| 1 |
/************************************************************************** |
| 2 |
* |
| 3 |
* Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. |
| 4 |
* All Rights Reserved. |
| 5 |
* |
| 6 |
* Permission is hereby granted, free of charge, to any person obtaining a |
| 7 |
* copy of this software and associated documentation files (the |
| 8 |
* "Software"), to deal in the Software without restriction, including |
| 9 |
* without limitation the rights to use, copy, modify, merge, publish, |
| 10 |
* distribute, sub license, and/or sell copies of the Software, and to |
| 11 |
* permit persons to whom the Software is furnished to do so, subject to |
| 12 |
* the following conditions: |
| 13 |
* |
| 14 |
* The above copyright notice and this permission notice (including the |
| 15 |
* next paragraph) shall be included in all copies or substantial portions |
| 16 |
* of the Software. |
| 17 |
* |
| 18 |
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS |
| 19 |
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
| 20 |
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. |
| 21 |
* IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR |
| 22 |
* ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, |
| 23 |
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE |
| 24 |
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. |
| 25 |
* |
| 26 |
**************************************************************************/ |
| 27 |
|
| 28 |
#include "glheader.h" |
| 29 |
#include "macros.h" |
| 30 |
#include "enums.h" |
| 31 |
#include "texenvprogram.h" |
| 32 |
|
| 33 |
#include "shader/program.h" |
| 34 |
#include "shader/nvfragprog.h" |
| 35 |
#include "shader/arbfragparse.h" |
| 36 |
|
| 37 |
|
| 38 |
#define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM) |
| 39 |
|
| 40 |
struct mode_opt { |
| 41 |
unsigned Source:4; |
| 42 |
unsigned Operand:3; |
| 43 |
}; |
| 44 |
|
| 45 |
struct state_key { |
| 46 |
GLuint enabled_units; |
| 47 |
unsigned separate_specular:1; |
| 48 |
unsigned fog_enabled:1; |
| 49 |
unsigned fog_mode:2; |
| 50 |
|
| 51 |
struct { |
| 52 |
unsigned enabled:1; |
| 53 |
unsigned source_index:3; |
| 54 |
unsigned ScaleShiftRGB:2; |
| 55 |
unsigned ScaleShiftA:2; |
| 56 |
|
| 57 |
unsigned NumArgsRGB:2; |
| 58 |
unsigned ModeRGB:4; |
| 59 |
struct mode_opt OptRGB[3]; |
| 60 |
|
| 61 |
unsigned NumArgsA:2; |
| 62 |
unsigned ModeA:4; |
| 63 |
struct mode_opt OptA[3]; |
| 64 |
} unit[8]; |
| 65 |
}; |
| 66 |
|
| 67 |
#define FOG_LINEAR 0 |
| 68 |
#define FOG_EXP 1 |
| 69 |
#define FOG_EXP2 2 |
| 70 |
#define FOG_UNKNOWN 3 |
| 71 |
|
| 72 |
static GLuint translate_fog_mode( GLenum mode ) |
| 73 |
{ |
| 74 |
switch (mode) { |
| 75 |
case GL_LINEAR: return FOG_LINEAR; |
| 76 |
case GL_EXP: return FOG_EXP; |
| 77 |
case GL_EXP2: return FOG_EXP2; |
| 78 |
default: return FOG_UNKNOWN; |
| 79 |
} |
| 80 |
} |
| 81 |
|
| 82 |
#define OPR_SRC_COLOR 0 |
| 83 |
#define OPR_ONE_MINUS_SRC_COLOR 1 |
| 84 |
#define OPR_SRC_ALPHA 2 |
| 85 |
#define OPR_ONE_MINUS_SRC_ALPHA 3 |
| 86 |
#define OPR_ZERO 4 |
| 87 |
#define OPR_ONE 5 |
| 88 |
#define OPR_UNKNOWN 7 |
| 89 |
|
| 90 |
static GLuint translate_operand( GLenum operand ) |
| 91 |
{ |
| 92 |
switch (operand) { |
| 93 |
case GL_SRC_COLOR: return OPR_SRC_COLOR; |
| 94 |
case GL_ONE_MINUS_SRC_COLOR: return OPR_ONE_MINUS_SRC_COLOR; |
| 95 |
case GL_SRC_ALPHA: return OPR_SRC_ALPHA; |
| 96 |
case GL_ONE_MINUS_SRC_ALPHA: return OPR_ONE_MINUS_SRC_ALPHA; |
| 97 |
case GL_ZERO: return OPR_ZERO; |
| 98 |
case GL_ONE: return OPR_ONE; |
| 99 |
default: return OPR_UNKNOWN; |
| 100 |
} |
| 101 |
} |
| 102 |
|
| 103 |
#define SRC_TEXTURE 0 |
| 104 |
#define SRC_TEXTURE0 1 |
| 105 |
#define SRC_TEXTURE1 2 |
| 106 |
#define SRC_TEXTURE2 3 |
| 107 |
#define SRC_TEXTURE3 4 |
| 108 |
#define SRC_TEXTURE4 5 |
| 109 |
#define SRC_TEXTURE5 6 |
| 110 |
#define SRC_TEXTURE6 7 |
| 111 |
#define SRC_TEXTURE7 8 |
| 112 |
#define SRC_CONSTANT 9 |
| 113 |
#define SRC_PRIMARY_COLOR 10 |
| 114 |
#define SRC_PREVIOUS 11 |
| 115 |
#define SRC_UNKNOWN 15 |
| 116 |
|
| 117 |
static GLuint translate_source( GLenum src ) |
| 118 |
{ |
| 119 |
switch (src) { |
| 120 |
case GL_TEXTURE: return SRC_TEXTURE; |
| 121 |
case GL_TEXTURE0: |
| 122 |
case GL_TEXTURE1: |
| 123 |
case GL_TEXTURE2: |
| 124 |
case GL_TEXTURE3: |
| 125 |
case GL_TEXTURE4: |
| 126 |
case GL_TEXTURE5: |
| 127 |
case GL_TEXTURE6: |
| 128 |
case GL_TEXTURE7: return SRC_TEXTURE0 + (src - GL_TEXTURE0); |
| 129 |
case GL_CONSTANT: return SRC_CONSTANT; |
| 130 |
case GL_PRIMARY_COLOR: return SRC_PRIMARY_COLOR; |
| 131 |
case GL_PREVIOUS: return SRC_PREVIOUS; |
| 132 |
default: return SRC_UNKNOWN; |
| 133 |
} |
| 134 |
} |
| 135 |
|
| 136 |
#define MODE_REPLACE 0 |
| 137 |
#define MODE_MODULATE 1 |
| 138 |
#define MODE_ADD 2 |
| 139 |
#define MODE_ADD_SIGNED 3 |
