2 * Mesa 3-D graphics library
5 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 * \file prog_execute.c
27 * Software interpreter for vertex/fragment programs.
32 * NOTE: we do everything in single-precision floating point; we don't
33 * currently observe the single/half/fixed-precision qualifiers.
38 #include "main/glheader.h"
39 #include "main/colormac.h"
40 #include "main/context.h"
42 #include "prog_execute.h"
43 #include "prog_instruction.h"
44 #include "prog_parameter.h"
45 #include "prog_print.h"
46 #include "prog_noise.h"
54 * Set x to positive or negative infinity.
56 #if defined(USE_IEEE) || defined(_WIN32)
57 #define SET_POS_INFINITY(x) \
63 #define SET_NEG_INFINITY(x) \
70 #define SET_POS_INFINITY(x) x = __MAXFLOAT
71 #define SET_NEG_INFINITY(x) x = -__MAXFLOAT
73 #define SET_POS_INFINITY(x) x = (GLfloat) HUGE_VAL
74 #define SET_NEG_INFINITY(x) x = (GLfloat) -HUGE_VAL
77 #define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
80 static const GLfloat ZeroVec
[4] = { 0.0F
, 0.0F
, 0.0F
, 0.0F
};
85 * Return a pointer to the 4-element float vector specified by the given
88 static INLINE
const GLfloat
*
89 get_src_register_pointer(const struct prog_src_register
*source
,
90 const struct gl_program_machine
*machine
)
92 const struct gl_program
*prog
= machine
->CurProgram
;
93 GLint reg
= source
->Index
;
95 if (source
->RelAddr
) {
96 /* add address register value to src index/offset */
97 reg
+= machine
->AddressReg
[0][0];
103 switch (source
->File
) {
104 case PROGRAM_TEMPORARY
:
105 if (reg
>= MAX_PROGRAM_TEMPS
)
107 return machine
->Temporaries
[reg
];
110 if (prog
->Target
== GL_VERTEX_PROGRAM_ARB
) {
111 if (reg
>= VERT_ATTRIB_MAX
)
113 return machine
->VertAttribs
[reg
];
116 if (reg
>= FRAG_ATTRIB_MAX
)
118 return machine
->Attribs
[reg
][machine
->CurElement
];
122 if (reg
>= MAX_PROGRAM_OUTPUTS
)
124 return machine
->Outputs
[reg
];
126 case PROGRAM_LOCAL_PARAM
:
127 if (reg
>= MAX_PROGRAM_LOCAL_PARAMS
)
129 return machine
->CurProgram
->LocalParams
[reg
];
131 case PROGRAM_ENV_PARAM
:
132 if (reg
>= MAX_PROGRAM_ENV_PARAMS
)
134 return machine
->EnvParams
[reg
];
136 case PROGRAM_STATE_VAR
:
138 case PROGRAM_CONSTANT
:
140 case PROGRAM_UNIFORM
:
142 case PROGRAM_NAMED_PARAM
:
143 if (reg
>= (GLint
) prog
->Parameters
->NumParameters
)
145 return prog
->Parameters
->ParameterValues
[reg
];
149 "Invalid src register file %d in get_src_register_pointer()",
157 * Return a pointer to the 4-element float vector specified by the given
158 * destination register.
160 static INLINE GLfloat
*
161 get_dst_register_pointer(const struct prog_dst_register
*dest
,
162 struct gl_program_machine
*machine
)
164 static GLfloat dummyReg
[4];
165 GLint reg
= dest
->Index
;
168 /* add address register value to src index/offset */
169 reg
+= machine
->AddressReg
[0][0];
175 switch (dest
->File
) {
176 case PROGRAM_TEMPORARY
:
177 if (reg
>= MAX_PROGRAM_TEMPS
)
179 return machine
->Temporaries
[reg
];
182 if (reg
>= MAX_PROGRAM_OUTPUTS
)
184 return machine
->Outputs
[reg
];
186 case PROGRAM_WRITE_ONLY
:
191 "Invalid dest register file %d in get_dst_register_pointer()",
200 * Fetch a 4-element float vector from the given source register.
201 * Apply swizzling and negating as needed.
204 fetch_vector4(const struct prog_src_register
*source
,
205 const struct gl_program_machine
*machine
, GLfloat result
[4])
207 const GLfloat
*src
= get_src_register_pointer(source
, machine
);
210 if (source
->Swizzle
== SWIZZLE_NOOP
) {
212 COPY_4V(result
, src
);
215 ASSERT(GET_SWZ(source
->Swizzle
, 0) <= 3);
216 ASSERT(GET_SWZ(source
->Swizzle
, 1) <= 3);
217 ASSERT(GET_SWZ(source
->Swizzle
, 2) <= 3);
218 ASSERT(GET_SWZ(source
->Swizzle
, 3) <= 3);
219 result
[0] = src
[GET_SWZ(source
->Swizzle
, 0)];
220 result
[1] = src
[GET_SWZ(source
->Swizzle
, 1)];
221 result
[2] = src
[GET_SWZ(source
->Swizzle
, 2)];
222 result
[3] = src
[GET_SWZ(source
->Swizzle
, 3)];
226 result
[0] = FABSF(result
[0]);
227 result
[1] = FABSF(result
[1]);
228 result
[2] = FABSF(result
[2]);
229 result
[3] = FABSF(result
[3]);
231 if (source
->Negate
) {
232 ASSERT(source
->Negate
== NEGATE_XYZW
);
233 result
[0] = -result
[0];
234 result
[1] = -result
[1];
235 result
[2] = -result
[2];
236 result
[3] = -result
[3];
240 assert(!IS_INF_OR_NAN(result
[0]));
241 assert(!IS_INF_OR_NAN(result
[0]));
242 assert(!IS_INF_OR_NAN(result
[0]));
243 assert(!IS_INF_OR_NAN(result
[0]));
249 * Fetch a 4-element uint vector from the given source register.
