   }
+
+   /* Deal with the cache miss: 
+    */
+   {
+      unsigned out;
+      
       assert(slot < Elements(draw->vcache.idx));
 
-      draw->vcache.idx[slot] = i;
+//      _mesa_printf("NEW %d %d\n", slot, i);
+      draw->vcache.idx[slot].in = i;
+      draw->vcache.idx[slot].out = out = draw->vs.queue_nr++;
+      draw->vcache.referenced |= (1 << slot);
+
 
       /* Add to vertex shader queue:
        */
       assert(draw->vs.queue_nr < VS_QUEUE_LENGTH);
-      draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot];
-      draw->vs.queue[draw->vs.queue_nr].elt = i;
-      draw->vs.queue_nr++;
+
+      draw->vs.queue[out].elt = i;
+      draw->vs.queue[out].vertex->clipmask = 0;
+      draw->vs.queue[out].vertex->edgeflag = 1; /*XXX use user's edge flag! */
+      draw->vs.queue[out].vertex->pad = 0;
+      draw->vs.queue[out].vertex->vertex_id = UNDEFINED_VERTEX_ID;
 
       /* Need to set the vertex's edge flag here.  If we're being called
        * by do_ef_triangle(), that function needs edge flag info!
        */
-      draw->vcache.vertex[slot]->clipmask = 0;
-      draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */
-      draw->vcache.vertex[slot]->pad = 0;
-      draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID;
-   }
 
-
-   /* primitive flushing may have cleared the bitfield but did not
-    * clear the idx[] array values.  Set the bit now.  This fixes a
-    * bug found when drawing long triangle fans.
-    */
-   draw->vcache.referenced |= (1 << slot);
-   return draw->vcache.vertex[slot];
+      return draw->vs.queue[draw->vcache.idx[slot].out].vertex;
+   }
 }
 
 
@@ -130,8 +138,8 @@ void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw )
 {
    unsigned i;
 
-   for (i = 0; i < Elements(draw->vcache.vertex); i++)
-      draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID;
+   for (i = 0; i < draw->vs.post_nr; i++)
+      draw->vs.queue[i].vertex->vertex_id = UNDEFINED_VERTEX_ID;
 }
 
 
diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c
index f68f6e3244..5d2f5c9c43 100644
--- a/src/gallium/auxiliary/draw/draw_vertex_shader.c
+++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c
@@ -55,7 +55,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw)
     */
    shader->prepare( shader, draw );
 
-//   fprintf(stderr, " q(%d) ", draw->vs.queue_nr );
+//   fprintf(stderr, "%s %d\n", __FUNCTION__, draw->vs.queue_nr );
 
    /* run vertex shader on vertex cache entries, four per invokation */
    for (i = 0; i < draw->vs.queue_nr; i += 4) {
@@ -65,12 +65,12 @@ draw_vertex_shader_queue_flush(struct draw_context *draw)
 
       for (j = 0; j < n; j++) {
          elts[j] = draw->vs.queue[i + j].elt;
-         dests[j] = draw->vs.queue[i + j].dest;
+         dests[j] = draw->vs.queue[i + j].vertex;
       }
 
       for ( ; j < 4; j++) {
 	 elts[j] = elts[0];
-	 dests[j] = dests[0];
+         dests[j] = draw->vs.queue[i + j].vertex;
       }
 
       assert(n > 0);
@@ -79,6 +79,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw)
       shader->run(shader, draw, elts, n, dests);
    }
 
+   draw->vs.post_nr = draw->vs.queue_nr;
    draw->vs.queue_nr = 0;
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c
index dc3a5ecd21..901ff20a7e 100644
--- a/src/gallium/auxiliary/draw/draw_vf.c
+++ b/src/gallium/auxiliary/draw/draw_vf.c
@@ -30,6 +30,7 @@
 
 #include "pipe/p_compiler.h"
 #include "pipe/p_util.h"
+#include "rtasm/rtasm_execmem.h"
 
 #include "draw_vf.h"
 
@@ -37,11 +38,6 @@
 #define DRAW_VF_DBG 0
 
 
-/* TODO: remove this */
-extern void 
-_mesa_exec_free( void *addr );
-
-
 static boolean match_fastpath( struct draw_vertex_fetch *vf,
 				 const struct draw_vf_fastpath *fp)
 {
@@ -414,12 +410,12 @@ void draw_vf_destroy( struct draw_vertex_fetch *vf )
       FREE(fp->attr);
 
       /* KW: At the moment, fp->func is constrained to be allocated by
-       * _mesa_exec_alloc(), as the hardwired fastpaths in
+       * rtasm_exec_alloc(), as the hardwired fastpaths in
        * t_vertex_generic.c are handled specially.  It would be nice
        * to unify them, but this probably won't change until this
        * module gets another overhaul.
        */
-      //_mesa_exec_free((void *) fp->func);
+      //rtasm_exec_free((void *) fp->func);
       FREE(fp);
    }
    
diff --git a/src/gallium/auxiliary/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c
index 1ad2ae756d..aff4ffd985 100644
--- a/src/gallium/auxiliary/draw/draw_vf_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vf_sse.c
@@ -26,17 +26,16 @@
  */
 