| 140 |
#define MODE_INTERPOLATE 4 |
| 141 |
#define MODE_SUBTRACT 5 |
| 142 |
#define MODE_DOT3_RGB 6 |
| 143 |
#define MODE_DOT3_RGB_EXT 7 |
| 144 |
#define MODE_DOT3_RGBA 8 |
| 145 |
#define MODE_DOT3_RGBA_EXT 9 |
| 146 |
#define MODE_MODULATE_ADD_ATI 10 |
| 147 |
#define MODE_MODULATE_SIGNED_ADD_ATI 11 |
| 148 |
#define MODE_MODULATE_SUBTRACT_ATI 12 |
| 149 |
#define MODE_UNKNOWN 15 |
| 150 |
|
| 151 |
static GLuint translate_mode( GLenum mode ) |
| 152 |
{ |
| 153 |
switch (mode) { |
| 154 |
case GL_REPLACE: return MODE_REPLACE; |
| 155 |
case GL_MODULATE: return MODE_MODULATE; |
| 156 |
case GL_ADD: return MODE_ADD; |
| 157 |
case GL_ADD_SIGNED: return MODE_ADD_SIGNED; |
| 158 |
case GL_INTERPOLATE: return MODE_INTERPOLATE; |
| 159 |
case GL_SUBTRACT: return MODE_SUBTRACT; |
| 160 |
case GL_DOT3_RGB: return MODE_DOT3_RGB; |
| 161 |
case GL_DOT3_RGB_EXT: return MODE_DOT3_RGB_EXT; |
| 162 |
case GL_DOT3_RGBA: return MODE_DOT3_RGBA; |
| 163 |
case GL_DOT3_RGBA_EXT: return MODE_DOT3_RGBA_EXT; |
| 164 |
case GL_MODULATE_ADD_ATI: return MODE_MODULATE_ADD_ATI; |
| 165 |
case GL_MODULATE_SIGNED_ADD_ATI: return MODE_MODULATE_SIGNED_ADD_ATI; |
| 166 |
case GL_MODULATE_SUBTRACT_ATI: return MODE_MODULATE_SUBTRACT_ATI; |
| 167 |
default: return MODE_UNKNOWN; |
| 168 |
} |
| 169 |
} |
| 170 |
|
| 171 |
#define TEXTURE_UNKNOWN_INDEX 7 |
| 172 |
static GLuint translate_tex_src_bit( GLuint bit ) |
| 173 |
{ |
| 174 |
switch (bit) { |
| 175 |
case TEXTURE_1D_BIT: return TEXTURE_1D_INDEX; |
| 176 |
case TEXTURE_2D_BIT: return TEXTURE_2D_INDEX; |
| 177 |
case TEXTURE_RECT_BIT: return TEXTURE_RECT_INDEX; |
| 178 |
case TEXTURE_3D_BIT: return TEXTURE_3D_INDEX; |
| 179 |
case TEXTURE_CUBE_BIT: return TEXTURE_CUBE_INDEX; |
| 180 |
default: return TEXTURE_UNKNOWN_INDEX; |
| 181 |
} |
| 182 |
} |
| 183 |
|
| 184 |
static struct state_key *make_state_key( GLcontext *ctx ) |
| 185 |
{ |
| 186 |
struct state_key *key = CALLOC_STRUCT(state_key); |
| 187 |
GLuint i, j; |
| 188 |
|
| 189 |
for (i=0;i<MAX_TEXTURE_UNITS;i++) { |
| 190 |
struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; |
| 191 |
|
| 192 |
if (!texUnit->_ReallyEnabled) |
| 193 |
continue; |
| 194 |
|
| 195 |
key->unit[i].enabled = 1; |
| 196 |
key->enabled_units |= (1<<i); |
| 197 |
|
| 198 |
key->unit[i].source_index = |
| 199 |
translate_tex_src_bit(texUnit->_ReallyEnabled); |
| 200 |
|
| 201 |
key->unit[i].NumArgsRGB = texUnit->_CurrentCombine->_NumArgsRGB; |
| 202 |
key->unit[i].NumArgsA = texUnit->_CurrentCombine->_NumArgsA; |
| 203 |
|
| 204 |
key->unit[i].ModeRGB = |
| 205 |
translate_mode(texUnit->_CurrentCombine->ModeRGB); |
| 206 |
key->unit[i].ModeA = |
| 207 |
translate_mode(texUnit->_CurrentCombine->ModeA); |
| 208 |
|
| 209 |
key->unit[i].ScaleShiftRGB = texUnit->_CurrentCombine->ScaleShiftRGB; |
| 210 |
key->unit[i].ScaleShiftA = texUnit->_CurrentCombine->ScaleShiftRGB; |
| 211 |
|
| 212 |
for (j=0;j<3;j++) { |
| 213 |
key->unit[i].OptRGB[j].Operand = |
| 214 |
translate_operand(texUnit->_CurrentCombine->OperandRGB[j]); |
| 215 |
key->unit[i].OptA[j].Operand = |
| 216 |
translate_operand(texUnit->_CurrentCombine->OperandA[j]); |
| 217 |
key->unit[i].OptRGB[j].Source = |
| 218 |
translate_source(texUnit->_CurrentCombine->SourceRGB[j]); |
| 219 |
key->unit[i].OptA[j].Source = |
| 220 |
translate_source(texUnit->_CurrentCombine->SourceA[j]); |
| 221 |
} |
| 222 |
} |
| 223 |
|
| 224 |
if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) |
| 225 |
key->separate_specular = 1; |
| 226 |
|
| 227 |
if (ctx->Fog.Enabled) { |
| 228 |
key->fog_enabled = 1; |
| 229 |
key->fog_mode = translate_fog_mode(ctx->Fog.Mode); |
| 230 |
} |
| 231 |
|
| 232 |
return key; |
| 233 |
} |
| 234 |
|
| 235 |
/* Use uregs to represent registers internally, translate to Mesa's |
| 236 |
* expected formats on emit. |
| 237 |
* |
| 238 |
* NOTE: These are passed by value extensively in this file rather |
| 239 |
* than as usual by pointer reference. If this disturbs you, try |
| 240 |
* remembering they are just 32bits in size. |
| 241 |
* |
| 242 |
* GCC is smart enough to deal with these dword-sized structures in |
| 243 |
* much the same way as if I had defined them as dwords and was using |
| 244 |
* macros to access and set the fields. This is much nicer and easier |
| 245 |
* to evolve. |
| 246 |
*/ |
| 247 |
struct ureg { |
| 248 |
GLuint file:4; |
| 249 |
GLuint idx:8; |
| 250 |
GLuint negatebase:1; |
| 251 |
GLuint abs:1; |
| 252 |
GLuint negateabs:1; |
| 253 |
GLuint swz:12; |
| 254 |
GLuint pad:5; |
| 255 |
}; |
| 256 |
|
| 257 |
const static struct ureg undef = { |
| 258 |
~0, |
| 259 |
~0, |
| 260 |
0, |
| 261 |
0, |
| 262 |
0, |
| 263 |
0, |
| 264 |
0 |
| 265 |
}; |
| 266 |
|
| 267 |
#define X 0 |
| 268 |
#define Y 1 |
| 269 |
#define Z 2 |
| 270 |
#define W 3 |
| 271 |
|
| 272 |
/* State used to build the fragment program: |