250 * Apply swizzling but not negation/abs.
253 fetch_vector4ui(const struct prog_src_register
*source
,
254 const struct gl_program_machine
*machine
, GLuint result
[4])
256 const GLuint
*src
= (GLuint
*) get_src_register_pointer(source
, machine
);
259 if (source
->Swizzle
== SWIZZLE_NOOP
) {
261 COPY_4V(result
, src
);
264 ASSERT(GET_SWZ(source
->Swizzle
, 0) <= 3);
265 ASSERT(GET_SWZ(source
->Swizzle
, 1) <= 3);
266 ASSERT(GET_SWZ(source
->Swizzle
, 2) <= 3);
267 ASSERT(GET_SWZ(source
->Swizzle
, 3) <= 3);
268 result
[0] = src
[GET_SWZ(source
->Swizzle
, 0)];
269 result
[1] = src
[GET_SWZ(source
->Swizzle
, 1)];
270 result
[2] = src
[GET_SWZ(source
->Swizzle
, 2)];
271 result
[3] = src
[GET_SWZ(source
->Swizzle
, 3)];
274 /* Note: no Negate or Abs here */
280 * Fetch the derivative with respect to X or Y for the given register.
281 * XXX this currently only works for fragment program input attribs.
284 fetch_vector4_deriv(GLcontext
* ctx
,
285 const struct prog_src_register
*source
,
286 const struct gl_program_machine
*machine
,
287 char xOrY
, GLfloat result
[4])
289 if (source
->File
== PROGRAM_INPUT
&&
290 source
->Index
< (GLint
) machine
->NumDeriv
) {
291 const GLint col
= machine
->CurElement
;
292 const GLfloat w
= machine
->Attribs
[FRAG_ATTRIB_WPOS
][col
][3];
293 const GLfloat invQ
= 1.0f
/ w
;
297 deriv
[0] = machine
->DerivX
[source
->Index
][0] * invQ
;
298 deriv
[1] = machine
->DerivX
[source
->Index
][1] * invQ
;
299 deriv
[2] = machine
->DerivX
[source
->Index
][2] * invQ
;
300 deriv
[3] = machine
->DerivX
[source
->Index
][3] * invQ
;
303 deriv
[0] = machine
->DerivY
[source
->Index
][0] * invQ
;
304 deriv
[1] = machine
->DerivY
[source
->Index
][1] * invQ
;
305 deriv
[2] = machine
->DerivY
[source
->Index
][2] * invQ
;
306 deriv
[3] = machine
->DerivY
[source
->Index
][3] * invQ
;
309 result
[0] = deriv
[GET_SWZ(source
->Swizzle
, 0)];
310 result
[1] = deriv
[GET_SWZ(source
->Swizzle
, 1)];
311 result
[2] = deriv
[GET_SWZ(source
->Swizzle
, 2)];
312 result
[3] = deriv
[GET_SWZ(source
->Swizzle
, 3)];
315 result
[0] = FABSF(result
[0]);
316 result
[1] = FABSF(result
[1]);
317 result
[2] = FABSF(result
[2]);
318 result
[3] = FABSF(result
[3]);
320 if (source
->Negate
) {
321 ASSERT(source
->Negate
== NEGATE_XYZW
);
322 result
[0] = -result
[0];
323 result
[1] = -result
[1];
324 result
[2] = -result
[2];
325 result
[3] = -result
[3];
329 ASSIGN_4V(result
, 0.0, 0.0, 0.0, 0.0);
335 * As above, but only return result[0] element.
338 fetch_vector1(const struct prog_src_register
*source
,
339 const struct gl_program_machine
*machine
, GLfloat result
[4])
341 const GLfloat
*src
= get_src_register_pointer(source
, machine
);
344 result
[0] = src
[GET_SWZ(source
->Swizzle
, 0)];
347 result
[0] = FABSF(result
[0]);
349 if (source
->Negate
) {
350 result
[0] = -result
[0];
356 * Fetch texel from texture. Use partial derivatives when possible.
359 fetch_texel(GLcontext
*ctx
,
360 const struct gl_program_machine
*machine
,
361 const struct prog_instruction
*inst
,
362 const GLfloat texcoord
[4], GLfloat lodBias
,
365 const GLuint unit
= machine
->Samplers
[inst
->TexSrcUnit
];
367 /* Note: we only have the right derivatives for fragment input attribs.
369 if (machine
->NumDeriv
> 0 &&
370 inst
->SrcReg
[0].File
== PROGRAM_INPUT
&&
371 inst
->SrcReg
[0].Index
== FRAG_ATTRIB_TEX0
+ inst
->TexSrcUnit
) {
372 /* simple texture fetch for which we should have derivatives */
373 GLuint attr
= inst
->SrcReg
[0].Index
;
374 machine
->FetchTexelDeriv(ctx
, texcoord
,
375 machine
->DerivX
[attr
],
376 machine
->DerivY
[attr
],
377 lodBias
, unit
, color
);
380 machine
->FetchTexelLod(ctx
, texcoord
, lodBias
, unit
, color
);
386 * Test value against zero and return GT, LT, EQ or UN if NaN.
389 generate_cc(float value
)
392 return COND_UN
; /* NaN */
402 * Test if the ccMaskRule is satisfied by the given condition code.
403 * Used to mask destination writes according to the current condition code.
405 static INLINE GLboolean
406 test_cc(GLuint condCode
, GLuint ccMaskRule
)
408 switch (ccMaskRule
) {
409 case COND_EQ
: return (condCode
== COND_EQ
);
410 case COND_NE
: return (condCode
!= COND_EQ
);
411 case COND_LT
: return (condCode
== COND_LT
);
412 case COND_GE
: return (condCode
== COND_GT
|| condCode
== COND_EQ
);
413 case COND_LE
: return (condCode
== COND_LT
|| condCode
== COND_EQ
);
414 case COND_GT
: return (condCode
== COND_GT
);
415 case COND_TR
: return GL_TRUE
;
416 case COND_FL
: return GL_FALSE
;
417 default: return GL_TRUE
;
423 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
424 * or GL_FALSE to indicate result.
