mesa: move fixed function vertex program builder from tnl to core mesa
authorKeith Whitwell <keith@tungstengraphics.com>
Sat, 20 Sep 2008 15:26:11 +0000 (08:26 -0700)
committerKeith Whitwell <keith@tungstengraphics.com>
Sun, 21 Sep 2008 16:46:00 +0000 (09:46 -0700)
Also unify caching of fragment and vertex programs in shader/prog_cache.c`

Brought across from gallium-0.2

17 files changed:
src/mesa/main/ffvertex_prog.c [new file with mode: 0644]
src/mesa/main/ffvertex_prog.h [new file with mode: 0644]
src/mesa/main/mtypes.h
src/mesa/main/state.c
src/mesa/main/texenvprogram.c
src/mesa/main/texenvprogram.h
src/mesa/main/texstate.c
src/mesa/shader/prog_cache.c [new file with mode: 0644]
src/mesa/shader/prog_cache.h [new file with mode: 0644]
src/mesa/shader/prog_statevars.c
src/mesa/shader/prog_statevars.h
src/mesa/shader/program.c
src/mesa/sources
src/mesa/tnl/t_context.c
src/mesa/tnl/t_context.h
src/mesa/tnl/t_vp_build.c
src/mesa/tnl/t_vp_build.h

diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c
new file mode 100644 (file)
index 0000000..787672b
--- /dev/null
@@ -0,0 +1,1753 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+/**
+ * \file ffvertex_prog.
+ *
+ * Create a vertex program to execute the current fixed function T&L pipeline.
+ * \author Keith Whitwell
+ */
+
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/enums.h"
+#include "main/ffvertex_prog.h"
+#include "shader/program.h"
+#include "shader/prog_cache.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+#include "shader/prog_statevars.h"
+
+
+struct state_key {
+   unsigned light_global_enabled:1;
+   unsigned light_local_viewer:1;
+   unsigned light_twoside:1;
+   unsigned light_color_material:1;
+   unsigned light_color_material_mask:12;
+   unsigned light_material_mask:12;
+   unsigned material_shininess_is_zero:1;
+
+   unsigned need_eye_coords:1;
+   unsigned normalize:1;
+   unsigned rescale_normals:1;
+   unsigned fog_source_is_depth:1;
+   unsigned tnl_do_vertex_fog:1;
+   unsigned separate_specular:1;
+   unsigned fog_mode:2;
+   unsigned point_attenuated:1;
+   unsigned point_array:1;
+   unsigned texture_enabled_global:1;
+   unsigned fragprog_inputs_read:12;
+
+   struct {
+      unsigned light_enabled:1;
+      unsigned light_eyepos3_is_zero:1;
+      unsigned light_spotcutoff_is_180:1;
+      unsigned light_attenuated:1;      
+      unsigned texunit_really_enabled:1;
+      unsigned texmat_enabled:1;
+      unsigned texgen_enabled:4;
+      unsigned texgen_mode0:4;
+      unsigned texgen_mode1:4;
+      unsigned texgen_mode2:4;
+      unsigned texgen_mode3:4;
+   } unit[8];
+};
+
+
+
+#define FOG_NONE   0
+#define FOG_LINEAR 1
+#define FOG_EXP    2
+#define FOG_EXP2   3
+
+static GLuint translate_fog_mode( GLenum mode )
+{
+   switch (mode) {
+   case GL_LINEAR: return FOG_LINEAR;
+   case GL_EXP: return FOG_EXP;
+   case GL_EXP2: return FOG_EXP2;
+   default: return FOG_NONE;
+   }
+}
+
+#define TXG_NONE           0
+#define TXG_OBJ_LINEAR     1
+#define TXG_EYE_LINEAR     2
+#define TXG_SPHERE_MAP     3
+#define TXG_REFLECTION_MAP 4
+#define TXG_NORMAL_MAP     5
+
+static GLuint translate_texgen( GLboolean enabled, GLenum mode )
+{
+   if (!enabled)
+      return TXG_NONE;
+
+   switch (mode) {
+   case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR;
+   case GL_EYE_LINEAR: return TXG_EYE_LINEAR;
+   case GL_SPHERE_MAP: return TXG_SPHERE_MAP;
+   case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP;
+   case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP;
+   default: return TXG_NONE;
+   }
+}
+
+
+/**
+ * Returns bitmask of flags indicating which materials are set per-vertex
+ * in the current VB.
+ * XXX get these from the VBO...
+ */
+static GLbitfield
+tnl_get_per_vertex_materials(GLcontext *ctx)
+{
+   GLbitfield mask = 0x0;
+#if 0
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+
+   for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
+      if (VB->AttribPtr[i] && VB->AttribPtr[i]->stride) 
+         mask |= 1 << (i - _TNL_FIRST_MAT);
+#endif
+   return mask;
+}
+
+/**
+ * Should fog be computed per-vertex?
+ */
+static GLboolean
+tnl_get_per_vertex_fog(GLcontext *ctx)
+{
+#if 0
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   return tnl->_DoVertexFog;
+#else
+   return GL_FALSE;
+#endif
+}
+
+static GLboolean check_active_shininess( GLcontext *ctx,
+                                         const struct state_key *key,
+                                         GLuint side )
+{
+   GLuint bit = 1 << (MAT_ATTRIB_FRONT_SHININESS + side);
+
+   if (key->light_color_material_mask & bit)
+      return GL_TRUE;
+
+   if (key->light_material_mask & bit)
+      return GL_TRUE;
+
+   if (ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SHININESS + side][0] != 0.0F)
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
+     
+
+
+
+static struct state_key *make_state_key( GLcontext *ctx )
+{
+   const struct gl_fragment_program *fp;
+   struct state_key *key = CALLOC_STRUCT(state_key);
+   GLuint i;
+
+   fp = ctx->FragmentProgram._Current;
+
+   /* This now relies on texenvprogram.c being active:
+    */
+   assert(fp);
+
+   key->need_eye_coords = ctx->_NeedEyeCoords;
+
+   key->fragprog_inputs_read = fp->Base.InputsRead;
+
+   if (ctx->RenderMode == GL_FEEDBACK) {
+      /* make sure the vertprog emits color and tex0 */
+      key->fragprog_inputs_read |= (FRAG_BIT_COL0 | FRAG_BIT_TEX0);
+   }
+
+   key->separate_specular = (ctx->Light.Model.ColorControl ==
+                            GL_SEPARATE_SPECULAR_COLOR);
+
+   if (ctx->Light.Enabled) {
+      key->light_global_enabled = 1;
+
+      if (ctx->Light.Model.LocalViewer)
+        key->light_local_viewer = 1;
+
+      if (ctx->Light.Model.TwoSide)
+        key->light_twoside = 1;
+
+      if (ctx->Light.ColorMaterialEnabled) {
+        key->light_color_material = 1;
+        key->light_color_material_mask = ctx->Light.ColorMaterialBitmask;
+      }
+
+      key->light_material_mask = tnl_get_per_vertex_materials(ctx);
+
+      for (i = 0; i < MAX_LIGHTS; i++) {
+        struct gl_light *light = &ctx->Light.Light[i];
+
+        if (light->Enabled) {
+           key->unit[i].light_enabled = 1;
+
+           if (light->EyePosition[3] == 0.0)
+              key->unit[i].light_eyepos3_is_zero = 1;
+           
+           if (light->SpotCutoff == 180.0)
+              key->unit[i].light_spotcutoff_is_180 = 1;
+
+           if (light->ConstantAttenuation != 1.0 ||
+               light->LinearAttenuation != 0.0 ||
+               light->QuadraticAttenuation != 0.0)
+              key->unit[i].light_attenuated = 1;
+        }
+      }
+
+      if (check_active_shininess(ctx, key, 0)) {
+         key->material_shininess_is_zero = 0;
+      }
+      else if (key->light_twoside &&
+               check_active_shininess(ctx, key, 1)) {
+         key->material_shininess_is_zero = 0;
+      }
+      else {
+         key->material_shininess_is_zero = 1;
+      }
+   }
+
+   if (ctx->Transform.Normalize)
+      key->normalize = 1;
+
+   if (ctx->Transform.RescaleNormals)
+      key->rescale_normals = 1;
+
+   key->fog_mode = translate_fog_mode(fp->FogOption);
+   
+   if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT)
+      key->fog_source_is_depth = 1;
+   
+   key->tnl_do_vertex_fog = tnl_get_per_vertex_fog(ctx);
+
+   if (ctx->Point._Attenuated)
+      key->point_attenuated = 1;
+
+#if FEATURE_point_size_array
+   if (ctx->Array.ArrayObj->PointSize.Enabled)
+      key->point_array = 1;
+#endif
+
+   if (ctx->Texture._TexGenEnabled ||
+       ctx->Texture._TexMatEnabled ||
+       ctx->Texture._EnabledUnits)
+      key->texture_enabled_global = 1;
+      
+   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+
+      if (texUnit->_ReallyEnabled)
+        key->unit[i].texunit_really_enabled = 1;
+
+      if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i))      
+        key->unit[i].texmat_enabled = 1;
+      
+      if (texUnit->TexGenEnabled) {
+        key->unit[i].texgen_enabled = 1;
+      
+        key->unit[i].texgen_mode0 = 
+           translate_texgen( texUnit->TexGenEnabled & (1<<0),
+                             texUnit->GenModeS );
+        key->unit[i].texgen_mode1 = 
+           translate_texgen( texUnit->TexGenEnabled & (1<<1),
+                             texUnit->GenModeT );
+        key->unit[i].texgen_mode2 = 
+           translate_texgen( texUnit->TexGenEnabled & (1<<2),
+                             texUnit->GenModeR );
+        key->unit[i].texgen_mode3 = 
+           translate_texgen( texUnit->TexGenEnabled & (1<<3),
+                             texUnit->GenModeQ );
+      }
+   }
+   
+   return key;
+}
+
+
+   
+/* Very useful debugging tool - produces annotated listing of
+ * generated program with line/function references for each
+ * instruction back into this file:
+ */
+#define DISASSEM 0
+
+/* Should be tunable by the driver - do we want to do matrix
+ * multiplications with DP4's or with MUL/MAD's?  SSE works better
+ * with the latter, drivers may differ.
+ */
+#define PREFER_DP4 0
+
+
+/* Use uregs to represent registers internally, translate to Mesa's
+ * expected formats on emit.  
+ *
+ * NOTE: These are passed by value extensively in this file rather
+ * than as usual by pointer reference.  If this disturbs you, try
+ * remembering they are just 32bits in size.
+ *
+ * GCC is smart enough to deal with these dword-sized structures in
+ * much the same way as if I had defined them as dwords and was using
+ * macros to access and set the fields.  This is much nicer and easier
+ * to evolve.
+ */
+struct ureg {
+   GLuint file:4;
+   GLint idx:9;      /* relative addressing may be negative */
+                     /* sizeof(idx) should == sizeof(prog_src_reg::Index) */
+   GLuint negate:1;
+   GLuint swz:12;
+   GLuint pad:6;
+};
+
+
+struct tnl_program {
+   const struct state_key *state;
+   struct gl_vertex_program *program;
+   GLint max_inst;  /** number of instructions allocated for program */
+   
+   GLuint temp_in_use;
+   GLuint temp_reserved;
+   
+   struct ureg eye_position;
+   struct ureg eye_position_z;
+   struct ureg eye_position_normalized;
+   struct ureg transformed_normal;
+   struct ureg identity;
+
+   GLuint materials;
+   GLuint color_materials;
+};
+
+
+static const struct ureg undef = { 
+   PROGRAM_UNDEFINED,
+   0,
+   0,
+   0,
+   0
+};
+
+/* Local shorthand:
+ */
+#define X    SWIZZLE_X
+#define Y    SWIZZLE_Y
+#define Z    SWIZZLE_Z
+#define W    SWIZZLE_W
+
+
+/* Construct a ureg:
+ */
+static struct ureg make_ureg(GLuint file, GLint idx)
+{
+   struct ureg reg;
+   reg.file = file;
+   reg.idx = idx;
+   reg.negate = 0;
+   reg.swz = SWIZZLE_NOOP;
+   reg.pad = 0;
+   return reg;
+}
+
+
+
+static struct ureg negate( struct ureg reg )
+{
+   reg.negate ^= 1;
+   return reg;
+} 
+
+
+static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
+{
+   reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x),
+                          GET_SWZ(reg.swz, y),
+                          GET_SWZ(reg.swz, z),
+                          GET_SWZ(reg.swz, w));
+
+   return reg;
+}
+
+static struct ureg swizzle1( struct ureg reg, int x )
+{
+   return swizzle(reg, x, x, x, x);
+}
+
+static struct ureg get_temp( struct tnl_program *p )
+{
+   int bit = _mesa_ffs( ~p->temp_in_use );
+   if (!bit) {
+      _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__);
+      _mesa_exit(1);
+   }
+
+   if ((GLuint) bit > p->program->Base.NumTemporaries)
+      p->program->Base.NumTemporaries = bit;
+
+   p->temp_in_use |= 1<<(bit-1);
+   return make_ureg(PROGRAM_TEMPORARY, bit-1);
+}
+
+static struct ureg reserve_temp( struct tnl_program *p )
+{
+   struct ureg temp = get_temp( p );
+   p->temp_reserved |= 1<<temp.idx;
+   return temp;
+}
+
+static void release_temp( struct tnl_program *p, struct ureg reg )
+{
+   if (reg.file == PROGRAM_TEMPORARY) {
+      p->temp_in_use &= ~(1<<reg.idx);
+      p->temp_in_use |= p->temp_reserved; /* can't release reserved temps */
+   }
+}
+
+static void release_temps( struct tnl_program *p )
+{
+   p->temp_in_use = p->temp_reserved;
+}
+
+
+
+/**
+ * \param input  one of VERT_ATTRIB_x tokens.
+ */
+static struct ureg register_input( struct tnl_program *p, GLuint input )
+{
+   p->program->Base.InputsRead |= (1<<input);
+   return make_ureg(PROGRAM_INPUT, input);
+}
+
+/**
+ * \param input  one of VERT_RESULT_x tokens.
+ */
+static struct ureg register_output( struct tnl_program *p, GLuint output )
+{
+   p->program->Base.OutputsWritten |= (1<<output);
+   return make_ureg(PROGRAM_OUTPUT, output);
+}
+
+static struct ureg register_const4f( struct tnl_program *p, 
+                             GLfloat s0,
+                             GLfloat s1,
+                             GLfloat s2,
+                             GLfloat s3)
+{
+   GLfloat values[4];
+   GLint idx;
+   GLuint swizzle;
+   values[0] = s0;
+   values[1] = s1;
+   values[2] = s2;
+   values[3] = s3;
+   idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
+                                     &swizzle );
+   ASSERT(swizzle == SWIZZLE_NOOP);
+   return make_ureg(PROGRAM_CONSTANT, idx);
+}
+
+#define register_const1f(p, s0)         register_const4f(p, s0, 0, 0, 1)
+#define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
+#define register_const2f(p, s0, s1)     register_const4f(p, s0, s1, 0, 1)
+#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
+
+static GLboolean is_undef( struct ureg reg )
+{
+   return reg.file == PROGRAM_UNDEFINED;
+}
+
+static struct ureg get_identity_param( struct tnl_program *p )
+{
+   if (is_undef(p->identity)) 
+      p->identity = register_const4f(p, 0,0,0,1);
+
+   return p->identity;
+}
+
+static struct ureg register_param5(struct tnl_program *p, 
+                                  GLint s0,
+                                  GLint s1,
+                                  GLint s2,
+                                  GLint s3,
+                                   GLint s4)
+{
+   gl_state_index tokens[STATE_LENGTH];
+   GLint idx;
+   tokens[0] = s0;
+   tokens[1] = s1;
+   tokens[2] = s2;
+   tokens[3] = s3;
+   tokens[4] = s4;
+   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
+   return make_ureg(PROGRAM_STATE_VAR, idx);
+}
+
+
+#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
+#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
+#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
+#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
+
+
+static void register_matrix_param5( struct tnl_program *p,
+                                   GLint s0, /* modelview, projection, etc */
+                                   GLint s1, /* texture matrix number */
+                                   GLint s2, /* first row */
+                                   GLint s3, /* last row */
+                                   GLint s4, /* inverse, transpose, etc */
+                                   struct ureg *matrix )
+{
+   GLint i;
+
+   /* This is a bit sad as the support is there to pull the whole
+    * matrix out in one go:
+    */
+   for (i = 0; i <= s3 - s2; i++) 
+      matrix[i] = register_param5( p, s0, s1, i, i, s4 );
+}
+
+
+static void emit_arg( struct prog_src_register *src,
+                     struct ureg reg )
+{
+   src->File = reg.file;
+   src->Index = reg.idx;
+   src->Swizzle = reg.swz;
+   src->NegateBase = reg.negate ? NEGATE_XYZW : 0;
+   src->Abs = 0;
+   src->NegateAbs = 0;
+   src->RelAddr = 0;
+   /* Check that bitfield sizes aren't exceeded */
+   ASSERT(src->Index == reg.idx);
+}
+
+static void emit_dst( struct prog_dst_register *dst,
+                     struct ureg reg, GLuint mask )
+{
+   dst->File = reg.file;
+   dst->Index = reg.idx;
+   /* allow zero as a shorthand for xyzw */
+   dst->WriteMask = mask ? mask : WRITEMASK_XYZW; 
+   dst->CondMask = COND_TR;  /* always pass cond test */
+   dst->CondSwizzle = SWIZZLE_NOOP;
+   dst->CondSrc = 0;
+   dst->pad = 0;
+   /* Check that bitfield sizes aren't exceeded */
+   ASSERT(dst->Index == reg.idx);
+}
+
+static void debug_insn( struct prog_instruction *inst, const char *fn,
+                       GLuint line )
+{
+   if (DISASSEM) {
+      static const char *last_fn;
+   
+      if (fn != last_fn) {
+        last_fn = fn;
+        _mesa_printf("%s:\n", fn);
+      }
+        
+      _mesa_printf("%d:\t", line);
+      _mesa_print_instruction(inst);
+   }
+}
+
+
+static void emit_op3fn(struct tnl_program *p,
+                       enum prog_opcode op,
+                      struct ureg dest,
+                      GLuint mask,
+                      struct ureg src0,
+                      struct ureg src1,
+                      struct ureg src2,
+                      const char *fn,
+                      GLuint line)
+{
+   GLuint nr;
+   struct prog_instruction *inst;
+      
+   assert((GLint) p->program->Base.NumInstructions <= p->max_inst);
+
+   if (p->program->Base.NumInstructions == p->max_inst) {
+      /* need to extend the program's instruction array */
+      struct prog_instruction *newInst;
+
+      /* double the size */
+      p->max_inst *= 2;
+
+      newInst = _mesa_alloc_instructions(p->max_inst);
+      if (!newInst) {
+         _mesa_error(NULL, GL_OUT_OF_MEMORY, "vertex program build");
+         return;
+      }
+
+      _mesa_copy_instructions(newInst,
+                              p->program->Base.Instructions,
+                              p->program->Base.NumInstructions);
+
+      _mesa_free_instructions(p->program->Base.Instructions,
+                              p->program->Base.NumInstructions);
+
+      p->program->Base.Instructions = newInst;
+   }
+      
+   nr = p->program->Base.NumInstructions++;
+
+   inst = &p->program->Base.Instructions[nr];
+   inst->Opcode = (enum prog_opcode) op; 
+   inst->StringPos = 0;
+   inst->Data = 0;
+   
+   emit_arg( &inst->SrcReg[0], src0 );
+   emit_arg( &inst->SrcReg[1], src1 );
+   emit_arg( &inst->SrcReg[2], src2 );   
+
+   emit_dst( &inst->DstReg, dest, mask );
+
+   debug_insn(inst, fn, line);
+}
+
+
+#define emit_op3(p, op, dst, mask, src0, src1, src2) \
+   emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
+
+#define emit_op2(p, op, dst, mask, src0, src1) \
+    emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
+
+#define emit_op1(p, op, dst, mask, src0) \
+    emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
+
+
+static struct ureg make_temp( struct tnl_program *p, struct ureg reg )
+{
+   if (reg.file == PROGRAM_TEMPORARY && 
+       !(p->temp_reserved & (1<<reg.idx)))
+      return reg;
+   else {
+      struct ureg temp = get_temp(p);
+      emit_op1(p, OPCODE_MOV, temp, 0, reg);
+      return temp;
+   }
+}
+
+
+/* Currently no tracking performed of input/output/register size or
+ * active elements.  Could be used to reduce these operations, as
+ * could the matrix type.
+ */
+static void emit_matrix_transform_vec4( struct tnl_program *p,
+                                       struct ureg dest,
+                                       const struct ureg *mat,
+                                       struct ureg src)
+{
+   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_X, src, mat[0]);
+   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Y, src, mat[1]);
+   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Z, src, mat[2]);
+   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_W, src, mat[3]);
+}
+
+/* This version is much easier to implement if writemasks are not
+ * supported natively on the target or (like SSE), the target doesn't
+ * have a clean/obvious dotproduct implementation.
+ */
+static void emit_transpose_matrix_transform_vec4( struct tnl_program *p,
+                                                 struct ureg dest,
+                                                 const struct ureg *mat,
+                                                 struct ureg src)
+{
+   struct ureg tmp;
+
+   if (dest.file != PROGRAM_TEMPORARY)
+      tmp = get_temp(p);
+   else
+      tmp = dest;
+
+   emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]);
+   emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp);
+   emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp);
+   emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp);
+
+   if (dest.file != PROGRAM_TEMPORARY)
+      release_temp(p, tmp);
+}
+
+static void emit_matrix_transform_vec3( struct tnl_program *p,
+                                       struct ureg dest,
+                                       const struct ureg *mat,
+                                       struct ureg src)
+{
+   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_X, src, mat[0]);
+   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Y, src, mat[1]);
+   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Z, src, mat[2]);
+}
+
+
+static void emit_normalize_vec3( struct tnl_program *p,
+                                struct ureg dest,
+                                struct ureg src )
+{
+   struct ureg tmp = get_temp(p);
+   emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src);
+   emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp);
+   emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X));
+   release_temp(p, tmp);
+}
+
+static void emit_passthrough( struct tnl_program *p, 
+                             GLuint input,
+                             GLuint output )
+{
+   struct ureg out = register_output(p, output);
+   emit_op1(p, OPCODE_MOV, out, 0, register_input(p, input)); 
+}
+
+static struct ureg get_eye_position( struct tnl_program *p )
+{
+   if (is_undef(p->eye_position)) {
+      struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
+      struct ureg modelview[4];
+
+      p->eye_position = reserve_temp(p);
+
+      if (PREFER_DP4) {
+        register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
+                                 0, modelview );
+
+        emit_matrix_transform_vec4(p, p->eye_position, modelview, pos);
+      }
+      else {
+        register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
+                                STATE_MATRIX_TRANSPOSE, modelview );
+
+        emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos);
+      }
+   }
+   
+   return p->eye_position;
+}
+
+
+static struct ureg get_eye_position_z( struct tnl_program *p )
+{
+   if (!is_undef(p->eye_position)) 
+      return swizzle1(p->eye_position, Z);
+
+   if (is_undef(p->eye_position_z)) {
+      struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
+      struct ureg modelview[4];
+
+      p->eye_position_z = reserve_temp(p);
+
+      register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
+                              0, modelview );
+
+      emit_op2(p, OPCODE_DP4, p->eye_position_z, 0, pos, modelview[2]);
+   }
+   
+   return p->eye_position_z;
+}
+   
+
+
+static struct ureg get_eye_position_normalized( struct tnl_program *p )
+{
+   if (is_undef(p->eye_position_normalized)) {
+      struct ureg eye = get_eye_position(p);
+      p->eye_position_normalized = reserve_temp(p);
+      emit_normalize_vec3(p, p->eye_position_normalized, eye);
+   }
+   
+   return p->eye_position_normalized;
+}
+
+
+static struct ureg get_transformed_normal( struct tnl_program *p )
+{
+   if (is_undef(p->transformed_normal) &&
+       !p->state->need_eye_coords &&
+       !p->state->normalize &&
+       !(p->state->need_eye_coords == p->state->rescale_normals))
+   {
+      p->transformed_normal = register_input(p, VERT_ATTRIB_NORMAL );
+   }
+   else if (is_undef(p->transformed_normal)) 
+   {
+      struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL );
+      struct ureg mvinv[3];
+      struct ureg transformed_normal = reserve_temp(p);
+
+      if (p->state->need_eye_coords) {
+         register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2,
+                                 STATE_MATRIX_INVTRANS, mvinv );
+
+         /* Transform to eye space:
+          */
+         emit_matrix_transform_vec3( p, transformed_normal, mvinv, normal );
+         normal = transformed_normal;
+      }
+
+      /* Normalize/Rescale:
+       */
+      if (p->state->normalize) {
+        emit_normalize_vec3( p, transformed_normal, normal );
+         normal = transformed_normal;
+      }
+      else if (p->state->need_eye_coords == p->state->rescale_normals) {
+         /* This is already adjusted for eye/non-eye rendering:
+          */
+        struct ureg rescale = register_param2(p, STATE_INTERNAL,
+                                               STATE_NORMAL_SCALE);
+
+        emit_op2( p, OPCODE_MUL, transformed_normal, 0, normal, rescale );
+         normal = transformed_normal;
+      }
+      
+      assert(normal.file == PROGRAM_TEMPORARY);
+      p->transformed_normal = normal;
+   }
+
+   return p->transformed_normal;
+}
+
+
+
+static void build_hpos( struct tnl_program *p )
+{
+   struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
+   struct ureg hpos = register_output( p, VERT_RESULT_HPOS );
+   struct ureg mvp[4];
+
+   if (PREFER_DP4) {
+      register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
+                             0, mvp );
+      emit_matrix_transform_vec4( p, hpos, mvp, pos );
+   }
+   else {
+      register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
+                             STATE_MATRIX_TRANSPOSE, mvp );
+      emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos );
+   }
+}
+
+
+static GLuint material_attrib( GLuint side, GLuint property )
+{
+   return ((property - STATE_AMBIENT) * 2 + 
+          side);
+}
+
+/* Get a bitmask of which material values vary on a per-vertex basis.
+ */
+static void set_material_flags( struct tnl_program *p )
+{
+   p->color_materials = 0;
+   p->materials = 0;
+
+   if (p->state->light_color_material) {
+      p->materials = 
+        p->color_materials = p->state->light_color_material_mask;
+   }
+
+   p->materials |= p->state->light_material_mask;
+}
+
+
+/* XXX temporary!!! */
+#define _TNL_ATTRIB_MAT_FRONT_AMBIENT 32
+
+static struct ureg get_material( struct tnl_program *p, GLuint side, 
+                                GLuint property )
+{
+   GLuint attrib = material_attrib(side, property);
+
+   if (p->color_materials & (1<<attrib))
+      return register_input(p, VERT_ATTRIB_COLOR0);
+   else if (p->materials & (1<<attrib)) 
+      return register_input( p, attrib + _TNL_ATTRIB_MAT_FRONT_AMBIENT );
+   else
+      return register_param3( p, STATE_MATERIAL, side, property );
+}
+
+#define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
+                                  MAT_BIT_FRONT_AMBIENT | \
+                                  MAT_BIT_FRONT_DIFFUSE) << (side))
+
+/* Either return a precalculated constant value or emit code to
+ * calculate these values dynamically in the case where material calls
+ * are present between begin/end pairs.
+ *
+ * Probably want to shift this to the program compilation phase - if
+ * we always emitted the calculation here, a smart compiler could
+ * detect that it was constant (given a certain set of inputs), and
+ * lift it out of the main loop.  That way the programs created here
+ * would be independent of the vertex_buffer details.
+ */
+static struct ureg get_scenecolor( struct tnl_program *p, GLuint side )
+{
+   if (p->materials & SCENE_COLOR_BITS(side)) {
+      struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT);
+      struct ureg material_emission = get_material(p, side, STATE_EMISSION);
+      struct ureg material_ambient = get_material(p, side, STATE_AMBIENT);
+      struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE);
+      struct ureg tmp = make_temp(p, material_diffuse);
+      emit_op3(p, OPCODE_MAD, tmp,  WRITEMASK_XYZ, lm_ambient, 
+              material_ambient, material_emission);
+      return tmp;
+   }
+   else
+      return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side );
+}
+
+
+static struct ureg get_lightprod( struct tnl_program *p, GLuint light, 
+                                 GLuint side, GLuint property )