| 273 |
*/ |
| 274 |
struct texenv_fragment_program { |
| 275 |
struct fragment_program *program; |
| 276 |
GLcontext *ctx; |
| 277 |
struct state_key *state; |
| 278 |
|
| 279 |
GLuint alu_temps; /* Track texture indirections, see spec. */ |
| 280 |
GLuint temps_output; /* Track texture indirections, see spec. */ |
| 281 |
|
| 282 |
GLuint temp_in_use; /* Tracks temporary regs which are in |
| 283 |
* use. |
| 284 |
*/ |
| 285 |
|
| 286 |
|
| 287 |
GLboolean error; |
| 288 |
|
| 289 |
struct ureg src_texture[MAX_TEXTURE_UNITS]; |
| 290 |
/* Reg containing each texture unit's sampled texture color, |
| 291 |
* else undef. |
| 292 |
*/ |
| 293 |
|
| 294 |
struct ureg src_previous; /* Reg containing color from previous |
| 295 |
* stage. May need to be decl'd. |
| 296 |
*/ |
| 297 |
|
| 298 |
GLuint last_tex_stage; /* Number of last enabled texture unit */ |
| 299 |
|
| 300 |
struct ureg half; |
| 301 |
struct ureg one; |
| 302 |
struct ureg zero; |
| 303 |
}; |
| 304 |
|
| 305 |
|
| 306 |
|
| 307 |
static struct ureg make_ureg(GLuint file, GLuint idx) |
| 308 |
{ |
| 309 |
struct ureg reg; |
| 310 |
reg.file = file; |
| 311 |
reg.idx = idx; |
| 312 |
reg.negatebase = 0; |
| 313 |
reg.abs = 0; |
| 314 |
reg.negateabs = 0; |
| 315 |
reg.swz = SWIZZLE_NOOP; |
| 316 |
reg.pad = 0; |
| 317 |
return reg; |
| 318 |
} |
| 319 |
|
| 320 |
static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w ) |
| 321 |
{ |
| 322 |
reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x), |
| 323 |
GET_SWZ(reg.swz, y), |
| 324 |
GET_SWZ(reg.swz, z), |
| 325 |
GET_SWZ(reg.swz, w)); |
| 326 |
|
| 327 |
return reg; |
| 328 |
} |
| 329 |
|
| 330 |
static struct ureg swizzle1( struct ureg reg, int x ) |
| 331 |
{ |
| 332 |
return swizzle(reg, x, x, x, x); |
| 333 |
} |
| 334 |
|
| 335 |
static struct ureg negate( struct ureg reg ) |
| 336 |
{ |
| 337 |
reg.negatebase ^= 1; |
| 338 |
return reg; |
| 339 |
} |
| 340 |
|
| 341 |
static GLboolean is_undef( struct ureg reg ) |
| 342 |
{ |
| 343 |
return reg.file == 0xf; |
| 344 |
} |
| 345 |
|
| 346 |
|
| 347 |
static struct ureg get_temp( struct texenv_fragment_program *p ) |
| 348 |
{ |
| 349 |
int bit; |
| 350 |
|
| 351 |
/* First try and reuse temps which have been used already: |
| 352 |
*/ |
| 353 |
bit = ffs( ~p->temp_in_use & p->alu_temps ); |
| 354 |
|
| 355 |
/* Then any unused temporary: |
| 356 |
*/ |
| 357 |
if (!bit) |
| 358 |
bit = ffs( ~p->temp_in_use ); |
| 359 |
|
| 360 |
if (!bit) { |
| 361 |
_mesa_problem(NULL, "%s: out of temporaries\n", __FILE__); |
| 362 |
exit(1); |
| 363 |
} |
| 364 |
|
| 365 |
p->temp_in_use |= 1<<(bit-1); |
| 366 |
return make_ureg(PROGRAM_TEMPORARY, (bit-1)); |
| 367 |
} |
| 368 |
|
| 369 |
static struct ureg get_tex_temp( struct texenv_fragment_program *p ) |
| 370 |
{ |
| 371 |
int bit; |
| 372 |
|
| 373 |
/* First try to find availble temp not previously used (to avoid |
| 374 |
* starting a new texture indirection). According to the spec, the |
| 375 |
* ~p->temps_output isn't necessary, but will keep it there for |
| 376 |
* now: |
| 377 |
*/ |
| 378 |
bit = ffs( ~p->temp_in_use & ~p->alu_temps & ~p->temps_output ); |
| 379 |
|
| 380 |
/* Then any unused temporary: |
| 381 |
*/ |
| 382 |
if (!bit) |
| 383 |
bit = ffs( ~p->temp_in_use ); |
| 384 |
|
| 385 |
if (!bit) { |
| 386 |
_mesa_problem(NULL, "%s: out of temporaries\n", __FILE__); |
| 387 |
exit(1); |
| 388 |
} |
| 389 |
|
| 390 |
p->temp_in_use |= 1<<(bit-1); |
| 391 |
return make_ureg(PROGRAM_TEMPORARY, (bit-1)); |
| 392 |
} |
| 393 |
|
| 394 |
|
| 395 |
static void release_temps( struct texenv_fragment_program *p ) |
| 396 |
{ |
| 397 |
GLuint max_temp = p->ctx->Const.MaxFragmentProgramTemps; |
| 398 |
|
| 399 |
/* KW: To support tex_env_crossbar, don't release the registers in |
| 400 |
* temps_output. |
| 401 |
*/ |
| 402 |
if (max_temp >= sizeof(int) * 8) |
| 403 |
p->temp_in_use = p->temps_output; |
| 404 |
else |
| 405 |
p->temp_in_use = ~((1<<max_temp)-1) | p->temps_output; |
| 406 |
} |
| 407 |
|
| 408 |
|
| 409 |
static struct ureg register_param6( struct texenv_fragment_program *p, |
| 410 |
GLint s0, |
| 411 |
GLint s1, |
| 412 |
GLint s2, |
| 413 |
GLint s3, |
| 414 |
GLint s4, |
| 415 |
GLint s5) |
| 416 |
{ |
| 417 |
GLint tokens[6]; |
| 418 |
GLuint idx; |
| 419 |
tokens[0] = s0; |
| 420 |
tokens[1] = s1; |
| 421 |
tokens[2] = s2; |
| 422 |
tokens[3] = s3; |
| 423 |
tokens[4] = s4; |
| 424 |
tokens[5] = s5; |
| 425 |
idx = _mesa_add_state_reference( p->program->Parameters, tokens ); |
| 426 |
return make_ureg(PROGRAM_STATE_VAR, idx); |
| 427 |
} |
| 428 |
|
| 429 |
|
| 430 |
#define register_param1(p,s0) register_param6(p,s0,0,0,0,0,0) |
| 431 |
#define register_param2(p,s0,s1) register_param6(p,s0,s1,0,0,0,0) |
| 432 |
#define register_param3(p,s0,s1,s2) register_param6(p,s0,s1,s2,0,0,0) |
| 433 |