426 static INLINE GLboolean
427 eval_condition(const struct gl_program_machine
*machine
,
428 const struct prog_instruction
*inst
)
430 const GLuint swizzle
= inst
->DstReg
.CondSwizzle
;
431 const GLuint condMask
= inst
->DstReg
.CondMask
;
432 if (test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 0)], condMask
) ||
433 test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 1)], condMask
) ||
434 test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 2)], condMask
) ||
435 test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 3)], condMask
)) {
446 * Store 4 floats into a register. Observe the instructions saturate and
447 * set-condition-code flags.
450 store_vector4(const struct prog_instruction
*inst
,
451 struct gl_program_machine
*machine
, const GLfloat value
[4])
453 const struct prog_dst_register
*dstReg
= &(inst
->DstReg
);
454 const GLboolean clamp
= inst
->SaturateMode
== SATURATE_ZERO_ONE
;
455 GLuint writeMask
= dstReg
->WriteMask
;
456 GLfloat clampedValue
[4];
457 GLfloat
*dst
= get_dst_register_pointer(dstReg
, machine
);
460 if (value
[0] > 1.0e10
||
461 IS_INF_OR_NAN(value
[0]) ||
462 IS_INF_OR_NAN(value
[1]) ||
463 IS_INF_OR_NAN(value
[2]) || IS_INF_OR_NAN(value
[3]))
464 printf("store %g %g %g %g\n", value
[0], value
[1], value
[2], value
[3]);
468 clampedValue
[0] = CLAMP(value
[0], 0.0F
, 1.0F
);
469 clampedValue
[1] = CLAMP(value
[1], 0.0F
, 1.0F
);
470 clampedValue
[2] = CLAMP(value
[2], 0.0F
, 1.0F
);
471 clampedValue
[3] = CLAMP(value
[3], 0.0F
, 1.0F
);
472 value
= clampedValue
;
475 if (dstReg
->CondMask
!= COND_TR
) {
476 /* condition codes may turn off some writes */
477 if (writeMask
& WRITEMASK_X
) {
478 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 0)],
480 writeMask
&= ~WRITEMASK_X
;
482 if (writeMask
& WRITEMASK_Y
) {
483 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 1)],
485 writeMask
&= ~WRITEMASK_Y
;
487 if (writeMask
& WRITEMASK_Z
) {
488 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 2)],
490 writeMask
&= ~WRITEMASK_Z
;
492 if (writeMask
& WRITEMASK_W
) {
493 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 3)],
495 writeMask
&= ~WRITEMASK_W
;
500 assert(!IS_INF_OR_NAN(value
[0]));
501 assert(!IS_INF_OR_NAN(value
[0]));
502 assert(!IS_INF_OR_NAN(value
[0]));
503 assert(!IS_INF_OR_NAN(value
[0]));
506 if (writeMask
& WRITEMASK_X
)
508 if (writeMask
& WRITEMASK_Y
)
510 if (writeMask
& WRITEMASK_Z
)
512 if (writeMask
& WRITEMASK_W
)
515 if (inst
->CondUpdate
) {
516 if (writeMask
& WRITEMASK_X
)
517 machine
->CondCodes
[0] = generate_cc(value
[0]);
518 if (writeMask
& WRITEMASK_Y
)
519 machine
->CondCodes
[1] = generate_cc(value
[1]);
520 if (writeMask
& WRITEMASK_Z
)
521 machine
->CondCodes
[2] = generate_cc(value
[2]);
522 if (writeMask
& WRITEMASK_W
)
523 machine
->CondCodes
[3] = generate_cc(value
[3]);
525 printf("CondCodes=(%s,%s,%s,%s) for:\n",
526 _mesa_condcode_string(machine
->CondCodes
[0]),
527 _mesa_condcode_string(machine
->CondCodes
[1]),
528 _mesa_condcode_string(machine
->CondCodes
[2]),
529 _mesa_condcode_string(machine
->CondCodes
[3]));
536 * Store 4 uints into a register. Observe the set-condition-code flags.
539 store_vector4ui(const struct prog_instruction
*inst
,
540 struct gl_program_machine
*machine
, const GLuint value
[4])
542 const struct prog_dst_register
*dstReg
= &(inst
->DstReg
);
543 GLuint writeMask
= dstReg
->WriteMask
;
544 GLuint
*dst
= (GLuint
*) get_dst_register_pointer(dstReg
, machine
);
546 if (dstReg
->CondMask
!= COND_TR
) {
547 /* condition codes may turn off some writes */
548 if (writeMask
& WRITEMASK_X
) {
549 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 0)],
551 writeMask
&= ~WRITEMASK_X
;
553 if (writeMask
& WRITEMASK_Y
) {
554 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 1)],
556 writeMask
&= ~WRITEMASK_Y
;
558 if (writeMask
& WRITEMASK_Z
) {
559 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 2)],
561 writeMask
&= ~WRITEMASK_Z
;
563 if (writeMask
& WRITEMASK_W
) {
564 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 3)],
566 writeMask
&= ~WRITEMASK_W
;
570 if (writeMask
& WRITEMASK_X
)
572 if (writeMask
& WRITEMASK_Y
)
574 if (writeMask
& WRITEMASK_Z
)
576 if (writeMask
& WRITEMASK_W
)
579 if (inst
->CondUpdate
) {
580 if (writeMask
& WRITEMASK_X
)
581 machine
->CondCodes
[0] = generate_cc(value
[0]);
582 if (writeMask
& WRITEMASK_Y
)
583 machine
->CondCodes
[1] = generate_cc(value
[1]);
584 if (writeMask
& WRITEMASK_Z
)
585 machine
->CondCodes
[2] = generate_cc(value
[2]);
586 if (writeMask
& WRITEMASK_W
)
587 machine
->CondCodes
[3] = generate_cc(value
[3]);
589 printf("CondCodes=(%s,%s,%s,%s) for:\n",
590 _mesa_condcode_string(machine
->CondCodes
[0]),
591 _mesa_condcode_string(machine
->CondCodes
[1]),
592 _mesa_condcode_string(machine
->CondCodes
[2]),
593 _mesa_condcode_string(machine
->CondCodes
[3]));
601 * Execute the given vertex/fragment program.