+{
+   GLuint attrib = material_attrib(side, property);
+   if (p->materials & (1<<attrib)) {
+      struct ureg light_value = 
+        register_param3(p, STATE_LIGHT, light, property);
+      struct ureg material_value = get_material(p, side, property);
+      struct ureg tmp = get_temp(p);
+      emit_op2(p, OPCODE_MUL, tmp,  0, light_value, material_value);
+      return tmp;
+   }
+   else
+      return register_param4(p, STATE_LIGHTPROD, light, side, property);
+}
+
+static struct ureg calculate_light_attenuation( struct tnl_program *p,
+                                               GLuint i, 
+                                               struct ureg VPpli,
+                                               struct ureg dist )
+{
+   struct ureg attenuation = register_param3(p, STATE_LIGHT, i,
+                                            STATE_ATTENUATION);
+   struct ureg att = get_temp(p);
+
+   /* Calculate spot attenuation:
+    */
+   if (!p->state->unit[i].light_spotcutoff_is_180) {
+      struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
+                                                 STATE_LIGHT_SPOT_DIR_NORMALIZED, i);
+      struct ureg spot = get_temp(p);
+      struct ureg slt = get_temp(p);
+
+      emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm);
+      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
+      emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
+      emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
+
+      release_temp(p, spot);
+      release_temp(p, slt);
+   }
+
+   /* Calculate distance attenuation:
+    */
+   if (p->state->unit[i].light_attenuated) {
+
+      /* 1/d,d,d,1/d */
+      emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist); 
+      /* 1,d,d*d,1/d */
+      emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y)); 
+      /* 1/dist-atten */
+      emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist); 
+
+      if (!p->state->unit[i].light_spotcutoff_is_180) {
+        /* dist-atten */
+        emit_op1(p, OPCODE_RCP, dist, 0, dist); 
+        /* spot-atten * dist-atten */
+        emit_op2(p, OPCODE_MUL, att, 0, dist, att);    
+      } else {
+        /* dist-atten */
+        emit_op1(p, OPCODE_RCP, att, 0, dist); 
+      }
+   }
+
+   return att;
+}
+                                               
+
+/**
+ * Compute:
+ *   lit.y = MAX(0, dots.x)
+ *   lit.z = SLT(0, dots.x)
+ */
+static void emit_degenerate_lit( struct tnl_program *p,
+                                 struct ureg lit,
+                                 struct ureg dots )
+{
+   struct ureg id = get_identity_param(p);  /* id = {0,0,0,1} */
+
+   /* Note that lit.x & lit.w will not be examined.  Note also that
+    * dots.xyzw == dots.xxxx.
+    */
+
+   /* MAX lit, id, dots;
+    */
+   emit_op2(p, OPCODE_MAX, lit, WRITEMASK_XYZW, id, dots); 
+
+   /* result[2] = (in > 0 ? 1 : 0)
+    * SLT lit.z, id.z, dots;   # lit.z = (0 < dots.z) ? 1 : 0
+    */
+   emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, swizzle1(id,Z), dots);
+}
+
+
+/* Need to add some addtional parameters to allow lighting in object
+ * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
+ * space lighting.
+ */
+static void build_lighting( struct tnl_program *p )
+{
+   const GLboolean twoside = p->state->light_twoside;
+   const GLboolean separate = p->state->separate_specular;
+   GLuint nr_lights = 0, count = 0;
+   struct ureg normal = get_transformed_normal(p);
+   struct ureg lit = get_temp(p);
+   struct ureg dots = get_temp(p);
+   struct ureg _col0 = undef, _col1 = undef;
+   struct ureg _bfc0 = undef, _bfc1 = undef;
+   GLuint i;
+
+   /*
+    * NOTE:
+    * dot.x = dot(normal, VPpli)
+    * dot.y = dot(normal, halfAngle)
+    * dot.z = back.shininess
+    * dot.w = front.shininess
+    */
+
+   for (i = 0; i < MAX_LIGHTS; i++) 
+      if (p->state->unit[i].light_enabled)
+        nr_lights++;
+   
+   set_material_flags(p);
+
+   {
+      if (!p->state->material_shininess_is_zero) {
+         struct ureg shininess = get_material(p, 0, STATE_SHININESS);
+         emit_op1(p, OPCODE_MOV, dots,  WRITEMASK_W, swizzle1(shininess,X));
+         release_temp(p, shininess);
+      }
+
+      _col0 = make_temp(p, get_scenecolor(p, 0));
+      if (separate)
+        _col1 = make_temp(p, get_identity_param(p));
+      else
+        _col1 = _col0;
+
+   }
+
+   if (twoside) {
+      if (!p->state->material_shininess_is_zero) {
+         struct ureg shininess = get_material(p, 1, STATE_SHININESS);
+         emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, 
+                  negate(swizzle1(shininess,X)));
+         release_temp(p, shininess);
+      }
+
+      _bfc0 = make_temp(p, get_scenecolor(p, 1));
+      if (separate)
+        _bfc1 = make_temp(p, get_identity_param(p));
+      else
+        _bfc1 = _bfc0;
+   }
+
+   /* If no lights, still need to emit the scenecolor.
+    */
+   {
+      struct ureg res0 = register_output( p, VERT_RESULT_COL0 );
+      emit_op1(p, OPCODE_MOV, res0, 0, _col0);
+   }
+
+   if (separate) {
+      struct ureg res1 = register_output( p, VERT_RESULT_COL1 );
+      emit_op1(p, OPCODE_MOV, res1, 0, _col1);
+   }
+
+   if (twoside) {
+      struct ureg res0 = register_output( p, VERT_RESULT_BFC0 );
+      emit_op1(p, OPCODE_MOV, res0, 0, _bfc0);
+   }
+      
+   if (twoside && separate) {
+      struct ureg res1 = register_output( p, VERT_RESULT_BFC1 );
+      emit_op1(p, OPCODE_MOV, res1, 0, _bfc1);
+   }
+      
+   if (nr_lights == 0) {
+      release_temps(p);
+      return;
+   }
+
+   for (i = 0; i < MAX_LIGHTS; i++) {
+      if (p->state->unit[i].light_enabled) {
+        struct ureg half = undef;
+        struct ureg att = undef, VPpli = undef;
+         
+        count++;
+
+        if (p->state->unit[i].light_eyepos3_is_zero) {
+           /* Can used precomputed constants in this case.
+            * Attenuation never applies to infinite lights.
+            */
+           VPpli = register_param3(p, STATE_INTERNAL, 
+                                   STATE_LIGHT_POSITION_NORMALIZED, i); 
+            
+            if (!p->state->material_shininess_is_zero) {
+               if (p->state->light_local_viewer) {
+                  struct ureg eye_hat = get_eye_position_normalized(p);
+                  half = get_temp(p);
+                  emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
+                  emit_normalize_vec3(p, half, half);
+               } else {
+                  half = register_param3(p, STATE_INTERNAL, 
+                                         STATE_LIGHT_HALF_VECTOR, i);
+               }
+            }
+        } 
+        else {
+           struct ureg Ppli = register_param3(p, STATE_INTERNAL, 
+                                              STATE_LIGHT_POSITION, i); 
+           struct ureg V = get_eye_position(p);
+           struct ureg dist = get_temp(p);
+
+           VPpli = get_temp(p); 
+           /* Calculate VPpli vector
+            */
+           emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V); 
+
+           /* Normalize VPpli.  The dist value also used in
+            * attenuation below.
+            */
+           emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli);
+           emit_op1(p, OPCODE_RSQ, dist, 0, dist);
+           emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist);
+
+           /* Calculate attenuation:
+            */ 
+           if (!p->state->unit[i].light_spotcutoff_is_180 ||
+               p->state->unit[i].light_attenuated) {
+              att = calculate_light_attenuation(p, i, VPpli, dist);
+           }
+
+           /* Calculate viewer direction, or use infinite viewer:
+            */
+            if (!p->state->material_shininess_is_zero) {
+               half = get_temp(p);
+
+               if (p->state->light_local_viewer) {
+                  struct ureg eye_hat = get_eye_position_normalized(p);
+                  emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
+               }
+               else {
+                  struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); 
+                  emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir);
+               }
+
+               emit_normalize_vec3(p, half, half);
+            }
+
+           release_temp(p, dist);
+        }
+
+        /* Calculate dot products:
+         */
+         if (p->state->material_shininess_is_zero) {
+            emit_op2(p, OPCODE_DP3, dots, 0, normal, VPpli);
+         }
+         else {
+            emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
+            emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
+         }
+
+        /* Front face lighting:
+         */
+        {
+           struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT);
+           struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE);
+           struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR);
+           struct ureg res0, res1;
+           GLuint mask0, mask1;
+
+           
+           if (count == nr_lights) {
+              if (separate) {
+                 mask0 = WRITEMASK_XYZ;
+                 mask1 = WRITEMASK_XYZ;
+                 res0 = register_output( p, VERT_RESULT_COL0 );
+                 res1 = register_output( p, VERT_RESULT_COL1 );
+              }
+              else {
+                 mask0 = 0;
+                 mask1 = WRITEMASK_XYZ;
+                 res0 = _col0;
+                 res1 = register_output( p, VERT_RESULT_COL0 );
+              }
+           } else {
+              mask0 = 0;
+              mask1 = 0;
+              res0 = _col0;
+              res1 = _col1;
+           }
+
+
+           if (!is_undef(att)) {
+               /* light is attenuated by distance */
+               emit_op1(p, OPCODE_LIT, lit, 0, dots);
+               emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
+               emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0);
+            } 
+            else if (!p->state->material_shininess_is_zero) {
+               /* there's a non-zero specular term */
+               emit_op1(p, OPCODE_LIT, lit, 0, dots);
+               emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
+            } 
+            else {
+               /* no attenutation, no specular */
+               emit_degenerate_lit(p, lit, dots);
+               emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
+            }
+
+           emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0);
+           emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1);
+      
+           release_temp(p, ambient);
+           release_temp(p, diffuse);
+           release_temp(p, specular);
+        }
+
+        /* Back face lighting:
+         */
+        if (twoside) {
+           struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT);
+           struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE);
+           struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR);
+           struct ureg res0, res1;
+           GLuint mask0, mask1;
+              
+           if (count == nr_lights) {
+              if (separate) {
+                 mask0 = WRITEMASK_XYZ;
+                 mask1 = WRITEMASK_XYZ;
+                 res0 = register_output( p, VERT_RESULT_BFC0 );
+                 res1 = register_output( p, VERT_RESULT_BFC1 );
+              }
+              else {
+                 mask0 = 0;
+                 mask1 = WRITEMASK_XYZ;
+                 res0 = _bfc0;
+                 res1 = register_output( p, VERT_RESULT_BFC0 );
+              }
+           } else {
+              res0 = _bfc0;
+              res1 = _bfc1;
+              mask0 = 0;
+              mask1 = 0;
+           }
+
+            dots = negate(swizzle(dots,X,Y,W,Z));
+
+           if (!is_undef(att)) {
+               emit_op1(p, OPCODE_LIT, lit, 0, dots);
+              emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
+               emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0);
+            }
+            else if (!p->state->material_shininess_is_zero) {
+               emit_op1(p, OPCODE_LIT, lit, 0, dots);
+               emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
+            } 
+            else {
+               emit_degenerate_lit(p, lit, dots);
+               emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0);
+            }
+
+           emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0);
+           emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
+           emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
+
+           release_temp(p, ambient);
+           release_temp(p, diffuse);
+           release_temp(p, specular);
+        }
+
+        release_temp(p, half);
+        release_temp(p, VPpli);
+        release_temp(p, att);
+      }
+   }
+
+   release_temps( p );
+}
+
+
+static void build_fog( struct tnl_program *p )
+{
+   struct ureg fog = register_output(p, VERT_RESULT_FOGC);
+   struct ureg input;
+
+   if (p->state->fog_source_is_depth) {
+      input = get_eye_position_z(p);
+   }
+   else {
+      input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
+   }
+
+   if (p->state->fog_mode && p->state->tnl_do_vertex_fog) {
+      struct ureg params = register_param2(p, STATE_INTERNAL,
+                                          STATE_FOG_PARAMS_OPTIMIZED);
+      struct ureg tmp = get_temp(p);
+      GLboolean useabs = (p->state->fog_mode != FOG_EXP2);
+
+      if (useabs) {
+        emit_op1(p, OPCODE_ABS, tmp, 0, input);
+      }
+
+      switch (p->state->fog_mode) {
+      case FOG_LINEAR: {
+        struct ureg id = get_identity_param(p);
+        emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input,
+                       swizzle1(params,X), swizzle1(params,Y));
+        emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
+        emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
+        break;
+      }
+      case FOG_EXP:
+        emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input,
+                       swizzle1(params,Z));
+        emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
+        break;
+      case FOG_EXP2:
+        emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
+        emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
+        emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
+        break;
+      }
+
+      release_temp(p, tmp);
+   }
+   else {
+      /* results = incoming fog coords (compute fog per-fragment later) 
+       *
+       * KW:  Is it really necessary to do anything in this case?
+       * BP: Yes, we always need to compute the absolute value, unless
+       * we want to push that down into the fragment program...
+       */
+      GLboolean useabs = GL_TRUE;
+      emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input);
+   }
+}
+static void build_reflect_texgen( struct tnl_program *p,
+                                 struct ureg dest,
+                                 GLuint writemask )
+{
+   struct ureg normal = get_transformed_normal(p);
+   struct ureg eye_hat = get_eye_position_normalized(p);
+   struct ureg tmp = get_temp(p);
+
+   /* n.u */
+   emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); 
+   /* 2n.u */
+   emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); 
+   /* (-2n.u)n + u */
+   emit_op3(p, OPCODE_MAD, dest, writemask, negate(tmp), normal, eye_hat);
+
+   release_temp(p, tmp);
+}
+
+static void build_sphere_texgen( struct tnl_program *p,
+                                struct ureg dest,
+                                GLuint writemask )
+{
+   struct ureg normal = get_transformed_normal(p);
+   struct ureg eye_hat = get_eye_position_normalized(p);
+   struct ureg tmp = get_temp(p);
+   struct ureg half = register_scalar_const(p, .5);
+   struct ureg r = get_temp(p);
+   struct ureg inv_m = get_temp(p);
+   struct ureg id = get_identity_param(p);
+
+   /* Could share the above calculations, but it would be
+    * a fairly odd state for someone to set (both sphere and
+    * reflection active for different texture coordinate
+    * components.  Of course - if two texture units enable
+    * reflect and/or sphere, things start to tilt in favour
+    * of seperating this out:
+    */
+
+   /* n.u */
+   emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); 
+   /* 2n.u */
+   emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); 
+   /* (-2n.u)n + u */
+   emit_op3(p, OPCODE_MAD, r, 0, negate(tmp), normal, eye_hat); 
+   /* r + 0,0,1 */
+   emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z)); 
+   /* rx^2 + ry^2 + (rz+1)^2 */
+   emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp); 
+   /* 2/m */
+   emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); 
+   /* 1/m */
+   emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half); 
+   /* r/m + 1/2 */
+   emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half); 
+              
+   release_temp(p, tmp);
+   release_temp(p, r);
+   release_temp(p, inv_m);
+}
+
+
+static void build_texture_transform( struct tnl_program *p )
+{
+   GLuint i, j;
+
+   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
+
+      if (!(p->state->fragprog_inputs_read & FRAG_BIT_TEX(i)))
+        continue;
+                                                            
+      if (p->state->unit[i].texgen_enabled || 
+         p->state->unit[i].texmat_enabled) {
+        
+        GLuint texmat_enabled = p->state->unit[i].texmat_enabled;
+        struct ureg out = register_output(p, VERT_RESULT_TEX0 + i);
+        struct ureg out_texgen = undef;
+
+        if (p->state->unit[i].texgen_enabled) {
+           GLuint copy_mask = 0;
+           GLuint sphere_mask = 0;
+           GLuint reflect_mask = 0;
+           GLuint normal_mask = 0;
+           GLuint modes[4];
+        
+           if (texmat_enabled) 
+              out_texgen = get_temp(p);
+           else
+              out_texgen = out;
+
+           modes[0] = p->state->unit[i].texgen_mode0;
+           modes[1] = p->state->unit[i].texgen_mode1;
+           modes[2] = p->state->unit[i].texgen_mode2;
+           modes[3] = p->state->unit[i].texgen_mode3;
+
+           for (j = 0; j < 4; j++) {
+              switch (modes[j]) {
+              case TXG_OBJ_LINEAR: {
+                 struct ureg obj = register_input(p, VERT_ATTRIB_POS);
+                 struct ureg plane = 
+                    register_param3(p, STATE_TEXGEN, i,
+                                    STATE_TEXGEN_OBJECT_S + j);
+
+                 emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, 
+                          obj, plane );
+                 break;
+              }
+              case TXG_EYE_LINEAR: {
+                 struct ureg eye = get_eye_position(p);
+                 struct ureg plane = 
+                    register_param3(p, STATE_TEXGEN, i, 
+                                    STATE_TEXGEN_EYE_S + j);
+
+                 emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, 
+                          eye, plane );
+                 break;
+              }
+              case TXG_SPHERE_MAP: 
+                 sphere_mask |= WRITEMASK_X << j;
+                 break;
+              case TXG_REFLECTION_MAP:
+                 reflect_mask |= WRITEMASK_X << j;
+                 break;
+              case TXG_NORMAL_MAP: 
+                 normal_mask |= WRITEMASK_X << j;
+                 break;
+              case TXG_NONE:
+                 copy_mask |= WRITEMASK_X << j;
+              }
+
+           }
+
+        
+           if (sphere_mask) {
+              build_sphere_texgen(p, out_texgen, sphere_mask);
+           }
+
+           if (reflect_mask) {
+              build_reflect_texgen(p, out_texgen, reflect_mask);
+           }
+
+           if (normal_mask) {
+              struct ureg normal = get_transformed_normal(p);
+              emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal );
+           }
+
+           if (copy_mask) {
+              struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i);
+              emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in );
+           }
+        }
+
+        if (texmat_enabled) {
+           struct ureg texmat[4];
+           struct ureg in = (!is_undef(out_texgen) ? 
+                             out_texgen : 
+                             register_input(p, VERT_ATTRIB_TEX0+i));
+           if (PREFER_DP4) {
+              register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
+                                      0, texmat );
+              emit_matrix_transform_vec4( p, out, texmat, in );
+           }
+           else {
+              register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
+                                      STATE_MATRIX_TRANSPOSE, texmat );
+              emit_transpose_matrix_transform_vec4( p, out, texmat, in );
+           }
+        }
+
+        release_temps(p);
+      } 
+      else {
+        emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
+      }
+   }
+}
+
+
+/**
+ * Point size attenuation computation.
+ */
+static void build_atten_pointsize( struct tnl_program *p )
+{
+   struct ureg eye = get_eye_position_z(p);
+   struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
+   struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
+   struct ureg out = register_output(p, VERT_RESULT_PSIZ);
+   struct ureg ut = get_temp(p);
+
+   /* dist = |eyez| */
+   emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z));
+   /* p1 + dist * (p2 + dist * p3); */
+   emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
+               swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
+   emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
+               ut, swizzle1(state_attenuation, X));
+
+   /* 1 / sqrt(factor) */
+   emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
+
+#if 0
+   /* out = pointSize / sqrt(factor) */
+   emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
+#else
+   /* this is a good place to clamp the point size since there's likely
+    * no hardware registers to clamp point size at rasterization time.
+    */
+   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
+   emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
+   emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
+#endif
+
+   release_temp(p, ut);
+}
+
+/**
+ * Emit constant point size.
+ */
+static void build_constant_pointsize( struct tnl_program *p )
+{
+   struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
+   struct ureg out = register_output(p, VERT_RESULT_PSIZ);
+   emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, state_size);
+}
+
+/**
+ * Pass-though per-vertex point size, from user's point size array.
+ */
+static void build_array_pointsize( struct tnl_program *p )
+{
+   struct ureg in = register_input(p, VERT_ATTRIB_POINT_SIZE);
+   struct ureg out = register_output(p, VERT_RESULT_PSIZ);
+   emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, in);
+}
+
+
+static void build_tnl_program( struct tnl_program *p )
+{   /* Emit the program, starting with modelviewproject:
+    */
+   build_hpos(p);
+
+   /* Lighting calculations:
+    */
+   if (p->state->fragprog_inputs_read & (FRAG_BIT_COL0|FRAG_BIT_COL1)) {
+      if (p->state->light_global_enabled)
+        build_lighting(p);
+      else {
+        if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
+           emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0);
+
+        if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
+           emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1);
+      }
+   }
+
+   if ((p->state->fragprog_inputs_read & FRAG_BIT_FOGC) ||
+       p->state->fog_mode != FOG_NONE)
+      build_fog(p);
+
+   if (p->state->fragprog_inputs_read & FRAG_BITS_TEX_ANY)
+      build_texture_transform(p);
+
+   if (p->state->point_attenuated)
+      build_atten_pointsize(p);
+   else if (p->state->point_array)
+      build_array_pointsize(p);
+#if 0
+   else
+      build_constant_pointsize(p);
+#else
+   (void) build_constant_pointsize;
+#endif
+
+   /* Finish up:
+    */
+   emit_op1(p, OPCODE_END, undef, 0, undef);
+
+   /* Disassemble:
+    */
+   if (DISASSEM) {
+      _mesa_printf ("\n");
+   }
+}
+
+
+static void
+create_new_program( const struct state_key *key,
+                    struct gl_vertex_program *program,
+                    GLuint max_temps)
+{
+   struct tnl_program p;
+
+   _mesa_memset(&p, 0, sizeof(p));
+   p.state = key;
+   p.program = program;
+   p.eye_position = undef;
+   p.eye_position_z = undef;
+   p.eye_position_normalized = undef;
+   p.transformed_normal = undef;
+   p.identity = undef;
+   p.temp_in_use = 0;
+   
+   if (max_temps >= sizeof(int) * 8)
+      p.temp_reserved = 0;
+   else
+      p.temp_reserved = ~((1<<max_temps)-1);
+
+   /* Start by allocating 32 instructions.
+    * If we need more, we'll grow the instruction array as needed.
+    */
+   p.max_inst = 32;
+   p.program->Base.Instructions = _mesa_alloc_instructions(p.max_inst);
+   p.program->Base.String = NULL;
+   p.program->Base.NumInstructions =
+   p.program->Base.NumTemporaries =
+   p.program->Base.NumParameters =
+   p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
+   p.program->Base.Parameters = _mesa_new_parameter_list();
+   p.program->Base.InputsRead = 0;
+   p.program->Base.OutputsWritten = 0;
+
+   build_tnl_program( &p );
+}
+
+
+/**
+ * Return a vertex program which implements the current fixed-function
+ * transform/lighting/texgen operations.
+ * XXX move this into core mesa (main/)
+ */
+struct gl_vertex_program *
+_mesa_get_fixed_func_vertex_program(GLcontext *ctx)
+{
+   struct gl_vertex_program *prog;
+   struct state_key *key;
+
+   /* Grab all the relevent state and put it in a single structure:
+    */
+   key = make_state_key(ctx);
+
+   /* Look for an already-prepared program for this state:
+    */
+   prog = (struct gl_vertex_program *)
+      _mesa_search_program_cache(ctx->VertexProgram.Cache, key, sizeof(*key));
+   
+   if (!prog) {
+      /* OK, we'll have to build a new one */
+      if (0)
+         _mesa_printf("Build new TNL program\n");
+        
+      prog = (struct gl_vertex_program *)
+         ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); 
+      if (!prog)
+         return NULL;
+
+      create_new_program( key, prog,
+                          ctx->Const.VertexProgram.MaxTemps );
+
+#if 0
+      if (ctx->Driver.ProgramStringNotify)
+         ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, 
+                                          &prog->Base );
+#endif
+      _mesa_program_cache_insert(ctx, ctx->VertexProgram.Cache,
+                                 key, sizeof(*key), &prog->Base);
+   }
+
+   _mesa_free(key);
+
+   return prog;
+}
diff --git a/src/mesa/main/ffvertex_prog.h b/src/mesa/main/ffvertex_prog.h
new file mode 100644 (file)
index 0000000..38dc5fb
--- /dev/null
@@ -0,0 +1,40 @@
+/**************************************************************************
+ * 
+ * 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef FFVERTEX_PROG_H
+#define FFVERTEX_PROG_H
+
+
+#include "main/mtypes.h"
+
+struct gl_vertex_program *
+_mesa_get_fixed_func_vertex_program(GLcontext *ctx);
+
+
+
+#endif /* FFVERTEX_PROG_H */
index e1abcf4db7d9f152f01cc3aa533ccc75b4a8b060..0a7a93063855bb14a7e22a376958d9b8ff0c4ca1 100644 (file)
@@ -123,6 +123,7 @@ typedef int GLfixed;
 /*@{*/
 struct _mesa_HashTable;
 struct gl_pixelstore_attrib;
+struct gl_program_cache;
 struct gl_texture_format;
 struct gl_texture_image;
 struct gl_texture_object;
@@ -149,6 +150,7 @@ enum
    VERT_ATTRIB_COLOR1 = 4,
    VERT_ATTRIB_FOG = 5,
    VERT_ATTRIB_COLOR_INDEX = 6,
+   VERT_ATTRIB_POINT_SIZE = 6,  /*alias*/
    VERT_ATTRIB_EDGEFLAG = 7,
    VERT_ATTRIB_TEX0 = 8,
    VERT_ATTRIB_TEX1 = 9,
@@ -1546,15 +1548,6 @@ struct gl_texture_unit
 };
 