#define register_param4(p,s0,s1,s2,s3) register_param6(p,s0,s1,s2,s3,0,0) |
| 434 |
|
| 435 |
|
| 436 |
static struct ureg register_input( struct texenv_fragment_program *p, GLuint input ) |
| 437 |
{ |
| 438 |
p->program->InputsRead |= (1<<input); |
| 439 |
return make_ureg(PROGRAM_INPUT, input); |
| 440 |
} |
| 441 |
|
| 442 |
|
| 443 |
static void emit_arg( struct fp_src_register *reg, |
| 444 |
struct ureg ureg ) |
| 445 |
{ |
| 446 |
reg->File = ureg.file; |
| 447 |
reg->Index = ureg.idx; |
| 448 |
reg->Swizzle = ureg.swz; |
| 449 |
reg->NegateBase = ureg.negatebase; |
| 450 |
reg->Abs = ureg.abs; |
| 451 |
reg->NegateAbs = ureg.negateabs; |
| 452 |
} |
| 453 |
|
| 454 |
static void emit_dst( struct fp_dst_register *dst, |
| 455 |
struct ureg ureg, GLuint mask ) |
| 456 |
{ |
| 457 |
dst->File = ureg.file; |
| 458 |
dst->Index = ureg.idx; |
| 459 |
dst->WriteMask = mask; |
| 460 |
dst->CondMask = 0; |
| 461 |
dst->CondSwizzle = 0; |
| 462 |
} |
| 463 |
|
| 464 |
static struct fp_instruction * |
| 465 |
emit_op(struct texenv_fragment_program *p, |
| 466 |
GLuint op, |
| 467 |
struct ureg dest, |
| 468 |
GLuint mask, |
| 469 |
GLuint saturate, |
| 470 |
struct ureg src0, |
| 471 |
struct ureg src1, |
| 472 |
struct ureg src2 ) |
| 473 |
{ |
| 474 |
GLuint nr = p->program->Base.NumInstructions++; |
| 475 |
struct fp_instruction *inst = &p->program->Instructions[nr]; |
| 476 |
|
| 477 |
memset(inst, 0, sizeof(*inst)); |
| 478 |
inst->Opcode = op; |
| 479 |
|
| 480 |
emit_arg( &inst->SrcReg[0], src0 ); |
| 481 |
emit_arg( &inst->SrcReg[1], src1 ); |
| 482 |
emit_arg( &inst->SrcReg[2], src2 ); |
| 483 |
|
| 484 |
inst->Saturate = saturate; |
| 485 |
|
| 486 |
emit_dst( &inst->DstReg, dest, mask ); |
| 487 |
|
| 488 |
/* Accounting for indirection tracking: |
| 489 |
*/ |
| 490 |
if (dest.file == PROGRAM_TEMPORARY) |
| 491 |
p->temps_output |= 1 << dest.idx; |
| 492 |
|
| 493 |
return inst; |
| 494 |
} |
| 495 |
|
| 496 |
|
| 497 |
static struct ureg emit_arith( struct texenv_fragment_program *p, |
| 498 |
GLuint op, |
| 499 |
struct ureg dest, |
| 500 |
GLuint mask, |
| 501 |
GLuint saturate, |
| 502 |
struct ureg src0, |
| 503 |
struct ureg src1, |
| 504 |
struct ureg src2 ) |
| 505 |
{ |
| 506 |
emit_op(p, op, dest, mask, saturate, src0, src1, src2); |
| 507 |
|
| 508 |
/* Accounting for indirection tracking: |
| 509 |
*/ |
| 510 |
if (src0.file == PROGRAM_TEMPORARY) |
| 511 |
p->alu_temps |= 1 << src0.idx; |
| 512 |
|
| 513 |
if (!is_undef(src1) && src1.file == PROGRAM_TEMPORARY) |
| 514 |
p->alu_temps |= 1 << src1.idx; |
| 515 |
|
| 516 |
if (!is_undef(src2) && src2.file == PROGRAM_TEMPORARY) |
| 517 |
p->alu_temps |= 1 << src2.idx; |
| 518 |
|
| 519 |
if (dest.file == PROGRAM_TEMPORARY) |
| 520 |
p->alu_temps |= 1 << dest.idx; |
| 521 |
|
| 522 |
p->program->NumAluInstructions++; |
| 523 |
return dest; |
| 524 |
} |
| 525 |
|
| 526 |
static struct ureg emit_texld( struct texenv_fragment_program *p, |
| 527 |
GLuint op, |
| 528 |
struct ureg dest, |
| 529 |
GLuint destmask, |
| 530 |
GLuint tex_unit, |
| 531 |
GLuint tex_idx, |
| 532 |
struct ureg coord ) |
| 533 |
{ |
| 534 |
struct fp_instruction *inst = emit_op( p, op, |
| 535 |
dest, destmask, |
| 536 |
0, /* don't saturate? */ |
| 537 |
coord, /* arg 0? */ |
| 538 |
undef, |
| 539 |
undef); |
| 540 |
|
| 541 |
inst->TexSrcIdx = tex_idx; |
| 542 |
inst->TexSrcUnit = tex_unit; |
| 543 |
|
| 544 |
p->program->NumTexInstructions++; |
| 545 |
|
| 546 |
/* Is this a texture indirection? |
| 547 |
*/ |
| 548 |
if ((coord.file == PROGRAM_TEMPORARY && |
| 549 |
(p->temps_output & (1<<coord.idx))) || |
| 550 |
(dest.file == PROGRAM_TEMPORARY && |
| 551 |
(p->alu_temps & (1<<dest.idx)))) { |
| 552 |
p->program->NumTexIndirections++; |
| 553 |
p->temps_output = 1<<coord.idx; |
| 554 |
p->alu_temps = 0; |
| 555 |
assert(0); /* KW: texture env crossbar */ |
| 556 |
} |
| 557 |
|
| 558 |
return dest; |
| 559 |
} |
| 560 |
|
| 561 |
|
| 562 |
static struct ureg register_const4f( struct texenv_fragment_program *p, |
| 563 |
GLfloat s0, |
| 564 |
GLfloat s1, |
| 565 |
GLfloat s2, |
| 566 |
GLfloat s3) |
| 567 |
{ |
| 568 |
GLfloat values[4]; |
| 569 |
GLuint idx; |
| 570 |
values[0] = s0; |
| 571 |
values[1] = s1; |
| 572 |
values[2] = s2; |
| 573 |
values[3] = s3; |
| 574 |
idx = _mesa_add_unnamed_constant( p->program->Parameters, values ); |
| 575 |
return make_ureg(PROGRAM_STATE_VAR, idx); |
| 576 |
} |
| 577 |
|
| 578 |
#define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0) |
| 579 |
#define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1) |
| 580 |
#define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1) |
| 581 |
#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1) |
| 582 |
|
| 583 |
|
| 584 |
|
| 585 |
|
| 586 |
static struct ureg get_one( struct texenv_fragment_program *p ) |
| 587 |
{ |
| 588 |