603 * \param ctx rendering context
604 * \param program the program to execute
605 * \param machine machine state (must be initialized)
606 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
609 _mesa_execute_program(GLcontext
* ctx
,
610 const struct gl_program
*program
,
611 struct gl_program_machine
*machine
)
613 const GLuint numInst
= program
->NumInstructions
;
614 const GLuint maxExec
= 10000;
615 GLuint pc
, numExec
= 0;
617 machine
->CurProgram
= program
;
620 printf("execute program %u --------------------\n", program
->Id
);
623 if (program
->Target
== GL_VERTEX_PROGRAM_ARB
) {
624 machine
->EnvParams
= ctx
->VertexProgram
.Parameters
;
627 machine
->EnvParams
= ctx
->FragmentProgram
.Parameters
;
630 for (pc
= 0; pc
< numInst
; pc
++) {
631 const struct prog_instruction
*inst
= program
->Instructions
+ pc
;
634 _mesa_print_instruction(inst
);
637 switch (inst
->Opcode
) {
640 GLfloat a
[4], result
[4];
641 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
642 result
[0] = FABSF(a
[0]);
643 result
[1] = FABSF(a
[1]);
644 result
[2] = FABSF(a
[2]);
645 result
[3] = FABSF(a
[3]);
646 store_vector4(inst
, machine
, result
);
651 GLfloat a
[4], b
[4], result
[4];
652 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
653 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
654 result
[0] = a
[0] + b
[0];
655 result
[1] = a
[1] + b
[1];
656 result
[2] = a
[2] + b
[2];
657 result
[3] = a
[3] + b
[3];
658 store_vector4(inst
, machine
, result
);
660 printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
661 result
[0], result
[1], result
[2], result
[3],
662 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
666 case OPCODE_AND
: /* bitwise AND */
668 GLuint a
[4], b
[4], result
[4];
669 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
670 fetch_vector4ui(&inst
->SrcReg
[1], machine
, b
);
671 result
[0] = a
[0] & b
[0];
672 result
[1] = a
[1] & b
[1];
673 result
[2] = a
[2] & b
[2];
674 result
[3] = a
[3] & b
[3];
675 store_vector4ui(inst
, machine
, result
);
681 fetch_vector4(&inst
->SrcReg
[0], machine
, t
);
682 machine
->AddressReg
[0][0] = IFLOOR(t
[0]);
689 /* subtract 1 here since pc is incremented by for(pc) loop */
690 pc
= inst
->BranchTarget
- 1; /* go to matching BNGLOOP */
692 case OPCODE_BGNSUB
: /* begin subroutine */
694 case OPCODE_ENDSUB
: /* end subroutine */
696 case OPCODE_BRA
: /* branch (conditional) */
698 case OPCODE_BRK
: /* break out of loop (conditional) */
700 case OPCODE_CONT
: /* continue loop (conditional) */
701 if (eval_condition(machine
, inst
)) {
703 /* Subtract 1 here since we'll do pc++ at end of for-loop */
704 pc
= inst
->BranchTarget
- 1;
707 case OPCODE_CAL
: /* Call subroutine (conditional) */
708 if (eval_condition(machine
, inst
)) {
709 /* call the subroutine */
710 if (machine
->StackDepth
>= MAX_PROGRAM_CALL_DEPTH
) {
711 return GL_TRUE
; /* Per GL_NV_vertex_program2 spec */
713 machine
->CallStack
[machine
->StackDepth
++] = pc
+ 1; /* next inst */
714 /* Subtract 1 here since we'll do pc++ at end of for-loop */
715 pc
= inst
->BranchTarget
- 1;
720 GLfloat a
[4], b
[4], c
[4], result
[4];
721 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
722 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
723 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
724 result
[0] = a
[0] < 0.0F
? b
[0] : c
[0];
725 result
[1] = a
[1] < 0.0F
? b
[1] : c
[1];
726 result
[2] = a
[2] < 0.0F
? b
[2] : c
[2];
727 result
[3] = a
[3] < 0.0F
? b
[3] : c
[3];
728 store_vector4(inst
, machine
, result
);
733 GLfloat a
[4], result
[4];
734 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
735 result
[0] = result
[1] = result
[2] = result
[3]
736 = (GLfloat
) _mesa_cos(a
[0]);
737 store_vector4(inst
, machine
, result
);
740 case OPCODE_DDX
: /* Partial derivative with respect to X */
743 fetch_vector4_deriv(ctx
, &inst
->SrcReg
[0], machine
,
745 store_vector4(inst
, machine
, result
);
748 case OPCODE_DDY
: /* Partial derivative with respect to Y */
751 fetch_vector4_deriv(ctx
, &inst
->SrcReg
[0], machine
,
753 store_vector4(inst
, machine
, result
);
758 GLfloat a
[4], b
[4], result
[4];
759 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
760 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
761 result
[0] = result
[1] = result
[2] = result
[3] = DOT2(a
, b
);
762 store_vector4(inst
, machine
, result
);
764 printf("DP2 %g = (%g %g) . (%g %g)\n",
765 result
[0], a
[0], a
[1], b
[0], b
[1]);
771 GLfloat a
[4], b
[4], c
, result
[4];
772 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
773 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
774 fetch_vector1(&inst
->SrcReg
[1], machine
, &c
);
775 result
[0] = result
[1] = result
[2] = result
[3] = DOT2(a
, b
) + c
;
776 store_vector4(inst
, machine
, result
);
778 printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
779 result
[0], a
[0], a
[1], b
[0], b
[1], c
);