 
-struct texenvprog_cache_item;
-
-struct texenvprog_cache
-{
-   struct texenvprog_cache_item **items;
-   GLuint size, n_items;
-   GLcontext *ctx;
-};
-
 
 /**
  * Texture attribute group (GL_TEXTURE_BIT).
@@ -1581,9 +1574,6 @@ struct gl_texture_attrib
    /** GL_EXT_shared_texture_palette */
    GLboolean SharedPalette;
    struct gl_color_table Palette;
-   
-   /** Cached texenv fragment programs */
-   struct texenvprog_cache env_fp_cache;
 };
 
 
@@ -1984,6 +1974,9 @@ struct gl_vertex_program_state
    /** Program to emulate fixed-function T&L (see above) */
    struct gl_vertex_program *_TnlProgram;
 
+   /** Cache of fixed-function programs */
+   struct gl_program_cache *Cache;
+
 #if FEATURE_MESA_program_debug
    GLprogramcallbackMESA Callback;
    GLvoid *CallbackData;
@@ -2017,6 +2010,9 @@ struct gl_fragment_program_state
    /** Program to emulate fixed-function texture env/combine (see above) */
    struct gl_fragment_program *_TexEnvProgram;
 
+   /** Cache of fixed-function programs */
+   struct gl_program_cache *Cache;
+
 #if FEATURE_MESA_program_debug
    GLprogramcallbackMESA Callback;
    GLvoid *CallbackData;
index d60b9a13d8b92a0183ccf8f9ec4596653eda8914..3323887cb0287d22efd8ea2d645257819a7b51ab 100644 (file)
@@ -62,6 +62,7 @@
 #if FEATURE_EXT_framebuffer_object
 #include "fbobject.h"
 #endif
+#include "ffvertex_prog.h"
 #include "framebuffer.h"
 #include "hint.h"
 #include "histogram.h"
@@ -951,13 +952,12 @@ update_arrays( GLcontext *ctx )
 }
 
 
-/**
- * Update derived vertex/fragment program state.
- */
 static void
 update_program(GLcontext *ctx)
 {
    const struct gl_shader_program *shProg = ctx->Shader.CurrentProgram;
+   const struct gl_vertex_program *prevVP = ctx->VertexProgram._Current;
+   const struct gl_fragment_program *prevFP = ctx->FragmentProgram._Current;
 
    /* These _Enabled flags indicate if the program is enabled AND valid. */
    ctx->VertexProgram._Enabled = ctx->VertexProgram.Enabled
@@ -969,13 +969,18 @@ update_program(GLcontext *ctx)
 
    /*
     * Set the ctx->VertexProgram._Current and ctx->FragmentProgram._Current
-    * pointers to the programs that should be enabled/used.
+    * pointers to the programs that should be enabled/used.  These will only
+    * be NULL if we need to use the fixed-function code.
     *
     * These programs may come from several sources.  The priority is as
     * follows:
     *   1. OpenGL 2.0/ARB vertex/fragment shaders
     *   2. ARB/NV vertex/fragment programs
     *   3. Programs derived from fixed-function state.
+    *
+    * Note: it's possible for a vertex shader to get used with a fragment
+    * program (and vice versa) here, but in practice that shouldn't ever
+    * come up, or matter.
     */
 
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
@@ -991,23 +996,6 @@ update_program(GLcontext *ctx)
                                shProg->FragmentProgram);
    }
    else {
-      if (ctx->VertexProgram._Enabled) {
-         /* use user-defined vertex program */
-         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current,
-                                  ctx->VertexProgram.Current);
-      }
-      else if (ctx->VertexProgram._MaintainTnlProgram) {
-         /* Use vertex program generated from fixed-function state.
-          * The _Current pointer will get set in
-          * _tnl_UpdateFixedFunctionProgram() later if appropriate.
-          */
-         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL);
-      }
-      else {
-         /* no vertex program */
-         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL);
-      }
-
       if (ctx->FragmentProgram._Enabled) {
          /* use user-defined vertex program */
          _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current,
@@ -1015,15 +1003,38 @@ update_program(GLcontext *ctx)
       }
       else if (ctx->FragmentProgram._MaintainTexEnvProgram) {
          /* Use fragment program generated from fixed-function state.
-          * The _Current pointer will get set in _mesa_UpdateTexEnvProgram()
-          * later if appropriate.
           */
-         _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
+         _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current,
+                                  _mesa_get_fixed_func_fragment_program(ctx));
+         _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram,
+                                  ctx->FragmentProgram._Current);
       }
       else {
          /* no fragment program */
          _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL);
       }
+
+      /* Examine vertex program after fragment program as
+       * _mesa_get_fixed_func_vertex_program() needs to know active
+       * fragprog inputs.
+       */
+      if (ctx->VertexProgram._Enabled) {
+         /* use user-defined vertex program */
+         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current,
+                                  ctx->VertexProgram.Current);
+      }
+      else if (ctx->VertexProgram._MaintainTnlProgram) {
+         /* Use vertex program generated from fixed-function state.
+          */
+         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current,
+                                  _mesa_get_fixed_func_vertex_program(ctx));
+         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram,
+                                  ctx->VertexProgram._Current);
+      }
+      else {
+         /* no vertex program */
+         _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL);
+      }
    }
 
    if (ctx->VertexProgram._Current)
@@ -1032,12 +1043,28 @@ update_program(GLcontext *ctx)
       assert(ctx->FragmentProgram._Current->Base.Parameters);
 
 
+   /* XXX: get rid of _Active flag.
+    */
+#if 1
    ctx->FragmentProgram._Active = ctx->FragmentProgram._Enabled;
    if (ctx->FragmentProgram._MaintainTexEnvProgram &&
        !ctx->FragmentProgram._Enabled) {
       if (ctx->FragmentProgram._UseTexEnvProgram)
         ctx->FragmentProgram._Active = GL_TRUE;
    }
+#endif
+
+   /* Let the driver know what's happening:
+    */
+   if (ctx->FragmentProgram._Current != prevFP && ctx->Driver.BindProgram) {
+      ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB,
+                              (struct gl_program *) ctx->FragmentProgram._Current);
+   }
+   
+   if (ctx->VertexProgram._Current != prevVP && ctx->Driver.BindProgram) {
+      ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+                              (struct gl_program *) ctx->VertexProgram._Current);
+   }
 }
 
 
@@ -1197,13 +1224,11 @@ void
 _mesa_update_state_locked( GLcontext *ctx )
 {
    GLbitfield new_state = ctx->NewState;
+   GLbitfield prog_flags = _NEW_PROGRAM;
 
    if (MESA_VERBOSE & VERBOSE_STATE)
       _mesa_print_state("_mesa_update_state", new_state);
 
-   if (new_state & _NEW_PROGRAM)
-      update_program( ctx );
-
    if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION))
       _mesa_update_modelview_project( ctx, new_state );
 
@@ -1249,11 +1274,6 @@ _mesa_update_state_locked( GLcontext *ctx )
       update_tricaps( ctx, new_state );
 #endif
 
-   if (ctx->FragmentProgram._MaintainTexEnvProgram) {
-      if (new_state & (_NEW_TEXTURE | _DD_NEW_SEPARATE_SPECULAR | _NEW_FOG))
-        _mesa_UpdateTexEnvProgram(ctx);
-   }
-
    /* ctx->_NeedEyeCoords is now up to date.
     *
     * If the truth value of this variable has changed, update for the
@@ -1266,6 +1286,20 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _MESA_NEW_NEED_EYE_COORDS) 
       _mesa_update_tnl_spaces( ctx, new_state );
 
+   if (ctx->FragmentProgram._MaintainTexEnvProgram) {
+      prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR);
+   }
+   if (ctx->VertexProgram._MaintainTnlProgram) {
+      prog_flags |= (_NEW_ARRAY | _NEW_TEXTURE | _NEW_TEXTURE_MATRIX |
+                     _NEW_TRANSFORM | _NEW_POINT |
+                     _NEW_FOG | _NEW_LIGHT |
+                     _MESA_NEW_NEED_EYE_COORDS);
+   }
+   if (new_state & prog_flags)
+      update_program( ctx );
+
+
+
    /*
     * Give the driver a chance to act upon the new_state flags.
     * The driver might plug in different span functions, for example.
index 512d52704d822c5659598d5ac7bd3248e9c1e15e..bfb22bbf02cd29f44bcb568e0f8194bb42b261cf 100644 (file)
@@ -1,6 +1,6 @@
 /**************************************************************************
  * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -30,6 +30,7 @@
 #include "enums.h"
 #include "shader/program.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_cache.h"
 #include "shader/prog_instruction.h"
 #include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
@@ -1138,107 +1139,36 @@ create_new_program(GLcontext *ctx, struct state_key *key,
 }
 
 
-static struct gl_fragment_program *
-search_cache(const struct texenvprog_cache *cache,
-             GLuint hash,
-             const void *key,
-             GLuint keysize)
+/**
+ * Return a fragment program which implements the current
+ * fixed-function texture, fog and color-sum operations.
+ */
+struct gl_fragment_program *
+_mesa_get_fixed_func_fragment_program(GLcontext *ctx)
 {
-   struct texenvprog_cache_item *c;
-
-   for (c = cache->items[hash % cache->size]; c; c = c->next) {
-      if (c->hash == hash && memcmp(c->key, key, keysize) == 0)
-        return c->data;
-   }
-
-   return NULL;
-}
+   struct gl_fragment_program *prog;
+   struct state_key key;
+       
+   make_state_key(ctx, &key);
+      
+   prog = (struct gl_fragment_program *)
+      _mesa_search_program_cache(ctx->FragmentProgram.Cache,
+                                 &key, sizeof(key));
 
-static void rehash( struct texenvprog_cache *cache )
-{
-   struct texenvprog_cache_item **items;
-   struct texenvprog_cache_item *c, *next;
-   GLuint size, i;
-
-   size = cache->size * 3;
-   items = (struct texenvprog_cache_item**) _mesa_malloc(size * sizeof(*items));
-   _mesa_memset(items, 0, size * sizeof(*items));
-
-   for (i = 0; i < cache->size; i++)
-      for (c = cache->items[i]; c; c = next) {
-        next = c->next;
-        c->next = items[c->hash % size];
-        items[c->hash % size] = c;
-      }
+   if (!prog) {
+      prog = (struct gl_fragment_program *) 
+         ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
 
-   _mesa_free(cache->items);
-   cache->items = items;
-   cache->size = size;
-}
+      create_new_program(ctx, &key, prog);
 
-static void clear_cache(GLcontext *ctx, struct texenvprog_cache *cache)
-{
-   struct texenvprog_cache_item *c, *next;
-   GLuint i;
-
-   for (i = 0; i < cache->size; i++) {
-      for (c = cache->items[i]; c; c = next) {
-        next = c->next;
-        _mesa_free(c->key);
-         _mesa_reference_fragprog(ctx, &c->data, NULL);
-        _mesa_free(c);
-      }
-      cache->items[i] = NULL;
+      _mesa_program_cache_insert(ctx, ctx->FragmentProgram.Cache,
+                                 &key, sizeof(key), &prog->Base);
    }
 
-
-   cache->n_items = 0;
+   return prog;
 }
 
 
-static void cache_item( GLcontext *ctx,
-                        struct texenvprog_cache *cache,
-                       GLuint hash,
-                       const struct state_key *key,
-                        struct gl_fragment_program *prog)
-{
-   struct texenvprog_cache_item *c = CALLOC_STRUCT(texenvprog_cache_item);
-   c->hash = hash;
-
-   c->key = _mesa_malloc(sizeof(*key));
-   memcpy(c->key, key, sizeof(*key));
-
-   c->data = prog;
-
-   if (cache->n_items > cache->size * 1.5) {
-      if (cache->size < 1000)
-        rehash(cache);
-      else 
-        clear_cache(ctx, cache);
-   }
-
-   cache->n_items++;
-   c->next = cache->items[hash % cache->size];
-   cache->items[hash % cache->size] = c;
-}
-
-static GLuint hash_key( const struct state_key *key )
-{
-   GLuint *ikey = (GLuint *)key;
-   GLuint hash = 0, i;
-
-   /* Make a slightly better attempt at a hash function:
-    */
-   for (i = 0; i < sizeof(*key)/sizeof(*ikey); i++)
-   {
-      hash += ikey[i];
-      hash += (hash << 10);
-      hash ^= (hash >> 6);
-   }
-
-   return hash;
-}
-
 