if (is_undef(p->one)) |
| 589 |
p->one = register_scalar_const(p, 1.0); |
| 590 |
return p->one; |
| 591 |
} |
| 592 |
|
| 593 |
static struct ureg get_half( struct texenv_fragment_program *p ) |
| 594 |
{ |
| 595 |
if (is_undef(p->half)) |
| 596 |
p->one = register_scalar_const(p, 0.5); |
| 597 |
return p->half; |
| 598 |
} |
| 599 |
|
| 600 |
static struct ureg get_zero( struct texenv_fragment_program *p ) |
| 601 |
{ |
| 602 |
if (is_undef(p->zero)) |
| 603 |
p->one = register_scalar_const(p, 0.0); |
| 604 |
return p->zero; |
| 605 |
} |
| 606 |
|
| 607 |
|
| 608 |
|
| 609 |
|
| 610 |
|
| 611 |
static void program_error( struct texenv_fragment_program *p, const char *msg ) |
| 612 |
{ |
| 613 |
_mesa_problem(NULL, msg); |
| 614 |
p->error = 1; |
| 615 |
} |
| 616 |
|
| 617 |
static struct ureg get_source( struct texenv_fragment_program *p, |
| 618 |
GLuint src, GLuint unit ) |
| 619 |
{ |
| 620 |
switch (src) { |
| 621 |
case SRC_TEXTURE: |
| 622 |
assert(!is_undef(p->src_texture[unit])); |
| 623 |
return p->src_texture[unit]; |
| 624 |
|
| 625 |
case SRC_TEXTURE0: |
| 626 |
case SRC_TEXTURE1: |
| 627 |
case SRC_TEXTURE2: |
| 628 |
case SRC_TEXTURE3: |
| 629 |
case SRC_TEXTURE4: |
| 630 |
case SRC_TEXTURE5: |
| 631 |
case SRC_TEXTURE6: |
| 632 |
case SRC_TEXTURE7: |
| 633 |
assert(!is_undef(p->src_texture[src - SRC_TEXTURE0])); |
| 634 |
return p->src_texture[src - SRC_TEXTURE0]; |
| 635 |
|
| 636 |
case SRC_CONSTANT: |
| 637 |
return register_param2(p, STATE_TEXENV_COLOR, unit); |
| 638 |
|
| 639 |
case SRC_PRIMARY_COLOR: |
| 640 |
return register_input(p, FRAG_ATTRIB_COL0); |
| 641 |
|
| 642 |
case SRC_PREVIOUS: |
| 643 |
default: |
| 644 |
if (is_undef(p->src_previous)) |
| 645 |
return register_input(p, FRAG_ATTRIB_COL0); |
| 646 |
else |
| 647 |
return p->src_previous; |
| 648 |
} |
| 649 |
} |
| 650 |
|
| 651 |
static struct ureg emit_combine_source( struct texenv_fragment_program *p, |
| 652 |
GLuint mask, |
| 653 |
GLuint unit, |
| 654 |
GLuint source, |
| 655 |
GLuint operand ) |
| 656 |
{ |
| 657 |
struct ureg arg, src, one; |
| 658 |
|
| 659 |
src = get_source(p, source, unit); |
| 660 |
|
| 661 |
switch (operand) { |
| 662 |
case OPR_ONE_MINUS_SRC_COLOR: |
| 663 |
/* Get unused tmp, |
| 664 |
* Emit tmp = 1.0 - arg.xyzw |
| 665 |
*/ |
| 666 |
arg = get_temp( p ); |
| 667 |
one = get_one( p ); |
| 668 |
return emit_arith( p, FP_OPCODE_SUB, arg, mask, 0, one, src, undef); |
| 669 |
|
| 670 |
case OPR_SRC_ALPHA: |
| 671 |
if (mask == WRITEMASK_W) |
| 672 |
return src; |
| 673 |
else |
| 674 |
return swizzle1( src, W ); |
| 675 |
case OPR_ONE_MINUS_SRC_ALPHA: |
| 676 |
/* Get unused tmp, |
| 677 |
* Emit tmp = 1.0 - arg.wwww |
| 678 |
*/ |
| 679 |
arg = get_temp(p); |
| 680 |
one = get_one(p); |
| 681 |
return emit_arith(p, FP_OPCODE_SUB, arg, mask, 0, |
| 682 |
one, swizzle1(src, W), undef); |
| 683 |
case OPR_ZERO: |
| 684 |
return get_zero(p); |
| 685 |
case OPR_ONE: |
| 686 |
return get_one(p); |
| 687 |
case OPR_SRC_COLOR: |
| 688 |
default: |
| 689 |
return src; |
| 690 |
} |
| 691 |
} |
| 692 |
|
| 693 |
static GLboolean args_match( struct state_key *key, GLuint unit ) |
| 694 |
{ |
| 695 |
int i, nr = key->unit[unit].NumArgsRGB; |
| 696 |
|
| 697 |
for (i = 0 ; i < nr ; i++) { |
| 698 |
if (key->unit[unit].OptA[i].Source != key->unit[unit].OptRGB[i].Source) |
| 699 |
return GL_FALSE; |
| 700 |
|
| 701 |
switch(key->unit[unit].OptA[i].Operand) { |
| 702 |
case OPR_SRC_ALPHA: |
| 703 |
switch(key->unit[unit].OptRGB[i].Operand) { |
| 704 |
case OPR_SRC_COLOR: |
| 705 |
case OPR_SRC_ALPHA: |
| 706 |
break; |
| 707 |
default: |
| 708 |
return GL_FALSE; |
| 709 |
} |
| 710 |
break; |
| 711 |
case OPR_ONE_MINUS_SRC_ALPHA: |
| 712 |
switch(key->unit[unit].OptRGB[i].Operand) { |
| 713 |
case OPR_ONE_MINUS_SRC_COLOR: |
| 714 |
case OPR_ONE_MINUS_SRC_ALPHA: |
| 715 |
break; |
| 716 |
default: |
| 717 |
return GL_FALSE; |
| 718 |
} |
| 719 |
break; |
| 720 |
default: |
| 721 |
return GL_FALSE; /* impossible */ |
| 722 |
} |
| 723 |
} |
| 724 |
|
| 725 |
return GL_TRUE; |
| 726 |
} |
| 727 |
|
| 728 |
static struct ureg emit_combine( struct texenv_fragment_program *p, |
| 729 |
struct ureg dest, |
| 730 |
GLuint mask, |
| 731 |
GLuint saturate, |
| 732 |
GLuint unit, |
| 733 |
GLuint nr, |
| 734 |
GLuint mode, |
| 735 |
struct mode_opt *opt) |
| 736 |
{ |
| 737 |
struct ureg src[3]; |
| 738 |
struct ureg tmp, half; |
| 739 |
int i; |
| 740 |
|
| 741 |
tmp = undef; /* silence warning (bug 5318) */ |
| 742 |
|
| 743 |
for (i = 0; i < nr; i++) |
| 744 |
src[i] = emit_combine_source( p, mask, unit, opt[i].Source, opt[i].Operand ); |
| 745 |
|
| 746 |
switch (mode) { |
| 747 |
case MODE_REPLACE: |
| 748 |
if (mask == WRITEMASK_XYZW && !saturate) |
| 749 |
return src[0]; |
| 750 |
else |
| 751 |
return emit_arith( p, FP_OPCODE_MOV, dest, mask, saturate, src[0], undef, undef ); |