785 GLfloat a
[4], b
[4], result
[4];
786 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
787 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
788 result
[0] = result
[1] = result
[2] = result
[3] = DOT3(a
, b
);
789 store_vector4(inst
, machine
, result
);
791 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
792 result
[0], a
[0], a
[1], a
[2], b
[0], b
[1], b
[2]);
798 GLfloat a
[4], b
[4], result
[4];
799 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
800 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
801 result
[0] = result
[1] = result
[2] = result
[3] = DOT4(a
, b
);
802 store_vector4(inst
, machine
, result
);
804 printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
805 result
[0], a
[0], a
[1], a
[2], a
[3],
806 b
[0], b
[1], b
[2], b
[3]);
812 GLfloat a
[4], b
[4], result
[4];
813 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
814 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
815 result
[0] = result
[1] = result
[2] = result
[3] = DOT3(a
, b
) + b
[3];
816 store_vector4(inst
, machine
, result
);
819 case OPCODE_DST
: /* Distance vector */
821 GLfloat a
[4], b
[4], result
[4];
822 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
823 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
825 result
[1] = a
[1] * b
[1];
828 store_vector4(inst
, machine
, result
);
833 GLfloat t
[4], q
[4], floor_t0
;
834 fetch_vector1(&inst
->SrcReg
[0], machine
, t
);
835 floor_t0
= FLOORF(t
[0]);
836 if (floor_t0
> FLT_MAX_EXP
) {
837 SET_POS_INFINITY(q
[0]);
838 SET_POS_INFINITY(q
[2]);
840 else if (floor_t0
< FLT_MIN_EXP
) {
845 q
[0] = LDEXPF(1.0, (int) floor_t0
);
846 /* Note: GL_NV_vertex_program expects
847 * result.z = result.x * APPX(result.y)
848 * We do what the ARB extension says.
850 q
[2] = (GLfloat
) _mesa_pow(2.0, t
[0]);
852 q
[1] = t
[0] - floor_t0
;
854 store_vector4( inst
, machine
, q
);
857 case OPCODE_EX2
: /* Exponential base 2 */
859 GLfloat a
[4], result
[4], val
;
860 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
861 val
= (GLfloat
) _mesa_pow(2.0, a
[0]);
863 if (IS_INF_OR_NAN(val))
866 result
[0] = result
[1] = result
[2] = result
[3] = val
;
867 store_vector4(inst
, machine
, result
);
872 GLfloat a
[4], result
[4];
873 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
874 result
[0] = FLOORF(a
[0]);
875 result
[1] = FLOORF(a
[1]);
876 result
[2] = FLOORF(a
[2]);
877 result
[3] = FLOORF(a
[3]);
878 store_vector4(inst
, machine
, result
);
883 GLfloat a
[4], result
[4];
884 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
885 result
[0] = a
[0] - FLOORF(a
[0]);
886 result
[1] = a
[1] - FLOORF(a
[1]);
887 result
[2] = a
[2] - FLOORF(a
[2]);
888 result
[3] = a
[3] - FLOORF(a
[3]);
889 store_vector4(inst
, machine
, result
);
896 if (inst
->SrcReg
[0].File
!= PROGRAM_UNDEFINED
) {
898 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
899 cond
= (a
[0] != 0.0);
902 cond
= eval_condition(machine
, inst
);
905 printf("IF: %d\n", cond
);
909 /* do if-clause (just continue execution) */
912 /* go to the instruction after ELSE or ENDIF */
913 assert(inst
->BranchTarget
>= 0);
914 pc
= inst
->BranchTarget
- 1;
920 assert(inst
->BranchTarget
>= 0);
921 pc
= inst
->BranchTarget
- 1;
926 case OPCODE_KIL_NV
: /* NV_f_p only (conditional) */
927 if (eval_condition(machine
, inst
)) {
931 case OPCODE_KIL
: /* ARB_f_p only */
934 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
936 printf("KIL if (%g %g %g %g) <= 0.0\n",
937 a
[0], a
[1], a
[2], a
[3]);
940 if (a
[0] < 0.0F
|| a
[1] < 0.0F
|| a
[2] < 0.0F
|| a
[3] < 0.0F
) {
945 case OPCODE_LG2
: /* log base 2 */
947 GLfloat a
[4], result
[4], val
;
948 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
949 /* The fast LOG2 macro doesn't meet the precision requirements.
955 val
= log(a
[0]) * 1.442695F
;
957 result
[0] = result
[1] = result
[2] = result
[3] = val
;
958 store_vector4(inst
, machine
, result
);
963 const GLfloat epsilon
= 1.0F
/ 256.0F
; /* from NV VP spec */
964 GLfloat a
[4], result
[4];
965 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
966 a
[0] = MAX2(a
[0], 0.0F
);
967 a
[1] = MAX2(a
[1], 0.0F
);
968 /* XXX ARB version clamps a[3], NV version doesn't */
969 a
[3] = CLAMP(a
[3], -(128.0F
- epsilon
), (128.0F
- epsilon
));
972 /* XXX we could probably just use pow() here */
974 if (a
[1] == 0.0 && a
[3] == 0.0)
977 result
[2] = (GLfloat
) _mesa_pow(a
[1], a
[3]);
983 store_vector4(inst
, machine
, result
);
985 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
986 result
[0], result
[1], result
[2], result
[3],
987 a
[0], a
[1], a
[2], a
[3]);
993 GLfloat t
[4], q
[4], abs_t0
;
994 fetch_vector1(&inst
->SrcReg
[0], machine
, t
);
995 abs_t0
= FABSF(t
[0]);
996 if (abs_t0
!= 0.0F
) {
997 /* Since we really can't handle infinite values on VMS
998 * like other OSes we'll use __MAXFLOAT to represent
999 * infinity. This may need some tweaking.