 /**
  * If _MaintainTexEnvProgram is set we'll generate a fragment program that
@@ -1248,44 +1178,22 @@ static GLuint hash_key( const struct state_key *key )
 void
 _mesa_UpdateTexEnvProgram( GLcontext *ctx )
 {
-   struct state_key key;
-   GLuint hash;
    const struct gl_fragment_program *prev = ctx->FragmentProgram._Current;
        
    ASSERT(ctx->FragmentProgram._MaintainTexEnvProgram);
 
    /* If a conventional fragment program/shader isn't in effect... */
    if (!ctx->FragmentProgram._Enabled &&
-       (!ctx->Shader.CurrentProgram || !ctx->Shader.CurrentProgram->FragmentProgram)) {
+       (!ctx->Shader.CurrentProgram ||
+        !ctx->Shader.CurrentProgram->FragmentProgram) ) 
+   {
       struct gl_fragment_program *newProg;
 
-      make_state_key(ctx, &key);
-      hash = hash_key(&key);
-      
-      newProg = search_cache(&ctx->Texture.env_fp_cache, hash, &key, sizeof(key));
-
-      if (!newProg) {
-         /* create new tex env program */
-
-         if (0)
-            _mesa_printf("Building new texenv proggy for key %x\n", hash);
-
-         newProg = (struct gl_fragment_program *) 
-            ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
-
-         create_new_program(ctx, &key, newProg);
-
-         /* Our ownership of newProg is transferred to the cache */
-         cache_item(ctx, &ctx->Texture.env_fp_cache, hash, &key, newProg);
-      }
+      newProg = _mesa_get_fixed_func_fragment_program(ctx);
 
       _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, newProg);
       _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, newProg);
    } 
-   else {
-      /* _Current pointer has been updated in update_program */
-      /* ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; */
-   }
 
    /* Tell the driver about the change.  Could define a new target for
     * this?
@@ -1295,21 +1203,3 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx )
                          (struct gl_program *) ctx->FragmentProgram._Current);
    }
 }
-
-
-void _mesa_TexEnvProgramCacheInit( GLcontext *ctx )
-{
-   ctx->Texture.env_fp_cache.ctx = ctx;
-   ctx->Texture.env_fp_cache.size = 17;
-   ctx->Texture.env_fp_cache.n_items = 0;
-   ctx->Texture.env_fp_cache.items = (struct texenvprog_cache_item **)
-      _mesa_calloc(ctx->Texture.env_fp_cache.size * 
-                  sizeof(struct texenvprog_cache_item));
-}
-
-
-void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx )
-{
-   clear_cache(ctx, &ctx->Texture.env_fp_cache);
-   _mesa_free(ctx->Texture.env_fp_cache.items);
-}
index 6f017767c8233714af52fabb3dc447d6a86bd44a..a7aa60cf374016821d6096b935bbcedd2f3720e4 100644 (file)
@@ -1,13 +1,8 @@
-/**
- * \file texenvprogram.h
- * Texture state management.
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  5.1
+ * Version:  7.1
  *
- * Copyright (C) 1999-2002  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   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"),
@@ -34,8 +29,9 @@
 
 #include "mtypes.h"
 
+extern struct gl_fragment_program *
+_mesa_get_fixed_func_fragment_program(GLcontext *ctx);
+
 extern void _mesa_UpdateTexEnvProgram( GLcontext *ctx );
-extern void _mesa_TexEnvProgramCacheInit( GLcontext *ctx );
-extern void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx );
 
 #endif
index dbaeaa201ad977dd99d7c686dc25b8c23a209274..4adfa6b385a4522d181926520857e951d5fd9a08 100644 (file)
@@ -3177,8 +3177,6 @@ _mesa_init_texture(GLcontext *ctx)
     */
    assert(ctx->Shared->Default1D->RefCount >= MAX_TEXTURE_UNITS + 1);
 
-   _mesa_TexEnvProgramCacheInit( ctx );
-
    /* Allocate proxy textures */
    if (!alloc_proxy_textures( ctx ))
       return GL_FALSE;
@@ -3214,7 +3212,6 @@ _mesa_free_texture_data(GLcontext *ctx)
    for (u = 0; u < MAX_TEXTURE_IMAGE_UNITS; u++)
       _mesa_free_colortable_data( &ctx->Texture.Unit[u].ColorTable );
 
-   _mesa_TexEnvProgramCacheDestroy( ctx );
 }
 
 
diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c
new file mode 100644 (file)
index 0000000..36a2537
--- /dev/null
@@ -0,0 +1,193 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 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.
+ * 
+ **************************************************************************/
+
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/imports.h"
+#include "shader/prog_cache.h"
+#include "shader/program.h"
+
+
+struct cache_item
+{
+   GLuint hash;
+   void *key;
+   struct gl_program *program;
+   struct cache_item *next;
+};
+
+struct gl_program_cache
+{
+   struct cache_item **items;
+   GLuint size, n_items;
+};
+
+
+
+/**
+ * Compute hash index from state key.
+ */
+static GLuint
+hash_key(const void *key, GLuint key_size)
+{
+   const GLuint *ikey = (const GLuint *) key;
+   GLuint hash = 0, i;
+
+   assert(key_size >= 4);
+
+   /* Make a slightly better attempt at a hash function:
+    */
+   for (i = 0; i < key_size / sizeof(*ikey); i++)
+   {
+      hash += ikey[i];
+      hash += (hash << 10);
+      hash ^= (hash >> 6);
+   }
+
+   return hash;
+}
+
+
+/**
+ * Rebuild/expand the hash table to accomodate more entries
+ */
+static void
+rehash(struct gl_program_cache *cache)
+{
+   struct cache_item **items;
+   struct cache_item *c, *next;
+   GLuint size, i;
+
+   size = cache->size * 3;
+   items = (struct cache_item**) _mesa_malloc(size * sizeof(*items));
+   _mesa_memset(items, 0, size * sizeof(*items));
+
+   for (i = 0; i < cache->size; i++)
+      for (c = cache->items[i]; c; c = next) {
+        next = c->next;
+        c->next = items[c->hash % size];
+        items[c->hash % size] = c;
+      }
+
+   _mesa_free(cache->items);
+   cache->items = items;
+   cache->size = size;
+}
+
+
+static void
+clear_cache(GLcontext *ctx, struct gl_program_cache *cache)
+{
+   struct cache_item *c, *next;
+   GLuint i;
+
+   for (i = 0; i < cache->size; i++) {
+      for (c = cache->items[i]; c; c = next) {
+        next = c->next;
+        _mesa_free(c->key);
+         _mesa_reference_program(ctx, &c->program, NULL);
+        _mesa_free(c);
+      }
+      cache->items[i] = NULL;
+   }
+
+
+   cache->n_items = 0;
+}
+
+
+
+struct gl_program_cache *
+_mesa_new_program_cache(void)
+{
+   struct gl_program_cache *cache = CALLOC_STRUCT(gl_program_cache);
+   if (cache) {
+      cache->size = 17;
+      cache->items = (struct cache_item **)
+         _mesa_calloc(cache->size * sizeof(struct cache_item));
+      if (!cache->items) {
+         _mesa_free(cache);
+         return NULL;
+      }
+   }
+   return cache;
+}
+
+
+void
+_mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *cache)
+{
+   clear_cache(ctx, cache);
+   _mesa_free(cache->items);
+   _mesa_free(cache);
+}
+
+
+struct gl_program *
+_mesa_search_program_cache(const struct gl_program_cache *cache,
+                           const void *key, GLuint keysize)
+{
+   const GLuint hash = hash_key(key, keysize);
+   struct cache_item *c;
+
+   for (c = cache->items[hash % cache->size]; c; c = c->next) {
+      if (c->hash == hash && memcmp(c->key, key, keysize) == 0)
+        return c->program;
+   }
+
+   return NULL;
+}
+
+
+void
+_mesa_program_cache_insert(GLcontext *ctx,
+                           struct gl_program_cache *cache,
+                           const void *key, GLuint keysize,
+                           struct gl_program *program)
+{
+   const GLuint hash = hash_key(key, keysize);
+   struct cache_item *c = CALLOC_STRUCT(cache_item);
+
+   c->hash = hash;
+
+   c->key = _mesa_malloc(keysize);
+   memcpy(c->key, key, keysize);
+
+   c->program = program;  /* no refcount change */
+
+   if (cache->n_items > cache->size * 1.5) {
+      if (cache->size < 1000)
+        rehash(cache);
+      else 
+        clear_cache(ctx, cache);
+   }
+
+   cache->n_items++;
+   c->next = cache->items[hash % cache->size];
+   cache->items[hash % cache->size] = c;
+}
diff --git a/src/mesa/shader/prog_cache.h b/src/mesa/shader/prog_cache.h
new file mode 100644 (file)
index 0000000..a8c91fb
--- /dev/null
@@ -0,0 +1,55 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 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.
+ * 
+ **************************************************************************/
+
+
+#ifndef PROG_CACHE_H
+#define PROG_CACHE_H
+
+
+/** Opaque type */
+struct gl_program_cache;
+
+
+extern struct gl_program_cache *
+_mesa_new_program_cache(void);
+
+extern void
+_mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *pc);
+
+
+extern struct gl_program *
+_mesa_search_program_cache(const struct gl_program_cache *cache,
+                           const void *key, GLuint keysize);
+
+extern void
+_mesa_program_cache_insert(GLcontext *ctx,
+                           struct gl_program_cache *cache,
+                           const void *key, GLuint keysize,
+                           struct gl_program *program);
+
+
+#endif /* PROG_CACHE_H */
index 63f028567d1d4b9c551d9d4db5ee36dbf8c8ca6e..9d543157f76df9dabc64b59863dd17eae4f4ac2c 100644 (file)
@@ -134,10 +134,6 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
               value[3] = 1.0;
             }
             return;
-        case STATE_POSITION_NORMALIZED:
-            COPY_4V(value, ctx->Light.Light[ln].EyePosition);
-           NORMALIZE_3FV( value );
-            return;
          default:
             _mesa_problem(ctx, "Invalid light state in fetch_state");
             return;
@@ -401,7 +397,11 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
    case STATE_INTERNAL:
       switch (state[1]) {
       case STATE_NORMAL_SCALE:
-         ASSIGN_4V(value, ctx->_ModelViewInvScale, 0, 0, 1);
+         ASSIGN_4V(value, 
+                   ctx->_ModelViewInvScale, 
+                   ctx->_ModelViewInvScale, 
+                   ctx->_ModelViewInvScale, 
+                   1);
          return;
       case STATE_TEXRECT_SCALE:
          {
@@ -431,15 +431,45 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[],
          value[2] = ctx->Fog.Density * ONE_DIV_LN2;
          value[3] = ctx->Fog.Density * ONE_DIV_SQRT_LN2;
          return;
-      case STATE_SPOT_DIR_NORMALIZED: {
+
+      case STATE_LIGHT_SPOT_DIR_NORMALIZED: {
          /* here, state[2] is the light number */
          /* pre-normalize spot dir */
          const GLuint ln = (GLuint) state[2];
-         COPY_3V(value, ctx->Light.Light[ln].EyeDirection);
-         NORMALIZE_3FV(value);
+         COPY_3V(value, ctx->Light.Light[ln]._NormDirection);
          value[3] = ctx->Light.Light[ln]._CosCutoff;
          return;
       }
+
+      case STATE_LIGHT_POSITION: {
+         const GLuint ln = (GLuint) state[2];
+         COPY_4V(value, ctx->Light.Light[ln]._Position);
+         return;
+      }
+
+      case STATE_LIGHT_POSITION_NORMALIZED: {
+         const GLuint ln = (GLuint) state[2];
+         COPY_4V(value, ctx->Light.Light[ln]._Position);
+         NORMALIZE_3FV( value );
+         return;
+      }
+
+      case STATE_LIGHT_HALF_VECTOR: {
+         const GLuint ln = (GLuint) state[2];
+         GLfloat p[3];
+         /* Compute infinite half angle vector:
+          *   halfVector = normalize(normalize(lightPos) + (0, 0, 1))
+          * light.EyePosition.w should be 0 for infinite lights.
+          */
+         COPY_3V(p, ctx->Light.Light[ln]._Position);
+         NORMALIZE_3FV(p);
+         ADD_3V(value, p, ctx->_EyeZDir);
+         NORMALIZE_3FV(value);
+         value[3] = 1.0;
+         return;
+      }                                                  
+
+
       case STATE_PT_SCALE:
          value[0] = ctx->Pixel.RedScale;
          value[1] = ctx->Pixel.GreenScale;
@@ -711,7 +741,6 @@ append_token(char *dst, gl_state_index k)
       append(dst, "normalScale");
       break;
    case STATE_INTERNAL:
-   case STATE_POSITION_NORMALIZED:
       append(dst, "(internal)");
       break;
    case STATE_PT_SCALE:
index 2efe166e1a872d593c86bea7f405a055b5d45e26..d3091147f82d9ca4797df70119791aca8b3abfa4 100644 (file)
@@ -106,9 +106,11 @@ typedef enum gl_state_index_ {
    STATE_INTERNAL,             /* Mesa additions */
    STATE_NORMAL_SCALE,
    STATE_TEXRECT_SCALE,
-   STATE_POSITION_NORMALIZED,   /* normalized light position */
    STATE_FOG_PARAMS_OPTIMIZED,  /* for faster fog calc */
-   STATE_SPOT_DIR_NORMALIZED,   /* pre-normalized spot dir */
+   STATE_LIGHT_SPOT_DIR_NORMALIZED,   /* pre-normalized spot dir */
+   STATE_LIGHT_POSITION,              /* object vs eye space */
+   STATE_LIGHT_POSITION_NORMALIZED,   /* object vs eye space */
+   STATE_LIGHT_HALF_VECTOR,           /* object vs eye space */
    STATE_PT_SCALE,              /**< Pixel transfer RGBA scale */
    STATE_PT_BIAS,               /**< Pixel transfer RGBA bias */
    STATE_PCM_SCALE,             /**< Post color matrix RGBA scale */
index 9f9d5955eaf5faccc0456c9648a3b46f92d64c83..2932c9ebc50232706a5df0d500832738f3a6dfeb 100644 (file)
@@ -33,6 +33,7 @@
 #include "main/context.h"
 #include "main/hash.h"
 #include "program.h"
+#include "prog_cache.h"
 #include "prog_parameter.h"
 #include "prog_instruction.h"
 
@@ -66,6 +67,7 @@ _mesa_init_program(GLcontext *ctx)
       ctx->VertexProgram.TrackMatrix[i] = GL_NONE;
       ctx->VertexProgram.TrackMatrixTransform[i] = GL_IDENTITY_NV;
    }
+   ctx->VertexProgram.Cache = _mesa_new_program_cache();
 #endif
 
 #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program
@@ -73,8 +75,10 @@ _mesa_init_program(GLcontext *ctx)
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current,
                             ctx->Shared->DefaultFragmentProgram);
    assert(ctx->FragmentProgram.Current);
+   ctx->FragmentProgram.Cache = _mesa_new_program_cache();
 #endif
 
+
    /* XXX probably move this stuff */
 #if FEATURE_ATI_fragment_shader
    ctx->ATIFragmentShader.Enabled = GL_FALSE;
@@ -93,9 +97,11 @@ _mesa_free_program_data(GLcontext *ctx)
 {
 #if FEATURE_NV_vertex_program || FEATURE_ARB_vertex_program
    _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL);
+   _mesa_delete_program_cache(ctx, ctx->VertexProgram.Cache);
 #endif
 #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program
    _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL);
+   _mesa_delete_program_cache(ctx, ctx->FragmentProgram.Cache);
 #endif
    /* XXX probably move this stuff */
 #if FEATURE_ATI_fragment_shader
index c5f717447ead8bd58ff504f74930ea71ced53122..7a3b9fb2bcbf403adc3a744990549c47c393a340 100644 (file)
@@ -27,6 +27,7 @@ MAIN_SOURCES = \
        main/extensions.c \
        main/fbobject.c \
        main/feedback.c \
+       main/ffvertex_prog.c \
        main/fog.c \
        main/framebuffer.c \
        main/get.c \
@@ -160,6 +161,7 @@ SHADER_SOURCES = \
        shader/nvprogram.c \
        shader/nvvertparse.c \
        shader/program.c \
+       shader/prog_cache.c \
        shader/prog_debug.c \
        shader/prog_execute.c \
        shader/prog_instruction.c \
index 00f43b553bd6850358d1ea70843d993773b2e015..8977fadcca23a11cfb493d7de6dbe09730551989 100644 (file)
@@ -61,7 +61,6 @@ _tnl_CreateContext( GLcontext *ctx )
    /* Initialize tnl state.
     */
    if (ctx->VertexProgram._MaintainTnlProgram) {
-      _tnl_ProgramCacheInit( ctx );
       _tnl_install_pipeline( ctx, _tnl_vp_pipeline );
    } else {
       _tnl_install_pipeline( ctx, _tnl_default_pipeline );
@@ -90,9 +89,6 @@ _tnl_DestroyContext( GLcontext *ctx )
 