 
-#include "simple_list.h"
-
 #include "pipe/p_compiler.h"
+#include "util/u_simple_list.h"
 
 #include "draw_vf.h"
 
 
 #if defined(USE_SSE_ASM)
 
-#include "x86/rtasm/x86sse.h"
-#include "x86/common_x86_asm.h"
+#include "rtasm/rtasm_cpu.h"
+#include "rtasm/rtasm_x86sse.h"
 
 
 #define X    0
@@ -576,7 +575,7 @@ void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf )
 {
    struct x86_program p;   
 
-   if (!cpu_has_xmm) {
+   if (!rtasm_cpu_has_sse()) {
       vf->codegen_emit = NULL;
       return;
    }
@@ -586,7 +585,7 @@ void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf )
    p.vf = vf;
    p.inputs_safe = 0;		/* for now */
    p.outputs_safe = 1;		/* for now */
-   p.have_sse2 = cpu_has_xmm2;
+   p.have_sse2 = rtasm_cpu_has_sse2();
    p.identity = x86_make_reg(file_XMM, 6);
    p.chan0 = x86_make_reg(file_XMM, 7);
 
diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c
index 44022b6e07..0fd557d667 100644
--- a/src/gallium/auxiliary/draw/draw_vs_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c
@@ -40,7 +40,7 @@
 
 #ifdef MESA_LLVM
 
-#include "llvm/gallivm.h"
+#include "gallivm/gallivm.h"
 
 struct draw_llvm_vertex_shader {
    struct draw_vertex_shader base;
diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c
index 27bc66812c..11ef0c503d 100644
--- a/src/gallium/auxiliary/draw/draw_vs_sse.c
+++ b/src/gallium/auxiliary/draw/draw_vs_sse.c
@@ -41,7 +41,7 @@
 #include "draw_private.h"
 #include "draw_context.h"
 
-#include "x86/rtasm/x86sse.h"
+#include "rtasm/rtasm_x86sse.h"
 #include "tgsi/exec/tgsi_sse2.h"
 
 
diff --git a/src/gallium/auxiliary/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c
index 655774b155..1f8069bdca 100644
--- a/src/gallium/auxiliary/draw/draw_wide_prims.c
+++ b/src/gallium/auxiliary/draw/draw_wide_prims.c
@@ -162,70 +162,6 @@ static void wide_line( struct draw_stage *stage,
 
 
 /**
- * Draw a wide line by drawing a quad, using geometry which will
- * fullfill GL's antialiased line requirements.
- */
-static void wide_line_aa(struct draw_stage *stage,
-                         struct prim_header *header)
-{
-   const struct wide_stage *wide = wide_stage(stage);
-   const float half_width = wide->half_line_width;
-   struct prim_header tri;
-   struct vertex_header *v[4];
-   float *pos;
-   float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0];
-   float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1];
-   const float len = (float) sqrt(dx * dx + dy * dy);
-   uint i;
-
-   dx = dx * half_width / len;
-   dy = dy * half_width / len;
-
-   /* allocate/dup new verts */
-   for (i = 0; i < 4; i++) {
-      v[i] = dup_vert(stage, header->v[i/2], i);
-   }
-
-   /*
-    * Quad for line from v0 to v1:
-    *
-    *  1                         3
-    *  +-------------------------+
-    *  |                         |
-    *  *v0                     v1*
-    *  |                         |
-    *  +-------------------------+
-    *  0                         2
-    */
-
-   pos = v[0]->data[0];
-   pos[0] += dy;
-   pos[1] -= dx;
-
-   pos = v[1]->data[0];
-   pos[0] -= dy;
-   pos[1] += dx;
-
-   pos = v[2]->data[0];
-   pos[0] += dy;
-   pos[1] -= dx;
-
-   pos = v[3]->data[0];
-   pos[0] -= dy;
-   pos[1] += dx;
-
-   tri.det = header->det;  /* only the sign matters */
-
-   tri.v[0] = v[2];  tri.v[1] = v[1];  tri.v[2] = v[0];
-   stage->next->tri( stage->next, &tri );
-
-   tri.v[0] = v[3];  tri.v[1] = v[1];  tri.v[2] = v[2];
-   stage->next->tri( stage->next, &tri );
-
-}
-
-
-/**
  * Set the vertex texcoords for sprite mode.
  * Coords may be left untouched or set to a right-side-up or upside-down
  * orientation.
@@ -379,10 +315,7 @@ static void wide_first_line( struct draw_stage *stage,
    wide->half_line_width = 0.5f * draw->rasterizer->line_width;
 
    if (draw->rasterizer->line_width != 1.0) {
-      if (draw->rasterizer->line_smooth)
-         wide->stage.line = wide_line_aa;
-      else
-         wide->stage.line = wide_line;
+      wide->stage.line = wide_line;
    }
    else {
       wide->stage.line = passthrough_line;