1002 if (abs_t0
== __MAXFLOAT
)
1004 if (IS_INF_OR_NAN(abs_t0
))
1007 SET_POS_INFINITY(q
[0]);
1009 SET_POS_INFINITY(q
[2]);
1013 GLfloat mantissa
= FREXPF(t
[0], &exponent
);
1014 q
[0] = (GLfloat
) (exponent
- 1);
1015 q
[1] = (GLfloat
) (2.0 * mantissa
); /* map [.5, 1) -> [1, 2) */
1017 /* The fast LOG2 macro doesn't meet the precision
1020 q
[2] = (log(t
[0]) * 1.442695F
);
1024 SET_NEG_INFINITY(q
[0]);
1026 SET_NEG_INFINITY(q
[2]);
1029 store_vector4(inst
, machine
, q
);
1034 GLfloat a
[4], b
[4], c
[4], result
[4];
1035 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1036 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1037 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
1038 result
[0] = a
[0] * b
[0] + (1.0F
- a
[0]) * c
[0];
1039 result
[1] = a
[1] * b
[1] + (1.0F
- a
[1]) * c
[1];
1040 result
[2] = a
[2] * b
[2] + (1.0F
- a
[2]) * c
[2];
1041 result
[3] = a
[3] * b
[3] + (1.0F
- a
[3]) * c
[3];
1042 store_vector4(inst
, machine
, result
);
1044 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
1045 "(%g %g %g %g), (%g %g %g %g)\n",
1046 result
[0], result
[1], result
[2], result
[3],
1047 a
[0], a
[1], a
[2], a
[3],
1048 b
[0], b
[1], b
[2], b
[3], c
[0], c
[1], c
[2], c
[3]);
1054 GLfloat a
[4], b
[4], c
[4], result
[4];
1055 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1056 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1057 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
1058 result
[0] = a
[0] * b
[0] + c
[0];
1059 result
[1] = a
[1] * b
[1] + c
[1];
1060 result
[2] = a
[2] * b
[2] + c
[2];
1061 result
[3] = a
[3] * b
[3] + c
[3];
1062 store_vector4(inst
, machine
, result
);
1064 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
1065 "(%g %g %g %g) + (%g %g %g %g)\n",
1066 result
[0], result
[1], result
[2], result
[3],
1067 a
[0], a
[1], a
[2], a
[3],
1068 b
[0], b
[1], b
[2], b
[3], c
[0], c
[1], c
[2], c
[3]);
1074 GLfloat a
[4], b
[4], result
[4];
1075 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1076 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1077 result
[0] = MAX2(a
[0], b
[0]);
1078 result
[1] = MAX2(a
[1], b
[1]);
1079 result
[2] = MAX2(a
[2], b
[2]);
1080 result
[3] = MAX2(a
[3], b
[3]);
1081 store_vector4(inst
, machine
, result
);
1083 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
1084 result
[0], result
[1], result
[2], result
[3],
1085 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
1091 GLfloat a
[4], b
[4], result
[4];
1092 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1093 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1094 result
[0] = MIN2(a
[0], b
[0]);
1095 result
[1] = MIN2(a
[1], b
[1]);
1096 result
[2] = MIN2(a
[2], b
[2]);
1097 result
[3] = MIN2(a
[3], b
[3]);
1098 store_vector4(inst
, machine
, result
);
1104 fetch_vector4(&inst
->SrcReg
[0], machine
, result
);
1105 store_vector4(inst
, machine
, result
);
1107 printf("MOV (%g %g %g %g)\n",
1108 result
[0], result
[1], result
[2], result
[3]);
1114 GLfloat a
[4], b
[4], result
[4];
1115 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1116 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1117 result
[0] = a
[0] * b
[0];
1118 result
[1] = a
[1] * b
[1];
1119 result
[2] = a
[2] * b
[2];
1120 result
[3] = a
[3] * b
[3];
1121 store_vector4(inst
, machine
, result
);
1123 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
1124 result
[0], result
[1], result
[2], result
[3],
1125 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
1131 GLfloat a
[4], result
[4];
1132 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1136 result
[3] = _mesa_noise1(a
[0]);
1137 store_vector4(inst
, machine
, result
);
1142 GLfloat a
[4], result
[4];
1143 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1146 result
[2] = result
[3] = _mesa_noise2(a
[0], a
[1]);
1147 store_vector4(inst
, machine
, result
);
1152 GLfloat a
[4], result
[4];
1153 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1157 result
[3] = _mesa_noise3(a
[0], a
[1], a
[2]);
1158 store_vector4(inst
, machine
, result
);
1163 GLfloat a
[4], result
[4];
1164 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1168 result
[3] = _mesa_noise4(a
[0], a
[1], a
[2], a
[3]);
1169 store_vector4(inst
, machine
, result
);
1174 case OPCODE_NOT
: /* bitwise NOT */
1176 GLuint a
[4], result
[4];
1177 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
1182 store_vector4ui(inst
, machine
, result
);
1185 case OPCODE_NRM3
: /* 3-component normalization */
1187 GLfloat a
[4], result
[4];
1189 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1190 tmp
= a
[0] * a
[0] + a
[1] * a
[1] + a
[2] * a
[2];
1192 tmp
= INV_SQRTF(tmp
);
1193 result
[0] = tmp
* a
[0];
1194 result
[1] = tmp
* a
[1];
1195 result
[2] = tmp
* a
[2];
1196 result