    _tnl_destroy_pipeline( ctx );
 
-   if (ctx->VertexProgram._MaintainTnlProgram)
-      _tnl_ProgramCacheDestroy( ctx );
-
    FREE(tnl);
    ctx->swtnl_context = NULL;
 }
index 1ac508f0330fd8e8b0b1f1cea9519aaee8d7b653..0a6ce046146e9e33cf0368a481d89836cfc70635 100644 (file)
@@ -385,19 +385,6 @@ struct tnl_clipspace
 };
 
 
-struct tnl_cache_item {
-   GLuint hash;
-   void *key;
-   struct gl_vertex_program *prog;
-   struct tnl_cache_item *next;
-};
-
-struct tnl_cache {
-   struct tnl_cache_item **items;
-   GLuint size, n_items;
-};
-
-
 struct tnl_device_driver
 {
    /***
@@ -549,10 +536,6 @@ typedef struct
    GLubyte *block[VERT_ATTRIB_MAX];
    GLuint nr_blocks;
 
-   /* Cache of fixed-function-replacing vertex programs:
-    */
-   struct tnl_cache *vp_cache;
-
 } TNLcontext;
 
 
index 1a2d9b8a57fe4f1220c3dab9d189558a0267437c..7be4d95af6b6c121a48331109443529b2c833fbc 100644 (file)
 
 
 #include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "shader/program.h"
-#include "shader/prog_instruction.h"
-#include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
-#include "shader/prog_statevars.h"
-#include "t_context.h" /* NOTE: very light dependency on this */
+#include "main/ffvertex_prog.h"
 #include "t_vp_build.h"
 