| 752 |
case MODE_MODULATE: |
| 753 |
return emit_arith( p, FP_OPCODE_MUL, dest, mask, saturate, |
| 754 |
src[0], src[1], undef ); |
| 755 |
case MODE_ADD: |
| 756 |
return emit_arith( p, FP_OPCODE_ADD, dest, mask, saturate, |
| 757 |
src[0], src[1], undef ); |
| 758 |
case MODE_ADD_SIGNED: |
| 759 |
/* tmp = arg0 + arg1 |
| 760 |
* result = tmp - .5 |
| 761 |
*/ |
| 762 |
half = get_half(p); |
| 763 |
emit_arith( p, FP_OPCODE_ADD, tmp, mask, 0, src[0], src[1], undef ); |
| 764 |
emit_arith( p, FP_OPCODE_SUB, dest, mask, saturate, tmp, half, undef ); |
| 765 |
return dest; |
| 766 |
case MODE_INTERPOLATE: |
| 767 |
/* Arg0 * (Arg2) + Arg1 * (1-Arg2) -- note arguments are reordered: |
| 768 |
*/ |
| 769 |
return emit_arith( p, FP_OPCODE_LRP, dest, mask, saturate, src[2], src[0], src[1] ); |
| 770 |
|
| 771 |
case MODE_SUBTRACT: |
| 772 |
return emit_arith( p, FP_OPCODE_SUB, dest, mask, saturate, src[0], src[1], undef ); |
| 773 |
|
| 774 |
case MODE_DOT3_RGBA: |
| 775 |
case MODE_DOT3_RGBA_EXT: |
| 776 |
case MODE_DOT3_RGB_EXT: |
| 777 |
case MODE_DOT3_RGB: { |
| 778 |
struct ureg tmp0 = get_temp( p ); |
| 779 |
struct ureg tmp1 = get_temp( p ); |
| 780 |
struct ureg neg1 = register_scalar_const(p, -1); |
| 781 |
struct ureg two = register_scalar_const(p, 2); |
| 782 |
|
| 783 |
/* tmp0 = 2*src0 - 1 |
| 784 |
* tmp1 = 2*src1 - 1 |
| 785 |
* |
| 786 |
* dst = tmp0 dot3 tmp1 |
| 787 |
*/ |
| 788 |
emit_arith( p, FP_OPCODE_MAD, tmp0, WRITEMASK_XYZW, 0, |
| 789 |
two, src[0], neg1); |
| 790 |
|
| 791 |
if (memcmp(&src[0], &src[1], sizeof(struct ureg)) == 0) |
| 792 |
tmp1 = tmp0; |
| 793 |
else |
| 794 |
emit_arith( p, FP_OPCODE_MAD, tmp1, WRITEMASK_XYZW, 0, |
| 795 |
two, src[1], neg1); |
| 796 |
emit_arith( p, FP_OPCODE_DP3, dest, mask, saturate, tmp0, tmp1, undef); |
| 797 |
return dest; |
| 798 |
} |
| 799 |
case MODE_MODULATE_ADD_ATI: |
| 800 |
/* Arg0 * Arg2 + Arg1 */ |
| 801 |
return emit_arith( p, FP_OPCODE_MAD, dest, mask, saturate, |
| 802 |
src[0], src[2], src[1] ); |
| 803 |
case MODE_MODULATE_SIGNED_ADD_ATI: { |
| 804 |
/* Arg0 * Arg2 + Arg1 - 0.5 */ |
| 805 |
struct ureg tmp0 = get_temp(p); |
| 806 |
half = get_half(p); |
| 807 |
emit_arith( p, FP_OPCODE_MAD, tmp0, mask, 0, src[0], src[2], src[1] ); |
| 808 |
emit_arith( p, FP_OPCODE_SUB, dest, mask, saturate, tmp0, half, undef ); |
| 809 |
return dest; |
| 810 |
} |
| 811 |
case MODE_MODULATE_SUBTRACT_ATI: |
| 812 |
/* Arg0 * Arg2 - Arg1 */ |
| 813 |
emit_arith( p, FP_OPCODE_MAD, dest, mask, 0, src[0], src[2], negate(src[1]) ); |
| 814 |
return dest; |
| 815 |
default: |
| 816 |
return src[0]; |
| 817 |
} |
| 818 |
} |
| 819 |
|
| 820 |
|
| 821 |
static struct ureg emit_texenv( struct texenv_fragment_program *p, int unit ) |
| 822 |
{ |
| 823 |
struct state_key *key = p->state; |
| 824 |
GLuint saturate = (unit < p->last_tex_stage); |
| 825 |
GLuint rgb_shift, alpha_shift; |
| 826 |
struct ureg out, shift; |
| 827 |
struct ureg dest; |
| 828 |
|
| 829 |
if (!key->unit[unit].enabled) { |
| 830 |
return get_source(p, SRC_PREVIOUS, 0); |
| 831 |
} |
| 832 |
|
| 833 |
switch (key->unit[unit].ModeRGB) { |
| 834 |
case MODE_DOT3_RGB_EXT: |
| 835 |
alpha_shift = key->unit[unit].ScaleShiftA; |
| 836 |
rgb_shift = 0; |
| 837 |
break; |
| 838 |
case MODE_DOT3_RGBA_EXT: |
| 839 |
alpha_shift = 0; |
| 840 |
rgb_shift = 0; |
| 841 |
break; |
| 842 |
default: |
| 843 |
rgb_shift = key->unit[unit].ScaleShiftRGB; |
| 844 |
alpha_shift = key->unit[unit].ScaleShiftA; |
| 845 |
break; |
| 846 |
} |
| 847 |
|
| 848 |
/* If this is the very last calculation, emit direct to output reg: |
| 849 |
*/ |
| 850 |
if (key->separate_specular || |
| 851 |
unit != p->last_tex_stage || |
| 852 |
alpha_shift || |
| 853 |
rgb_shift) |
| 854 |
dest = get_temp( p ); |
| 855 |
else |
| 856 |
dest = make_ureg(PROGRAM_OUTPUT, FRAG_OUTPUT_COLR); |
| 857 |
|
| 858 |
/* Emit the RGB and A combine ops |
| 859 |
*/ |
| 860 |
if (key->unit[unit].ModeRGB == key->unit[unit].ModeA && |
| 861 |
args_match(key, unit)) { |
| 862 |
out = emit_combine( p, dest, WRITEMASK_XYZW, saturate, |
| 863 |
unit, |
| 864 |
key->unit[unit].NumArgsRGB, |
| 865 |
key->unit[unit].ModeRGB, |
| 866 |
key->unit[unit].OptRGB); |
| 867 |
} |
| 868 |
else if (key->unit[unit].ModeRGB == MODE_DOT3_RGBA_EXT || |
| 869 |
key->unit[unit].ModeA == MODE_DOT3_RGBA) { |
| 870 |
|
| 871 |
out = emit_combine( p, dest, WRITEMASK_XYZW, saturate, |
| 872 |
unit, |
| 873 |
key->unit[unit].NumArgsRGB, |
| 874 |
key->unit[unit].ModeRGB, |
| 875 |
key->unit[unit].OptRGB); |
| 876 |
} |
| 877 |
else { |
| 878 |
/* Need to do something to stop from re-emitting identical |
| 879 |
* argument calculations here: |
| 880 |
*/ |
| 881 |
out = emit_combine( p, dest, WRITEMASK_XYZ, saturate, |
| 882 |
unit, |
| 883 |
key->unit[unit].NumArgsRGB, |
| 884 |
key->unit[unit].ModeRGB, |
| 885 |