[3] = 0.0; /* undefined, but prevent valgrind warnings */
1197 store_vector4(inst
, machine
, result
);
1200 case OPCODE_NRM4
: /* 4-component normalization */
1202 GLfloat a
[4], result
[4];
1204 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1205 tmp
= a
[0] * a
[0] + a
[1] * a
[1] + a
[2] * a
[2] + a
[3] * a
[3];
1207 tmp
= INV_SQRTF(tmp
);
1208 result
[0] = tmp
* a
[0];
1209 result
[1] = tmp
* a
[1];
1210 result
[2] = tmp
* a
[2];
1211 result
[3] = tmp
* a
[3];
1212 store_vector4(inst
, machine
, result
);
1215 case OPCODE_OR
: /* bitwise OR */
1217 GLuint a
[4], b
[4], result
[4];
1218 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
1219 fetch_vector4ui(&inst
->SrcReg
[1], machine
, b
);
1220 result
[0] = a
[0] | b
[0];
1221 result
[1] = a
[1] | b
[1];
1222 result
[2] = a
[2] | b
[2];
1223 result
[3] = a
[3] | b
[3];
1224 store_vector4ui(inst
, machine
, result
);
1227 case OPCODE_PK2H
: /* pack two 16-bit floats in one 32-bit float */
1232 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1233 hx
= _mesa_float_to_half(a
[0]);
1234 hy
= _mesa_float_to_half(a
[1]);
1238 result
[3] = hx
| (hy
<< 16);
1239 store_vector4ui(inst
, machine
, result
);
1242 case OPCODE_PK2US
: /* pack two GLushorts into one 32-bit float */
1245 GLuint result
[4], usx
, usy
;
1246 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1247 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
1248 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
1249 usx
= IROUND(a
[0] * 65535.0F
);
1250 usy
= IROUND(a
[1] * 65535.0F
);
1254 result
[3] = usx
| (usy
<< 16);
1255 store_vector4ui(inst
, machine
, result
);
1258 case OPCODE_PK4B
: /* pack four GLbytes into one 32-bit float */
1261 GLuint result
[4], ubx
, uby
, ubz
, ubw
;
1262 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1263 a
[0] = CLAMP(a
[0], -128.0F
/ 127.0F
, 1.0F
);
1264 a
[1] = CLAMP(a
[1], -128.0F
/ 127.0F
, 1.0F
);
1265 a
[2] = CLAMP(a
[2], -128.0F
/ 127.0F
, 1.0F
);
1266 a
[3] = CLAMP(a
[3], -128.0F
/ 127.0F
, 1.0F
);
1267 ubx
= IROUND(127.0F
* a
[0] + 128.0F
);
1268 uby
= IROUND(127.0F
* a
[1] + 128.0F
);
1269 ubz
= IROUND(127.0F
* a
[2] + 128.0F
);
1270 ubw
= IROUND(127.0F
* a
[3] + 128.0F
);
1274 result
[3] = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
1275 store_vector4ui(inst
, machine
, result
);
1278 case OPCODE_PK4UB
: /* pack four GLubytes into one 32-bit float */
1281 GLuint result
[4], ubx
, uby
, ubz
, ubw
;
1282 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1283 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
1284 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
1285 a
[2] = CLAMP(a
[2], 0.0F
, 1.0F
);
1286 a
[3] = CLAMP(a
[3], 0.0F
, 1.0F
);
1287 ubx
= IROUND(255.0F
* a
[0]);
1288 uby
= IROUND(255.0F
* a
[1]);
1289 ubz
= IROUND(255.0F
* a
[2]);
1290 ubw
= IROUND(255.0F
* a
[3]);
1294 result
[3] = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
1295 store_vector4ui(inst
, machine
, result
);
1300 GLfloat a
[4], b
[4], result
[4];
1301 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1302 fetch_vector1(&inst
->SrcReg
[1], machine
, b
);
1303 result
[0] = result
[1] = result
[2] = result
[3]
1304 = (GLfloat
) _mesa_pow(a
[0], b
[0]);
1305 store_vector4(inst
, machine
, result
);
1310 GLfloat a
[4], result
[4];
1311 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1315 else if (IS_INF_OR_NAN(a
[0]))
1316 printf("RCP(inf)\n");
1318 result
[0] = result
[1] = result
[2] = result
[3] = 1.0F
/ a
[0];
1319 store_vector4(inst
, machine
, result
);
1322 case OPCODE_RET
: /* return from subroutine (conditional) */
1323 if (eval_condition(machine
, inst
)) {
1324 if (machine
->StackDepth
== 0) {
1325 return GL_TRUE
; /* Per GL_NV_vertex_program2 spec */
1327 /* subtract one because of pc++ in the for loop */
1328 pc
= machine
->CallStack
[--machine
->StackDepth
] - 1;
1331 case OPCODE_RFL
: /* reflection vector */
1333 GLfloat axis
[4], dir
[4], result
[4], tmpX
, tmpW
;
1334 fetch_vector4(&inst
->SrcReg
[0], machine
, axis
);
1335 fetch_vector4(&inst
->SrcReg
[1], machine
, dir
);
1336 tmpW
= DOT3(axis
, axis
);
1337 tmpX
= (2.0F
* DOT3(axis
, dir
)) / tmpW
;
1338 result
[0] = tmpX
* axis
[0] - dir
[0];
1339 result
[1] = tmpX
* axis
[1] - dir
[1];
1340 result
[2] = tmpX
* axis
[2] - dir
[2];
1341 /* result[3] is never written! XXX enforce in parser! */
1342 store_vector4(inst
, machine
, result
);
1345 case OPCODE_RSQ
: /* 1 / sqrt() */
1347 GLfloat a
[4], result
[4];
1348 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1350 result
[0] = result
[1] = result
[2] = result
[3] = INV_SQRTF(a
[0]);
1351 store_vector4(inst
, machine
, result
);
1353 printf("RSQ %g = 1/sqrt(|%g|)\n", result
[0], a
[0]);
1357 case OPCODE_SCS
: /* sine and cos */
1359 GLfloat a
[4], result
[4];
1360 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1361 result