 
-struct state_key {
-   unsigned light_global_enabled:1;
-   unsigned light_local_viewer:1;
-   unsigned light_twoside:1;
-   unsigned light_color_material:1;
-   unsigned light_color_material_mask:12;
-   unsigned light_material_mask:12;
-
-   unsigned normalize:1;
-   unsigned rescale_normals:1;
-   unsigned fog_source_is_depth:1;
-   unsigned tnl_do_vertex_fog:1;
-   unsigned separate_specular:1;
-   unsigned fog_mode:2;
-   unsigned point_attenuated:1;
-   unsigned texture_enabled_global:1;
-   unsigned fragprog_inputs_read:12;
-
-   struct {
-      unsigned light_enabled:1;
-      unsigned light_eyepos3_is_zero:1;
-      unsigned light_spotcutoff_is_180:1;
-      unsigned light_attenuated:1;      
-      unsigned texunit_really_enabled:1;
-      unsigned texmat_enabled:1;
-      unsigned texgen_enabled:4;
-      unsigned texgen_mode0:4;
-      unsigned texgen_mode1:4;
-      unsigned texgen_mode2:4;
-      unsigned texgen_mode3:4;
-   } unit[8];
-};
-
-
-
-#define FOG_NONE   0
-#define FOG_LINEAR 1
-#define FOG_EXP    2
-#define FOG_EXP2   3
-
-static GLuint translate_fog_mode( GLenum mode )
-{
-   switch (mode) {
-   case GL_LINEAR: return FOG_LINEAR;
-   case GL_EXP: return FOG_EXP;
-   case GL_EXP2: return FOG_EXP2;
-   default: return FOG_NONE;
-   }
-}
-
-#define TXG_NONE           0
-#define TXG_OBJ_LINEAR     1
-#define TXG_EYE_LINEAR     2
-#define TXG_SPHERE_MAP     3
-#define TXG_REFLECTION_MAP 4
-#define TXG_NORMAL_MAP     5
-
-static GLuint translate_texgen( GLboolean enabled, GLenum mode )
-{
-   if (!enabled)
-      return TXG_NONE;
-
-   switch (mode) {
-   case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR;
-   case GL_EYE_LINEAR: return TXG_EYE_LINEAR;
-   case GL_SPHERE_MAP: return TXG_SPHERE_MAP;
-   case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP;
-   case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP;
-   default: return TXG_NONE;
-   }
-}
-
-static struct state_key *make_state_key( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
-   struct state_key *key = CALLOC_STRUCT(state_key);
-   GLuint i;
-
-   /* This now relies on texenvprogram.c being active:
-    */
-   assert(fp);
-
-   key->fragprog_inputs_read = fp->Base.InputsRead;
-
-   key->separate_specular = (ctx->Light.Model.ColorControl ==
-                            GL_SEPARATE_SPECULAR_COLOR);
-
-   if (ctx->Light.Enabled) {
-      key->light_global_enabled = 1;
-
-      if (ctx->Light.Model.LocalViewer)
-        key->light_local_viewer = 1;
-
-      if (ctx->Light.Model.TwoSide)
-        key->light_twoside = 1;
-
-      if (ctx->Light.ColorMaterialEnabled) {
-        key->light_color_material = 1;
-        key->light_color_material_mask = ctx->Light.ColorMaterialBitmask;
-      }
-
-      for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) 
-        if (VB->AttribPtr[i]->stride) 
-           key->light_material_mask |= 1<<(i-_TNL_ATTRIB_MAT_FRONT_AMBIENT);
-
-      for (i = 0; i < MAX_LIGHTS; i++) {
-        struct gl_light *light = &ctx->Light.Light[i];
-
-        if (light->Enabled) {
-           key->unit[i].light_enabled = 1;
-
-           if (light->EyePosition[3] == 0.0)
-              key->unit[i].light_eyepos3_is_zero = 1;
-           
-           if (light->SpotCutoff == 180.0)
-              key->unit[i].light_spotcutoff_is_180 = 1;
-
-           if (light->ConstantAttenuation != 1.0 ||
-               light->LinearAttenuation != 0.0 ||
-               light->QuadraticAttenuation != 0.0)
-              key->unit[i].light_attenuated = 1;
-        }
-      }
-   }
-
-   if (ctx->Transform.Normalize)
-      key->normalize = 1;
-
-   if (ctx->Transform.RescaleNormals)
-      key->rescale_normals = 1;
-
-   key->fog_mode = translate_fog_mode(fp->FogOption);
-   
-   if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT)
-      key->fog_source_is_depth = 1;
-   
-   if (tnl->_DoVertexFog)
-      key->tnl_do_vertex_fog = 1;
-
-   if (ctx->Point._Attenuated)
-      key->point_attenuated = 1;
-
-   if (ctx->Texture._TexGenEnabled ||
-       ctx->Texture._TexMatEnabled ||
-       ctx->Texture._EnabledUnits)
-      key->texture_enabled_global = 1;
-      
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
-
-      if (texUnit->_ReallyEnabled)
-        key->unit[i].texunit_really_enabled = 1;
-
-      if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i))      
-        key->unit[i].texmat_enabled = 1;
-      
-      if (texUnit->TexGenEnabled) {
-        key->unit[i].texgen_enabled = 1;
-      
-        key->unit[i].texgen_mode0 = 
-           translate_texgen( texUnit->TexGenEnabled & (1<<0),
-                             texUnit->GenModeS );
-        key->unit[i].texgen_mode1 = 
-           translate_texgen( texUnit->TexGenEnabled & (1<<1),
-                             texUnit->GenModeT );
-        key->unit[i].texgen_mode2 = 
-           translate_texgen( texUnit->TexGenEnabled & (1<<2),
-                             texUnit->GenModeR );
-        key->unit[i].texgen_mode3 = 
-           translate_texgen( texUnit->TexGenEnabled & (1<<3),
-                             texUnit->GenModeQ );
-      }
-   }
-   
-   return key;
-}
-
-
-   
-/* Very useful debugging tool - produces annotated listing of
- * generated program with line/function references for each
- * instruction back into this file:
- */
-#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM)
-
-/* Should be tunable by the driver - do we want to do matrix
- * multiplications with DP4's or with MUL/MAD's?  SSE works better
- * with the latter, drivers may differ.
- */
-#define PREFER_DP4 0
-
-#define MAX_INSN 350
-
-/* Use uregs to represent registers internally, translate to Mesa's
- * expected formats on emit.  
- *
- * NOTE: These are passed by value extensively in this file rather
- * than as usual by pointer reference.  If this disturbs you, try
- * remembering they are just 32bits in size.
- *
- * GCC is smart enough to deal with these dword-sized structures in
- * much the same way as if I had defined them as dwords and was using
- * macros to access and set the fields.  This is much nicer and easier
- * to evolve.
- */
-struct ureg {
-   GLuint file:4;
-   GLint idx:8;      /* relative addressing may be negative */
-   GLuint negate:1;
-   GLuint swz:12;
-   GLuint pad:7;
-};
-
-
-struct tnl_program {
-   const struct state_key *state;
-   struct gl_vertex_program *program;
-   
-   GLuint temp_in_use;
-   GLuint temp_reserved;
-   
-   struct ureg eye_position;
-   struct ureg eye_position_normalized;
-   struct ureg eye_normal;
-   struct ureg identity;
-
-   GLuint materials;
-   GLuint color_materials;
-};
-
-
-static const struct ureg undef = { 
-   PROGRAM_UNDEFINED,
-   ~0,
-   0,
-   0,
-   0
-};
-
-/* Local shorthand:
- */
-#define X    SWIZZLE_X
-#define Y    SWIZZLE_Y
-#define Z    SWIZZLE_Z
-#define W    SWIZZLE_W
-
-
-/* Construct a ureg:
- */
-static struct ureg make_ureg(GLuint file, GLint idx)
-{
-   struct ureg reg;
-   reg.file = file;
-   reg.idx = idx;
-   reg.negate = 0;
-   reg.swz = SWIZZLE_NOOP;
-   reg.pad = 0;
-   return reg;
-}
-
-
-
-static struct ureg negate( struct ureg reg )
-{
-   reg.negate ^= 1;
-   return reg;
-} 
-
-
-static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
-{
-   reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x),
-                          GET_SWZ(reg.swz, y),
-                          GET_SWZ(reg.swz, z),
-                          GET_SWZ(reg.swz, w));
-
-   return reg;
-}
-
-static struct ureg swizzle1( struct ureg reg, int x )
-{
-   return swizzle(reg, x, x, x, x);
-}
-
-static struct ureg get_temp( struct tnl_program *p )
-{
-   int bit = _mesa_ffs( ~p->temp_in_use );
-   if (!bit) {
-      _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__);
-      _mesa_exit(1);
-   }
-
-   if ((GLuint) bit > p->program->Base.NumTemporaries)
-      p->program->Base.NumTemporaries = bit;
-
-   p->temp_in_use |= 1<<(bit-1);
-   return make_ureg(PROGRAM_TEMPORARY, bit-1);
-}
-
-static struct ureg reserve_temp( struct tnl_program *p )
-{
-   struct ureg temp = get_temp( p );
-   p->temp_reserved |= 1<<temp.idx;
-   return temp;
-}
-
-static void release_temp( struct tnl_program *p, struct ureg reg )
-{
-   if (reg.file == PROGRAM_TEMPORARY) {
-      p->temp_in_use &= ~(1<<reg.idx);
-      p->temp_in_use |= p->temp_reserved; /* can't release reserved temps */
-   }
-}
-
-static void release_temps( struct tnl_program *p )
-{
-   p->temp_in_use = p->temp_reserved;
-}
-
-
-
-static struct ureg register_input( struct tnl_program *p, GLuint input )
-{
-   p->program->Base.InputsRead |= (1<<input);
-   return make_ureg(PROGRAM_INPUT, input);
-}
-
-static struct ureg register_output( struct tnl_program *p, GLuint output )
-{
-   p->program->Base.OutputsWritten |= (1<<output);
-   return make_ureg(PROGRAM_OUTPUT, output);
-}
-
-static struct ureg register_const4f( struct tnl_program *p, 
-                             GLfloat s0,
-                             GLfloat s1,
-                             GLfloat s2,
-                             GLfloat s3)
-{
-   GLfloat values[4];
-   GLint idx;
-   GLuint swizzle;
-   values[0] = s0;
-   values[1] = s1;
-   values[2] = s2;
-   values[3] = s3;
-   idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
-                                     &swizzle );
-   ASSERT(swizzle == SWIZZLE_NOOP);
-   return make_ureg(PROGRAM_STATE_VAR, idx);
-}
-
-#define register_const1f(p, s0)         register_const4f(p, s0, 0, 0, 1)
-#define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
-#define register_const2f(p, s0, s1)     register_const4f(p, s0, s1, 0, 1)
-#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
-
-static GLboolean is_undef( struct ureg reg )
-{
-   return reg.file == PROGRAM_UNDEFINED;
-}
-
-static struct ureg get_identity_param( struct tnl_program *p )
-{
-   if (is_undef(p->identity)) 
-      p->identity = register_const4f(p, 0,0,0,1);
-
-   return p->identity;
-}
-
-static struct ureg register_param5(struct tnl_program *p, 
-                                  GLint s0,
-                                  GLint s1,
-                                  GLint s2,
-                                  GLint s3,
-                                   GLint s4)
-{
-   gl_state_index tokens[STATE_LENGTH];
-   GLint idx;
-   tokens[0] = s0;
-   tokens[1] = s1;
-   tokens[2] = s2;
-   tokens[3] = s3;
-   tokens[4] = s4;
-   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
-   return make_ureg(PROGRAM_STATE_VAR, idx);
-}
-
-
-#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
-#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
-#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
-#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
-
-
-static void register_matrix_param5( struct tnl_program *p,
-                                   GLint s0, /* modelview, projection, etc */
-                                   GLint s1, /* texture matrix number */
-                                   GLint s2, /* first row */
-                                   GLint s3, /* last row */
-                                   GLint s4, /* inverse, transpose, etc */
-                                   struct ureg *matrix )
-{
-   GLint i;
-
-   /* This is a bit sad as the support is there to pull the whole
-    * matrix out in one go:
-    */
-   for (i = 0; i <= s3 - s2; i++) 
-      matrix[i] = register_param5( p, s0, s1, i, i, s4 );
-}
-
-
-/**
- * Convert a ureg source register to a prog_src_register.
- */
-static void emit_arg( struct prog_src_register *src,
-                     struct ureg reg )
-{
-   assert(reg.file != PROGRAM_OUTPUT);
-   src->File = reg.file;
-   src->Index = reg.idx;
-   src->Swizzle = reg.swz;
-   src->NegateBase = reg.negate ? NEGATE_XYZW : 0;
-   src->Abs = 0;
-   src->NegateAbs = 0;
-   src->RelAddr = 0;
-}
-
 /**
- * Convert a ureg dest register to a prog_dst_register.
- */
-static void emit_dst( struct prog_dst_register *dst,
-                     struct ureg reg, GLuint mask )
-{
-   /* Check for legal output register type.  UNDEFINED will occur in
-    * instruction that don't produce a result (like END).
-    */
-   assert(reg.file == PROGRAM_TEMPORARY ||
-          reg.file == PROGRAM_OUTPUT ||
-          reg.file == PROGRAM_UNDEFINED);
-   dst->File = reg.file;
-   dst->Index = reg.idx;
-   /* allow zero as a shorthand for xyzw */
-   dst->WriteMask = mask ? mask : WRITEMASK_XYZW; 
-   dst->CondMask = COND_TR;  /* always pass cond test */
-   dst->CondSwizzle = SWIZZLE_NOOP;
-   dst->CondSrc = 0;
-   dst->pad = 0;
-}
-
-static void debug_insn( struct prog_instruction *inst, const char *fn,
-                       GLuint line )
-{
-   if (DISASSEM) {
-      static const char *last_fn;
-   
-      if (fn != last_fn) {
-        last_fn = fn;
-        _mesa_printf("%s:\n", fn);
-      }
-        
-      _mesa_printf("%d:\t", line);
-      _mesa_print_instruction(inst);
-   }
-}
-
-
-static void emit_op3fn(struct tnl_program *p,
-                       enum prog_opcode op,
-                      struct ureg dest,
-                      GLuint mask,
-                      struct ureg src0,
-                      struct ureg src1,
-                      struct ureg src2,
-                      const char *fn,
-                      GLuint line)
-{
-   GLuint nr = p->program->Base.NumInstructions++;
-   struct prog_instruction *inst = &p->program->Base.Instructions[nr];
-      
-   if (p->program->Base.NumInstructions > MAX_INSN) {
-      _mesa_problem(0, "Out of instructions in emit_op3fn\n");
-      return;
-   }
-      
-   inst->Opcode = (enum prog_opcode) op; 
-   inst->StringPos = 0;
-   inst->Data = 0;
-   
-   emit_arg( &inst->SrcReg[0], src0 );
-   emit_arg( &inst->SrcReg[1], src1 );
-   emit_arg( &inst->SrcReg[2], src2 );   
-
-   emit_dst( &inst->DstReg, dest, mask );
-
-   debug_insn(inst, fn, line);
-}
-
-
-#define emit_op3(p, op, dst, mask, src0, src1, src2) \
-   emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
-
-#define emit_op2(p, op, dst, mask, src0, src1) \
-    emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
-
-#define emit_op1(p, op, dst, mask, src0) \
-    emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
-
-
-static struct ureg make_temp( struct tnl_program *p, struct ureg reg )
-{
-   if (reg.file == PROGRAM_TEMPORARY && 
-       !(p->temp_reserved & (1<<reg.idx)))
-      return reg;
-   else {
-      struct ureg temp = get_temp(p);
-      emit_op1(p, OPCODE_MOV, temp, 0, reg);
-      return temp;
-   }
-}
-
-
-/* Currently no tracking performed of input/output/register size or
- * active elements.  Could be used to reduce these operations, as
- * could the matrix type.
- */
-static void emit_matrix_transform_vec4( struct tnl_program *p,
-                                       struct ureg dest,
-                                       const struct ureg *mat,
-                                       struct ureg src)
-{
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_X, src, mat[0]);
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Y, src, mat[1]);
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Z, src, mat[2]);
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_W, src, mat[3]);
-}
-
-/* This version is much easier to implement if writemasks are not
- * supported natively on the target or (like SSE), the target doesn't
- * have a clean/obvious dotproduct implementation.
- */
-static void emit_transpose_matrix_transform_vec4( struct tnl_program *p,
-                                                 struct ureg dest,
-                                                 const struct ureg *mat,
-                                                 struct ureg src)
-{
-   struct ureg tmp;
-
-   if (dest.file != PROGRAM_TEMPORARY)
-      tmp = get_temp(p);
-   else
-      tmp = dest;
-
-   emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]);
-   emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp);
-   emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp);
-   emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp);
-
-   if (dest.file != PROGRAM_TEMPORARY)
-      release_temp(p, tmp);
-}
-
-static void emit_matrix_transform_vec3( struct tnl_program *p,
-                                       struct ureg dest,
-                                       const struct ureg *mat,
-                                       struct ureg src)
-{
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_X, src, mat[0]);
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Y, src, mat[1]);
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Z, src, mat[2]);
-}
-
-
-static void emit_normalize_vec3( struct tnl_program *p,
-                                struct ureg dest,
-                                struct ureg src )
-{
-   struct ureg tmp = get_temp(p);
-   emit_op2(p, OPCODE_DP3, tmp, 0, src, src);
-   emit_op1(p, OPCODE_RSQ, tmp, 0, tmp);
-   emit_op2(p, OPCODE_MUL, dest, 0, src, tmp);
-   release_temp(p, tmp);
-}
-
-static void emit_passthrough( struct tnl_program *p, 
-                             GLuint input,
-                             GLuint output )
-{
-   struct ureg out = register_output(p, output);
-   emit_op1(p, OPCODE_MOV, out, 0, register_input(p, input)); 
-}
-
-static struct ureg get_eye_position( struct tnl_program *p )
-{
-   if (is_undef(p->eye_position)) {
-      struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
-      struct ureg modelview[4];
-
-      p->eye_position = reserve_temp(p);
-
-      if (PREFER_DP4) {
-        register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
-                                 0, modelview );
-
-        emit_matrix_transform_vec4(p, p->eye_position, modelview, pos);
-      }
-      else {
-        register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3,
-                                STATE_MATRIX_TRANSPOSE, modelview );
-
-        emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos);
-      }
-   }
-   
-   return p->eye_position;
-}
-
-
-static struct ureg get_eye_position_normalized( struct tnl_program *p )
-{
-   if (is_undef(p->eye_position_normalized)) {
-      struct ureg eye = get_eye_position(p);
-      p->eye_position_normalized = reserve_temp(p);
-      emit_normalize_vec3(p, p->eye_position_normalized, eye);
-   }
-   
-   return p->eye_position_normalized;
-}
-
-
-static struct ureg get_eye_normal( struct tnl_program *p )
-{
-   if (is_undef(p->eye_normal)) {
-      struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL );
-      struct ureg mvinv[3];
-
-      register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2,
-                             STATE_MATRIX_INVTRANS, mvinv );
-
-      p->eye_normal = reserve_temp(p);
-
-      /* Transform to eye space:
-       */
-      emit_matrix_transform_vec3( p, p->eye_normal, mvinv, normal );
-
-      /* Normalize/Rescale:
-       */
-      if (p->state->normalize) {
-        emit_normalize_vec3( p, p->eye_normal, p->eye_normal );
-      }
-      else if (p->state->rescale_normals) {
-        struct ureg rescale = register_param2(p, STATE_INTERNAL,
-                                              STATE_NORMAL_SCALE);
-
-        emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal,
-                  swizzle1(rescale, X));
-      }
-   }
-
-   return p->eye_normal;
-}
-
-
-
-static void build_hpos( struct tnl_program *p )
-{
-   struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
-   struct ureg hpos = register_output( p, VERT_RESULT_HPOS );
-   struct ureg mvp[4];
-
-   if (PREFER_DP4) {
-      register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
-                             0, mvp );
-      emit_matrix_transform_vec4( p, hpos, mvp, pos );
-   }
-   else {
-      register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
-                             STATE_MATRIX_TRANSPOSE, mvp );
-      emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos );
-   }
-}
-
-
-static GLuint material_attrib( GLuint side, GLuint property )
-{
-   return ((property - STATE_AMBIENT) * 2 + 
-          side);
-}
-
-/* Get a bitmask of which material values vary on a per-vertex basis.
- */
-static void set_material_flags( struct tnl_program *p )
-{
-   p->color_materials = 0;
-   p->materials = 0;
-
-   if (p->state->light_color_material) {
-      p->materials = 
-        p->color_materials = p->state->light_color_material_mask;
-   }
-
-   p->materials |= p->state->light_material_mask;
-}
-
-
-static struct ureg get_material( struct tnl_program *p, GLuint side, 
-                                GLuint property )
-{
-   GLuint attrib = material_attrib(side, property);
-
-   if (p->color_materials & (1<<attrib))
-      return register_input(p, VERT_ATTRIB_COLOR0);
-   else if (p->materials & (1<<attrib)) 
-      return register_input( p, attrib + _TNL_ATTRIB_MAT_FRONT_AMBIENT );
-   else
-      return register_param3( p, STATE_MATERIAL, side, property );
-}
-
-#define SCENE_COLOR_BITS(side) (( MAT_BIT_FRONT_EMISSION | \
-                                  MAT_BIT_FRONT_AMBIENT | \
-                                  MAT_BIT_FRONT_DIFFUSE) << (side))
-
-/* Either return a precalculated constant value or emit code to
- * calculate these values dynamically in the case where material calls
- * are present between begin/end pairs.
- *
- * Probably want to shift this to the program compilation phase - if
- * we always emitted the calculation here, a smart compiler could
- * detect that it was constant (given a certain set of inputs), and
- * lift it out of the main loop.  That way the programs created here
- * would be independent of the vertex_buffer details.
+ * XXX This should go away someday, but still referenced by some drivers...
  */
-static struct ureg get_scenecolor( struct tnl_program *p, GLuint side )
-{
-   if (p->materials & SCENE_COLOR_BITS(side)) {
-      struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT);
-      struct ureg material_emission = get_material(p, side, STATE_EMISSION);
-      struct ureg material_ambient = get_material(p, side, STATE_AMBIENT);
-      struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE);
-      struct ureg tmp = make_temp(p, material_diffuse);
-      emit_op3(p, OPCODE_MAD, tmp,  WRITEMASK_XYZ, lm_ambient, 
-              material_ambient, material_emission);
-      return tmp;
-   }
-   else
-      return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side );
-}
-
-
-static struct ureg get_lightprod( struct tnl_program *p, GLuint light, 
-                                 GLuint side, GLuint property )
-{
-   GLuint attrib = material_attrib(side, property);
-   if (p->materials & (1<<attrib)) {
-      struct ureg light_value = 
-        register_param3(p, STATE_LIGHT, light, property);
-      struct ureg material_value = get_material(p, side, property);
-      struct ureg tmp = get_temp(p);
-      emit_op2(p, OPCODE_MUL, tmp,  0, light_value, material_value);
-      return tmp;
-   }
-   else
-      return register_param4(p, STATE_LIGHTPROD, light, side, property);
-}
-
-static struct ureg calculate_light_attenuation( struct tnl_program *p,
-                                               GLuint i, 
-                                               struct ureg VPpli,
-                                               struct ureg dist )
-{
-   struct ureg attenuation = register_param3(p, STATE_LIGHT, i,
-                                            STATE_ATTENUATION);
-   struct ureg att = get_temp(p);