key->unit[unit].OptRGB); |
| 886 |
out = emit_combine( p, dest, WRITEMASK_W, saturate, |
| 887 |
unit, |
| 888 |
key->unit[unit].NumArgsA, |
| 889 |
key->unit[unit].ModeA, |
| 890 |
key->unit[unit].OptA); |
| 891 |
} |
| 892 |
|
| 893 |
/* Deal with the final shift: |
| 894 |
*/ |
| 895 |
if (alpha_shift || rgb_shift) { |
| 896 |
if (rgb_shift == alpha_shift) { |
| 897 |
shift = register_scalar_const(p, 1<<rgb_shift); |
| 898 |
} |
| 899 |
else { |
| 900 |
shift = register_const4f(p, |
| 901 |
1<<rgb_shift, |
| 902 |
1<<rgb_shift, |
| 903 |
1<<rgb_shift, |
| 904 |
1<<alpha_shift); |
| 905 |
} |
| 906 |
return emit_arith( p, FP_OPCODE_MUL, dest, WRITEMASK_XYZW, |
| 907 |
saturate, out, shift, undef ); |
| 908 |
} |
| 909 |
else |
| 910 |
return out; |
| 911 |
} |
| 912 |
|
| 913 |
|
| 914 |
|
| 915 |
static void load_texture( struct texenv_fragment_program *p, GLuint unit ) |
| 916 |
{ |
| 917 |
if (is_undef(p->src_texture[unit])) { |
| 918 |
GLuint dim = p->state->unit[unit].source_index; |
| 919 |
struct ureg texcoord = register_input(p, FRAG_ATTRIB_TEX0+unit); |
| 920 |
struct ureg tmp = get_tex_temp( p ); |
| 921 |
|
| 922 |
if (dim == TEXTURE_UNKNOWN_INDEX) |
| 923 |
program_error(p, "TexSrcBit"); |
| 924 |
|
| 925 |
/* TODO: Use D0_MASK_XY where possible. |
| 926 |
*/ |
| 927 |
p->src_texture[unit] = emit_texld( p, FP_OPCODE_TXP, |
| 928 |
tmp, WRITEMASK_XYZW, |
| 929 |
unit, dim, texcoord ); |
| 930 |
} |
| 931 |
} |
| 932 |
|
| 933 |
static GLboolean load_texenv_source( struct texenv_fragment_program *p, |
| 934 |
GLuint src, GLuint unit ) |
| 935 |
{ |
| 936 |
switch (src) { |
| 937 |
case SRC_TEXTURE: |
| 938 |
load_texture(p, unit); |
| 939 |
break; |
| 940 |
|
| 941 |
case SRC_TEXTURE0: |
| 942 |
case SRC_TEXTURE1: |
| 943 |
case SRC_TEXTURE2: |
| 944 |
case SRC_TEXTURE3: |
| 945 |
case SRC_TEXTURE4: |
| 946 |
case SRC_TEXTURE5: |
| 947 |
case SRC_TEXTURE6: |
| 948 |
case SRC_TEXTURE7: |
| 949 |
if (!p->state->unit[src - SRC_TEXTURE0].enabled) |
| 950 |
return GL_FALSE; |
| 951 |
load_texture(p, src - SRC_TEXTURE0); |
| 952 |
break; |
| 953 |
|
| 954 |
default: |
| 955 |
break; |
| 956 |
} |
| 957 |
|
| 958 |
return GL_TRUE; |
| 959 |
} |
| 960 |
|
| 961 |
static GLboolean load_texunit_sources( struct texenv_fragment_program *p, int unit ) |
| 962 |
{ |
| 963 |
struct state_key *key = p->state; |
| 964 |
int i, nr = key->unit[unit].NumArgsRGB; |
| 965 |
for (i = 0; i < nr; i++) { |
| 966 |
if (!load_texenv_source( p, key->unit[unit].OptRGB[i].Source, unit) || |
| 967 |
!load_texenv_source( p, key->unit[unit].OptA[i].Source, unit )) |
| 968 |
return GL_FALSE; |
| 969 |
} |
| 970 |
return GL_TRUE; |
| 971 |
} |
| 972 |
|
| 973 |
static void create_new_program(struct state_key *key, GLcontext *ctx, |
| 974 |
struct fragment_program *program) |
| 975 |
{ |
| 976 |
struct texenv_fragment_program p; |
| 977 |
GLuint unit; |
| 978 |
struct ureg cf, out; |
| 979 |
|
| 980 |
_mesa_memset(&p, 0, sizeof(p)); |
| 981 |
p.ctx = ctx; |
| 982 |
p.state = key; |
| 983 |
p.program = program; |
| 984 |
|
| 985 |
p.program->Instructions = MALLOC(sizeof(struct fp_instruction) * 100); |
| 986 |
p.program->Base.NumInstructions = 0; |
| 987 |
p.program->Base.Target = GL_FRAGMENT_PROGRAM_ARB; |
| 988 |
p.program->NumTexIndirections = 1; /* correct? */ |
| 989 |
p.program->NumTexInstructions = 0; |
| 990 |
p.program->NumAluInstructions = 0; |
| 991 |
p.program->Base.String = 0; |
| 992 |
p.program->Base.NumInstructions = |
| 993 |
p.program->Base.NumTemporaries = |
| 994 |
p.program->Base.NumParameters = |
| 995 |
p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0; |
| 996 |
p.program->Parameters = _mesa_new_parameter_list(); |
| 997 |
|
| 998 |
p.program->InputsRead = 0; |
| 999 |
p.program->OutputsWritten = 1 << FRAG_OUTPUT_COLR; |
| 1000 |
|
| 1001 |
for (unit = 0; unit < MAX_TEXTURE_UNITS; unit++) |
| 1002 |
p.src_texture[unit] = undef; |
| 1003 |
|
| 1004 |
p.src_previous = undef; |
| 1005 |
p.last_tex_stage = 0; |
| 1006 |
release_temps(&p); |
| 1007 |
|
| 1008 |
if (key->enabled_units) { |
| 1009 |
/* First pass - to support texture_env_crossbar, first identify |
| 1010 |
* all referenced texture sources and emit texld instructions |
| 1011 |
* for each: |
| 1012 |
*/ |
| 1013 |
for (unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++) |
| 1014 |
if (key->unit[unit].enabled) { |
| 1015 |
if (load_texunit_sources( &p, unit )) |
| 1016 |
p.last_tex_stage = unit; |
| 1017 |
} |
| 1018 |
|
| 1019 |
/* Second pass - emit combine instructions to build final color: |
| 1020 |
*/ |
| 1021 |
for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) |
| 1022 |
if (key->enabled_units & (1<<unit)) { |
| 1023 |
p.src_previous = emit_texenv( &p, unit ); |
| 1024 |
release_temps(&p); /* release all temps */ |
| 1025 |
} |
| 1026 |
} |
| 1027 |
|
| 1028 |