-
-   /* Calculate spot attenuation:
-    */
-   if (!p->state->unit[i].light_spotcutoff_is_180) {
-      struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
-                                                 STATE_SPOT_DIR_NORMALIZED, i);
-      struct ureg spot = get_temp(p);
-      struct ureg slt = get_temp(p);
-
-      emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm);
-      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
-      emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
-      emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
-
-      release_temp(p, spot);
-      release_temp(p, slt);
-   }
-
-   /* Calculate distance attenuation:
-    */
-   if (p->state->unit[i].light_attenuated) {
-
-      /* 1/d,d,d,1/d */
-      emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist); 
-      /* 1,d,d*d,1/d */
-      emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y)); 
-      /* 1/dist-atten */
-      emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist); 
-
-      if (!p->state->unit[i].light_spotcutoff_is_180) {
-        /* dist-atten */
-        emit_op1(p, OPCODE_RCP, dist, 0, dist); 
-        /* spot-atten * dist-atten */
-        emit_op2(p, OPCODE_MUL, att, 0, dist, att);    
-      } else {
-        /* dist-atten */
-        emit_op1(p, OPCODE_RCP, att, 0, dist); 
-      }
-   }
-
-   return att;
-}
-                                               
-
-
-
-
-/* Need to add some addtional parameters to allow lighting in object
- * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
- * space lighting.
- */
-static void build_lighting( struct tnl_program *p )
-{
-   const GLboolean twoside = p->state->light_twoside;
-   const GLboolean separate = p->state->separate_specular;
-   GLuint nr_lights = 0, count = 0;
-   struct ureg normal = get_eye_normal(p);
-   struct ureg lit = get_temp(p);
-   struct ureg dots = get_temp(p);
-   struct ureg _col0 = undef, _col1 = undef;
-   struct ureg _bfc0 = undef, _bfc1 = undef;
-   GLuint i;
-
-   for (i = 0; i < MAX_LIGHTS; i++) 
-      if (p->state->unit[i].light_enabled)
-        nr_lights++;
-   
-   set_material_flags(p);
-
-   {
-      struct ureg shininess = get_material(p, 0, STATE_SHININESS);
-      emit_op1(p, OPCODE_MOV, dots,  WRITEMASK_W, swizzle1(shininess,X));
-
-      _col0 = make_temp(p, get_scenecolor(p, 0));
-      if (separate)
-        _col1 = make_temp(p, get_identity_param(p));
-      else
-        _col1 = _col0;
-
-   }
-
-   if (twoside) {
-      struct ureg shininess = get_material(p, 1, STATE_SHININESS);
-      emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, 
-              negate(swizzle1(shininess,X)));
-
-      _bfc0 = make_temp(p, get_scenecolor(p, 1));
-      if (separate)
-        _bfc1 = make_temp(p, get_identity_param(p));
-      else
-        _bfc1 = _bfc0;
-   }
-
-
-   /* If no lights, still need to emit the scenecolor.
-    */
-      {
-        struct ureg res0 = register_output( p, VERT_RESULT_COL0 );
-        emit_op1(p, OPCODE_MOV, res0, 0, _col0);
-      }
-
-      if (separate) {
-        struct ureg res1 = register_output( p, VERT_RESULT_COL1 );
-        emit_op1(p, OPCODE_MOV, res1, 0, _col1);
-      }
-
-      if (twoside) {
-        struct ureg res0 = register_output( p, VERT_RESULT_BFC0 );
-        emit_op1(p, OPCODE_MOV, res0, 0, _bfc0);
-      }
-      
-      if (twoside && separate) {
-        struct ureg res1 = register_output( p, VERT_RESULT_BFC1 );
-        emit_op1(p, OPCODE_MOV, res1, 0, _bfc1);
-      }
-      
-   if (nr_lights == 0) {
-      release_temps(p);
-      return;
-   }
-
-
-   for (i = 0; i < MAX_LIGHTS; i++) {
-      if (p->state->unit[i].light_enabled) {
-        struct ureg half = undef;
-        struct ureg att = undef, VPpli = undef;
-         
-        count++;
-
-        if (p->state->unit[i].light_eyepos3_is_zero) {
-           /* Can used precomputed constants in this case.
-            * Attenuation never applies to infinite lights.
-            */
-           VPpli = register_param3(p, STATE_LIGHT, i, 
-                                   STATE_POSITION_NORMALIZED); 
-            if (p->state->light_local_viewer) {
-                struct ureg eye_hat = get_eye_position_normalized(p);
-                half = get_temp(p);
-                emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
-                emit_normalize_vec3(p, half, half);
-            } else {
-                half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR);
-            }
-        } 
-        else {
-           struct ureg Ppli = register_param3(p, STATE_LIGHT, i, 
-                                              STATE_POSITION); 
-           struct ureg V = get_eye_position(p);
-           struct ureg dist = get_temp(p);
-           struct ureg tmpPpli = get_temp(p);
-
-           VPpli = get_temp(p); 
-            /* In homogeneous object coordinates
-             */
-            emit_op1(p, OPCODE_RCP, dist, 0, swizzle1(Ppli, W));
-            emit_op2(p, OPCODE_MUL, tmpPpli, 0, Ppli, dist);
-
-           /* Calculate VPpli vector
-            */
-           emit_op2(p, OPCODE_SUB, VPpli, 0, tmpPpli, V); 
-
-            /* we're done with tmpPpli now */
-           release_temp(p, tmpPpli);
-
-           /* Normalize VPpli.  The dist value also used in
-            * attenuation below.
-            */
-           emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli);
-           emit_op1(p, OPCODE_RSQ, dist, 0, dist);
-           emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist);
-
-
-           /* Calculate  attenuation:
-            */ 
-           if (!p->state->unit[i].light_spotcutoff_is_180 ||
-               p->state->unit[i].light_attenuated) {
-              att = calculate_light_attenuation(p, i, VPpli, dist);
-           }
-           
-           /* We're done with dist now */
-           release_temp(p, dist);
-        
-      
-           /* Calculate viewer direction, or use infinite viewer:
-            */
-           half = get_temp(p);
-           if (p->state->light_local_viewer) {
-              struct ureg eye_hat = get_eye_position_normalized(p);
-              emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
-           }
-           else {
-              struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); 
-              emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir);
-           }
-
-           emit_normalize_vec3(p, half, half);
-        }
-
-        /* Calculate dot products:
-         */
-        emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
-        emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
-
-        /* we're done with VPpli and half now, so free them as to not drive up
-           our temp usage unnecessary */
-        release_temp(p, VPpli);
-        release_temp(p, half);
-       
-        /* Front face lighting:
-         */
-        {
-           struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT);
-           struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE);
-           struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR);
-           struct ureg res0, res1;
-           GLuint mask0, mask1;
-
-           emit_op1(p, OPCODE_LIT, lit, 0, dots);
-   
-           if (!is_undef(att)) 
-              emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
-   
-           
-           if (count == nr_lights) {
-              if (separate) {
-                 mask0 = WRITEMASK_XYZ;
-                 mask1 = WRITEMASK_XYZ;
-                 res0 = register_output( p, VERT_RESULT_COL0 );
-                 res1 = register_output( p, VERT_RESULT_COL1 );
-              }
-              else {
-                 mask0 = 0;
-                 mask1 = WRITEMASK_XYZ;
-                 res0 = _col0;
-                 res1 = register_output( p, VERT_RESULT_COL0 );
-              }
-           } else {
-              mask0 = 0;
-              mask1 = 0;
-              res0 = _col0;
-              res1 = _col1;
-           }
-
-           emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0);
-           emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0);
-           emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1);
-      
-           release_temp(p, ambient);
-           release_temp(p, diffuse);
-           release_temp(p, specular);
-        }
-
-        /* Back face lighting:
-         */
-        if (twoside) {
-           struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT);
-           struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE);
-           struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR);
-           struct ureg res0, res1;
-           GLuint mask0, mask1;
-              
-           emit_op1(p, OPCODE_LIT, lit, 0, negate(swizzle(dots,X,Y,W,Z)));
-
-           if (!is_undef(att)) 
-              emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
-
-           if (count == nr_lights) {
-              if (separate) {
-                 mask0 = WRITEMASK_XYZ;
-                 mask1 = WRITEMASK_XYZ;
-                 res0 = register_output( p, VERT_RESULT_BFC0 );
-                 res1 = register_output( p, VERT_RESULT_BFC1 );
-              }
-              else {
-                 mask0 = 0;
-                 mask1 = WRITEMASK_XYZ;
-                 res0 = _bfc0;
-                 res1 = register_output( p, VERT_RESULT_BFC0 );
-              }
-           } else {
-              res0 = _bfc0;
-              res1 = _bfc1;
-              mask0 = 0;
-              mask1 = 0;
-           }
-
-           emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0);
-           emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
-           emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
-
-           release_temp(p, ambient);
-           release_temp(p, diffuse);
-           release_temp(p, specular);
-        }
-
-        release_temp(p, att);
-      }
-   }
-
-   release_temps( p );
-}
-
-
-static void build_fog( struct tnl_program *p )
-{
-   struct ureg fog = register_output(p, VERT_RESULT_FOGC);
-   struct ureg input;
-
-   if (p->state->fog_source_is_depth) {
-      input = swizzle1(get_eye_position(p), Z);
-   }
-   else {
-      input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
-   }
-
-   if (p->state->fog_mode && p->state->tnl_do_vertex_fog) {
-      struct ureg params = register_param2(p, STATE_INTERNAL,
-                                          STATE_FOG_PARAMS_OPTIMIZED);
-      struct ureg tmp = get_temp(p);
-      GLboolean useabs = (p->state->fog_mode != FOG_EXP2);
-
-      if (useabs) {
-        emit_op1(p, OPCODE_ABS, tmp, 0, input);
-      }
-
-      switch (p->state->fog_mode) {
-      case FOG_LINEAR: {
-        struct ureg id = get_identity_param(p);
-        emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input,
-                       swizzle1(params,X), swizzle1(params,Y));
-        emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
-        emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
-        break;
-      }
-      case FOG_EXP:
-        emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input,
-                       swizzle1(params,Z));
-        emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
-        break;
-      case FOG_EXP2:
-        emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
-        emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
-        emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
-        break;
-      }
-
-      release_temp(p, tmp);
-   }
-   else {
-      /* results = incoming fog coords (compute fog per-fragment later) 
-       *
-       * KW:  Is it really necessary to do anything in this case?
-       * BP: Yes, we always need to compute the absolute value, unless
-       * we want to push that down into the fragment program...
-       */
-      GLboolean useabs = GL_TRUE;
-      emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input);
-   }
-}
-static void build_reflect_texgen( struct tnl_program *p,
-                                 struct ureg dest,
-                                 GLuint writemask )
-{
-   struct ureg normal = get_eye_normal(p);
-   struct ureg eye_hat = get_eye_position_normalized(p);
-   struct ureg tmp = get_temp(p);
-
-   /* n.u */
-   emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); 
-   /* 2n.u */
-   emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); 
-   /* (-2n.u)n + u */
-   emit_op3(p, OPCODE_MAD, dest, writemask, negate(tmp), normal, eye_hat);
-
-   release_temp(p, tmp);
-}
-
-static void build_sphere_texgen( struct tnl_program *p,
-                                struct ureg dest,
-                                GLuint writemask )
-{
-   struct ureg normal = get_eye_normal(p);
-   struct ureg eye_hat = get_eye_position_normalized(p);
-   struct ureg tmp = get_temp(p);
-   struct ureg half = register_scalar_const(p, .5);
-   struct ureg r = get_temp(p);
-   struct ureg inv_m = get_temp(p);
-   struct ureg id = get_identity_param(p);
-
-   /* Could share the above calculations, but it would be
-    * a fairly odd state for someone to set (both sphere and
-    * reflection active for different texture coordinate
-    * components.  Of course - if two texture units enable
-    * reflect and/or sphere, things start to tilt in favour
-    * of seperating this out:
-    */
-
-   /* n.u */
-   emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); 
-   /* 2n.u */
-   emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); 
-   /* (-2n.u)n + u */
-   emit_op3(p, OPCODE_MAD, r, 0, negate(tmp), normal, eye_hat); 
-   /* r + 0,0,1 */
-   emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z)); 
-   /* rx^2 + ry^2 + (rz+1)^2 */
-   emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp); 
-   /* 2/m */
-   emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); 
-   /* 1/m */
-   emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half); 
-   /* r/m + 1/2 */
-   emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half); 
-              
-   release_temp(p, tmp);
-   release_temp(p, r);
-   release_temp(p, inv_m);
-}
-
-
-static void build_texture_transform( struct tnl_program *p )
-{
-   GLuint i, j;
-
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-
-      if (!(p->state->fragprog_inputs_read & FRAG_BIT_TEX(i)))
-        continue;
-                                                            
-      if (p->state->unit[i].texgen_enabled || 
-         p->state->unit[i].texmat_enabled) {
-        
-        GLuint texmat_enabled = p->state->unit[i].texmat_enabled;
-        struct ureg out = register_output(p, VERT_RESULT_TEX0 + i);
-        struct ureg out_texgen = undef;
-
-        if (p->state->unit[i].texgen_enabled) {
-           GLuint copy_mask = 0;
-           GLuint sphere_mask = 0;
-           GLuint reflect_mask = 0;
-           GLuint normal_mask = 0;
-           GLuint modes[4];
-        
-           if (texmat_enabled) 
-              out_texgen = get_temp(p);
-           else
-              out_texgen = out;
-
-           modes[0] = p->state->unit[i].texgen_mode0;
-           modes[1] = p->state->unit[i].texgen_mode1;
-           modes[2] = p->state->unit[i].texgen_mode2;
-           modes[3] = p->state->unit[i].texgen_mode3;
-
-           for (j = 0; j < 4; j++) {
-              switch (modes[j]) {
-              case TXG_OBJ_LINEAR: {
-                 struct ureg obj = register_input(p, VERT_ATTRIB_POS);
-                 struct ureg plane = 
-                    register_param3(p, STATE_TEXGEN, i,
-                                    STATE_TEXGEN_OBJECT_S + j);
-
-                 emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, 
-                          obj, plane );
-                 break;
-              }
-              case TXG_EYE_LINEAR: {
-                 struct ureg eye = get_eye_position(p);
-                 struct ureg plane = 
-                    register_param3(p, STATE_TEXGEN, i, 
-                                    STATE_TEXGEN_EYE_S + j);
-
-                 emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, 
-                          eye, plane );
-                 break;
-              }
-              case TXG_SPHERE_MAP: 
-                 sphere_mask |= WRITEMASK_X << j;
-                 break;
-              case TXG_REFLECTION_MAP:
-                 reflect_mask |= WRITEMASK_X << j;
-                 break;
-              case TXG_NORMAL_MAP: 
-                 normal_mask |= WRITEMASK_X << j;
-                 break;
-              case TXG_NONE:
-                 copy_mask |= WRITEMASK_X << j;
-              }
-
-           }
-
-        
-           if (sphere_mask) {
-              build_sphere_texgen(p, out_texgen, sphere_mask);
-           }
-
-           if (reflect_mask) {
-              build_reflect_texgen(p, out_texgen, reflect_mask);
-           }
-
-           if (normal_mask) {
-              struct ureg normal = get_eye_normal(p);
-              emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal );
-           }
-
-           if (copy_mask) {
-              struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i);
-              emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in );
-           }
-        }
-
-        if (texmat_enabled) {
-           struct ureg texmat[4];
-           struct ureg in = (!is_undef(out_texgen) ? 
-                             out_texgen : 
-                             register_input(p, VERT_ATTRIB_TEX0+i));
-           if (PREFER_DP4) {
-              register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
-                                      0, texmat );
-              emit_matrix_transform_vec4( p, out, texmat, in );
-           }
-           else {
-              register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
-                                      STATE_MATRIX_TRANSPOSE, texmat );
-              emit_transpose_matrix_transform_vec4( p, out, texmat, in );
-           }
-        }
-
-        release_temps(p);
-      } 
-      else {
-        emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
-      }
-   }
-}
-
-
-static void build_pointsize( struct tnl_program *p )
-{
-   struct ureg eye = get_eye_position(p);
-   struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
-   struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
-   struct ureg out = register_output(p, VERT_RESULT_PSIZ);
-   struct ureg ut = get_temp(p);
-
-   /* dist = |eyez| */
-   emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z));
-   /* p1 + dist * (p2 + dist * p3); */
-   emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
-               swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
-   emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y),
-               ut, swizzle1(state_attenuation, X));
-
-   /* 1 / sqrt(factor) */
-   emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut );
-
-#if 1
-   /* out = pointSize / sqrt(factor) */
-   emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
-#else
-   /* not sure, might make sense to do clamping here,
-      but it's not done in t_vb_points neither */
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size);
-   emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y));
-   emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
-#endif
-
-   release_temp(p, ut);
-}
-
-static void build_tnl_program( struct tnl_program *p )
-{   /* Emit the program, starting with modelviewproject:
-    */
-   build_hpos(p);
-
-   /* Lighting calculations:
-    */
-   if (p->state->fragprog_inputs_read & (FRAG_BIT_COL0|FRAG_BIT_COL1)) {
-      if (p->state->light_global_enabled)
-        build_lighting(p);
-      else {
-        if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-           emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0);
-
-        if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
-           emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1);
-      }
-   }
-
-   if ((p->state->fragprog_inputs_read & FRAG_BIT_FOGC) ||
-       p->state->fog_mode != FOG_NONE)
-      build_fog(p);
-
-   if (p->state->fragprog_inputs_read & FRAG_BITS_TEX_ANY)
-      build_texture_transform(p);
-
-   if (p->state->point_attenuated)
-      build_pointsize(p);
-
-   /* Finish up:
-    */
-   emit_op1(p, OPCODE_END, undef, 0, undef);
-
-   /* Disassemble:
-    */
-   if (DISASSEM) {
-      _mesa_printf ("\n");
-   }
-}
-
-
-static void
-create_new_program( const struct state_key *key,
-                    struct gl_vertex_program *program,
-                    GLuint max_temps)
-{
-   struct tnl_program p;
-
-   _mesa_memset(&p, 0, sizeof(p));
-   p.state = key;
-   p.program = program;
-   p.eye_position = undef;
-   p.eye_position_normalized = undef;
-   p.eye_normal = undef;
-   p.identity = undef;
-   p.temp_in_use = 0;
-   
-   if (max_temps >= sizeof(int) * 8)
-      p.temp_reserved = 0;
-   else
-      p.temp_reserved = ~((1<<max_temps)-1);
-
-   p.program->Base.Instructions = _mesa_alloc_instructions(MAX_INSN);
-   p.program->Base.String = NULL;
-   p.program->Base.NumInstructions =
-   p.program->Base.NumTemporaries =
-   p.program->Base.NumParameters =
-   p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
-   p.program->Base.Parameters = _mesa_new_parameter_list();
-   p.program->Base.InputsRead = 0;
-   p.program->Base.OutputsWritten = 0;
-
-   build_tnl_program( &p );
-}
-
-
-static struct gl_vertex_program *
-search_cache(struct tnl_cache *cache, GLuint hash,
-             const void *key, GLuint keysize)
-{
-   struct tnl_cache_item *c;
-
-   for (c = cache->items[hash % cache->size]; c; c = c->next) {
-      if (c->hash == hash && _mesa_memcmp(c->key, key, keysize) == 0)
-        return c->prog;
-   }
-
-   return NULL;
-}
-
-
-static void rehash( struct tnl_cache *cache )
-{
-   struct tnl_cache_item **items;
-   struct tnl_cache_item *c, *next;
-   GLuint size, i;
-
-   size = cache->size * 3;
-   items = (struct tnl_cache_item**) _mesa_malloc(size * sizeof(*items));
-   _mesa_memset(items, 0, size * sizeof(*items));
-
-   for (i = 0; i < cache->size; i++)
-      for (c = cache->items[i]; c; c = next) {
-        next = c->next;
-        c->next = items[c->hash % size];
-        items[c->hash % size] = c;
-      }
-
-   FREE(cache->items);
-   cache->items = items;
-   cache->size = size;
-}
-
-static void cache_item( GLcontext *ctx,
-                        struct tnl_cache *cache,
-                       GLuint hash,
-                       void *key,
-                       struct gl_vertex_program *prog )
-{
-   struct tnl_cache_item *c = CALLOC_STRUCT(tnl_cache_item);
-   c->hash = hash;
-   c->key = key;
-
-   c->prog = prog;
-
-   if (++cache->n_items > cache->size * 1.5)
-      rehash(cache);
-
-   c->next = cache->items[hash % cache->size];
-   cache->items[hash % cache->size] = c;
-}
-
-static GLuint hash_key( struct state_key *key )
-{
-   GLuint *ikey = (GLuint *)key;
-   GLuint hash = 0, i;
-
-   /* I'm sure this can be improved on, but speed is important:
-    */
-   for (i = 0; i < sizeof(*key)/sizeof(GLuint); i++)
-      hash ^= ikey[i];
-
-   return hash;
-}
-
 void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx )
 {
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct state_key *key;
-   GLuint hash;
    const struct gl_vertex_program *prev = ctx->VertexProgram._Current;
 
    if (!ctx->VertexProgram._Current ||
        ctx->VertexProgram._Current == ctx->VertexProgram._TnlProgram) {
-      struct gl_vertex_program *newProg;
-
-      /* Grab all the relevent state and put it in a single structure:
-       */
-      key = make_state_key(ctx);
-      hash = hash_key(key);
-
-      /* Look for an already-prepared program for this state:
-       */
-      newProg = search_cache( tnl->vp_cache, hash, key, sizeof(*key));
-   
-      /* OK, we'll have to build a new one:
-       */
-      if (!newProg) {
-
-        if (0)
-           _mesa_printf("Build new TNL program\n");
-        
-        newProg = (struct gl_vertex_program *)
-           ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); 
-
-        create_new_program( key, newProg, ctx->Const.VertexProgram.MaxTemps );
-
-        if (ctx->Driver.ProgramStringNotify)
-           ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, 
-                                             &newProg->Base );
-
-         /* Our ownership of newProg is transferred to the cache */
-        cache_item(ctx, tnl->vp_cache, hash, key, newProg);
-      }
-      else {
-        FREE(key);
-      }
-
-      _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, newProg);
-      _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, newProg);
+      ctx->VertexProgram._Current
+         = ctx->VertexProgram._TnlProgram
+         = _mesa_get_fixed_func_vertex_program(ctx);
    }
 
    /* Tell the driver about the change.  Could define a new target for
@@ -1589,32 +57,3 @@ void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx )
                             (struct gl_program *) ctx->VertexProgram._Current);
    }
 }
-
-void _tnl_ProgramCacheInit( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   tnl->vp_cache = (struct tnl_cache *) MALLOC(sizeof(*tnl->vp_cache));
-   tnl->vp_cache->size = 17;
-   tnl->vp_cache->n_items = 0;
-   tnl->vp_cache->items = (struct tnl_cache_item**)
-      _mesa_calloc(tnl->vp_cache->size * sizeof(*tnl->vp_cache->items));
-}
-
-void _tnl_ProgramCacheDestroy( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_cache_item *c, *next;
-   GLuint i;
-
-   for (i = 0; i < tnl->vp_cache->size; i++)
-      for (c = tnl->vp_cache->items[i]; c; c = next) {
-        next = c->next;
-        FREE(c->key);
-        _mesa_reference_vertprog(ctx, &c->prog, NULL);
-        FREE(c);
-      }
-
-   FREE(tnl->vp_cache->items);
-   FREE(tnl->vp_cache);
-}
index 2ad3abe45aae256a907a60141b3278340b94aa02..d1c147ff032d34e9c8e922c58a34780cb043d226 100644 (file)
  * 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 NONINFRINGEMENT.  IN NO EVENT SHALL
- * TUNGSTEN GRAPHICS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * TUNGSTEN GRAPHICS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
+ * WHETHER IN
  * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  */
 
 
-#ifndef _T_ARB_BUILD_H
-#define _T_ARB_BUILD_H
+#ifndef T_VP_BUILD_H
+#define T_VP_BUILD_H
 
 #include "main/mtypes.h"
 
-extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx );
 
-extern void _tnl_ProgramCacheInit( GLcontext *ctx );
-extern void _tnl_ProgramCacheDestroy( GLcontext *ctx );
+extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx );
 
 #endif