cf = get_source( &p, SRC_PREVIOUS, 0 ); |
| 1029 |
out = make_ureg( PROGRAM_OUTPUT, FRAG_OUTPUT_COLR ); |
| 1030 |
|
| 1031 |
if (key->separate_specular) { |
| 1032 |
/* Emit specular add. |
| 1033 |
*/ |
| 1034 |
struct ureg s = register_input(&p, FRAG_ATTRIB_COL1); |
| 1035 |
emit_arith( &p, FP_OPCODE_ADD, out, WRITEMASK_XYZ, 0, cf, s, undef ); |
| 1036 |
} |
| 1037 |
else if (memcmp(&cf, &out, sizeof(cf)) != 0) { |
| 1038 |
/* Will wind up in here if no texture enabled or a couple of |
| 1039 |
* other scenarios (GL_REPLACE for instance). |
| 1040 |
*/ |
| 1041 |
emit_arith( &p, FP_OPCODE_MOV, out, WRITEMASK_XYZW, 0, cf, undef, undef ); |
| 1042 |
} |
| 1043 |
|
| 1044 |
/* Finish up: |
| 1045 |
*/ |
| 1046 |
emit_arith( &p, FP_OPCODE_END, undef, WRITEMASK_XYZW, 0, undef, undef, undef); |
| 1047 |
|
| 1048 |
if (key->fog_enabled) { |
| 1049 |
/* Pull fog mode from GLcontext, the value in the state key is |
| 1050 |
* a reduced value and not what is expected in FogOption |
| 1051 |
*/ |
| 1052 |
p.program->FogOption = ctx->Fog.Mode; |
| 1053 |
} else |
| 1054 |
p.program->FogOption = GL_NONE; |
| 1055 |
|
| 1056 |
if (p.program->NumTexIndirections > ctx->Const.MaxFragmentProgramTexIndirections) |
| 1057 |
program_error(&p, "Exceeded max nr indirect texture lookups"); |
| 1058 |
|
| 1059 |
if (p.program->NumTexInstructions > ctx->Const.MaxFragmentProgramTexInstructions) |
| 1060 |
program_error(&p, "Exceeded max TEX instructions"); |
| 1061 |
|
| 1062 |
if (p.program->NumAluInstructions > ctx->Const.MaxFragmentProgramAluInstructions) |
| 1063 |
program_error(&p, "Exceeded max ALU instructions"); |
| 1064 |
|
| 1065 |
|
| 1066 |
/* Notify driver the fragment program has (actually) changed. |
| 1067 |
*/ |
| 1068 |
if (ctx->Driver.ProgramStringNotify || DISASSEM) { |
| 1069 |
if (ctx->Driver.ProgramStringNotify) |
| 1070 |
ctx->Driver.ProgramStringNotify( ctx, GL_FRAGMENT_PROGRAM_ARB, |
| 1071 |
&p.program->Base ); |
| 1072 |
|
| 1073 |
if (DISASSEM) { |
| 1074 |
_mesa_debug_fp_inst(p.program->NumTexInstructions + p.program->NumAluInstructions, |
| 1075 |
p.program->Instructions); |
| 1076 |
_mesa_printf("\n"); |
| 1077 |
} |
| 1078 |
|
| 1079 |
} |
| 1080 |
|
| 1081 |
} |
| 1082 |
|
| 1083 |
static void *search_cache( struct texenvprog_cache *cache, |
| 1084 |
GLuint hash, |
| 1085 |
const void *key, |
| 1086 |
GLuint keysize) |
| 1087 |
{ |
| 1088 |
struct texenvprog_cache *c; |
| 1089 |
|
| 1090 |
for (c = cache; c; c = c->next) { |
| 1091 |
if (c->hash == hash && memcmp(c->key, key, keysize) == 0) |
| 1092 |
return c->data; |
| 1093 |
} |
| 1094 |
|
| 1095 |
return NULL; |
| 1096 |
} |
| 1097 |
|
| 1098 |
static void cache_item( struct texenvprog_cache **cache, |
| 1099 |
GLuint hash, |
| 1100 |
void *key, |
| 1101 |
void *data ) |
| 1102 |
{ |
| 1103 |
struct texenvprog_cache *c = MALLOC(sizeof(*c)); |
| 1104 |
c->hash = hash; |
| 1105 |
c->key = key; |
| 1106 |
c->data = data; |
| 1107 |
c->next = *cache; |
| 1108 |
*cache = c; |
| 1109 |
} |
| 1110 |
|
| 1111 |
static GLuint hash_key( struct state_key *key ) |
| 1112 |
{ |
| 1113 |
GLuint *ikey = (GLuint *)key; |
| 1114 |
GLuint hash = 0, i; |
| 1115 |
|
| 1116 |
/* I'm sure this can be improved on, but speed is important: |
| 1117 |
*/ |
| 1118 |
for (i = 0; i < sizeof(*key)/sizeof(GLuint); i++) |
| 1119 |
hash ^= ikey[i]; |
| 1120 |
|
| 1121 |
return hash; |
| 1122 |
} |
| 1123 |
|
| 1124 |
void _mesa_UpdateTexEnvProgram( GLcontext *ctx ) |
| 1125 |
{ |
| 1126 |
struct state_key *key; |
| 1127 |
GLuint hash; |
| 1128 |
|
| 1129 |
if (ctx->FragmentProgram._Enabled) |
| 1130 |
return; |
| 1131 |
|
| 1132 |
key = make_state_key(ctx); |
| 1133 |
hash = hash_key(key); |
| 1134 |
|
| 1135 |
ctx->FragmentProgram._Current = ctx->_TexEnvProgram = |
| 1136 |
(struct fragment_program *) |
| 1137 |
search_cache(ctx->Texture.env_fp_cache, hash, key, sizeof(*key)); |
| 1138 |
|
| 1139 |
if (!ctx->_TexEnvProgram) { |
| 1140 |
if (0) _mesa_printf("Building new texenv proggy for key %x\n", hash); |
| 1141 |
|
| 1142 |
ctx->FragmentProgram._Current = ctx->_TexEnvProgram = |
| 1143 |
(struct fragment_program *) |
| 1144 |
ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); |
| 1145 |
|
| 1146 |
create_new_program(key, ctx, ctx->_TexEnvProgram); |
| 1147 |
|
| 1148 |
cache_item(&ctx->Texture.env_fp_cache, hash, key, ctx->_TexEnvProgram); |
| 1149 |
} else { |
| 1150 |
FREE(key); |
| 1151 |
if (0) _mesa_printf("Found existing texenv program for key %x\n", hash); |
| 1152 |
} |
| 1153 |
|
| 1154 |
} |
| 1155 |
|
| 1156 |
void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx ) |
| 1157 |
{ |
| 1158 |
struct texenvprog_cache *a, *tmp; |
| 1159 |
|
| 1160 |
for (a = ctx->Texture.env_fp_cache; a; a = tmp) { |
| 1161 |
tmp = a->next; |
| 1162 |
FREE(a->key); |
| 1163 |
FREE(a->data); |
| 1164 |
FREE(a); |
| 1165 |
} |
| 1166 |
} |
| 1167 |
|