[0] = (GLfloat
) _mesa_cos(a
[0]);
1362 result
[1] = (GLfloat
) _mesa_sin(a
[0]);
1363 result
[2] = 0.0; /* undefined! */
1364 result
[3] = 0.0; /* undefined! */
1365 store_vector4(inst
, machine
, result
);
1368 case OPCODE_SEQ
: /* set on equal */
1370 GLfloat a
[4], b
[4], result
[4];
1371 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1372 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1373 result
[0] = (a
[0] == b
[0]) ? 1.0F
: 0.0F
;
1374 result
[1] = (a
[1] == b
[1]) ? 1.0F
: 0.0F
;
1375 result
[2] = (a
[2] == b
[2]) ? 1.0F
: 0.0F
;
1376 result
[3] = (a
[3] == b
[3]) ? 1.0F
: 0.0F
;
1377 store_vector4(inst
, machine
, result
);
1379 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1380 result
[0], result
[1], result
[2], result
[3],
1381 a
[0], a
[1], a
[2], a
[3],
1382 b
[0], b
[1], b
[2], b
[3]);
1386 case OPCODE_SFL
: /* set false, operands ignored */
1388 static const GLfloat result
[4] = { 0.0F
, 0.0F
, 0.0F
, 0.0F
};
1389 store_vector4(inst
, machine
, result
);
1392 case OPCODE_SGE
: /* set on greater or equal */
1394 GLfloat a
[4], b
[4], result
[4];
1395 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1396 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1397 result
[0] = (a
[0] >= b
[0]) ? 1.0F
: 0.0F
;
1398 result
[1] = (a
[1] >= b
[1]) ? 1.0F
: 0.0F
;
1399 result
[2] = (a
[2] >= b
[2]) ? 1.0F
: 0.0F
;
1400 result
[3] = (a
[3] >= b
[3]) ? 1.0F
: 0.0F
;
1401 store_vector4(inst
, machine
, result
);
1403 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1404 result
[0], result
[1], result
[2], result
[3],
1405 a
[0], a
[1], a
[2], a
[3],
1406 b
[0], b
[1], b
[2], b
[3]);
1410 case OPCODE_SGT
: /* set on greater */
1412 GLfloat a
[4], b
[4], result
[4];
1413 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1414 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1415 result
[0] = (a
[0] > b
[0]) ? 1.0F
: 0.0F
;
1416 result
[1] = (a
[1] > b
[1]) ? 1.0F
: 0.0F
;
1417 result
[2] = (a
[2] > b
[2]) ? 1.0F
: 0.0F
;
1418 result
[3] = (a
[3] > b
[3]) ? 1.0F
: 0.0F
;
1419 store_vector4(inst
, machine
, result
);
1421 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1422 result
[0], result
[1], result
[2], result
[3],
1423 a
[0], a
[1], a
[2], a
[3],
1424 b
[0], b
[1], b
[2], b
[3]);
1430 GLfloat a
[4], result
[4];
1431 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1432 result
[0] = result
[1] = result
[2] = result
[3]
1433 = (GLfloat
) _mesa_sin(a
[0]);
1434 store_vector4(inst
, machine
, result
);
1437 case OPCODE_SLE
: /* set on less or equal */
1439 GLfloat a
[4], b
[4], result
[4];
1440 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1441 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1442 result
[0] = (a
[0] <= b
[0]) ? 1.0F
: 0.0F
;
1443 result
[1] = (a
[1] <= b
[1]) ? 1.0F
: 0.0F
;
1444 result
[2] = (a
[2] <= b
[2]) ? 1.0F
: 0.0F
;
1445 result
[3] = (a
[3] <= b
[3]) ? 1.0F
: 0.0F
;
1446 store_vector4(inst
, machine
, result
);
1448 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1449 result
[0], result
[1], result
[2], result
[3],
1450 a
[0], a
[1], a
[2], a
[3],
1451 b
[0], b
[1], b
[2], b
[3]);
1455 case OPCODE_SLT
: /* set on less */
1457 GLfloat a
[4], b
[4], result
[4];
1458 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1459 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1460 result
[0] = (a
[0] < b
[0]) ? 1.0F
: 0.0F
;
1461 result
[1] = (a
[1] < b
[1]) ? 1.0F
: 0.0F
;
1462 result
[2] = (a
[2] < b
[2]) ? 1.0F
: 0.0F
;
1463 result
[3] = (a
[3] < b
[3]) ? 1.0F
: 0.0F
;
1464 store_vector4(inst
, machine
, result
);
1466 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1467 result
[0], result
[1], result
[2], result
[3],
1468 a
[0], a
[1], a
[2], a
[3],
1469 b
[0], b
[1], b
[2], b
[3]);
1473 case OPCODE_SNE
: /* set on not equal */
1475 GLfloat a
[4], b
[4], result
[4];
1476 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1477 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1478 result
[0] = (a
[0] != b
[0]) ? 1.0F
: 0.0F
;
1479 result
[1] = (a
[1] != b
[1]) ? 1.0F
: 0.0F
;
1480 result
[2] = (a
[2] != b
[2]) ? 1.0F
: 0.0F
;
1481 result
[3] = (a
[3] != b
[3]) ? 1.0F
: 0.0F
;
1482 store_vector4(inst
, machine
, result
);
1484 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1485 result
[0], result
[1], result
[2], result
[3],
1486 a
[0], a
[1], a
[2], a
[3],
1487 b
[0], b
[1], b
[2], b
[3]);
1491 case OPCODE_SSG
: /* set sign (-1, 0 or +1) */
1493 GLfloat a
[4], result
[4];
1494 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1495 result
[0] = (GLfloat
) ((a
[0] > 0.0F
) - (a
[0] < 0.0F
));
1496 result
[1] = (GLfloat
) ((a
[1] > 0.0F
) - (a
[1] < 0.0F
));
1497 result
[2] = (GLfloat
) ((a
[2] > 0.0F
) - (a
[2] < 0.0F
));
1498 result
[3] = (GLfloat
) ((a
[3] > 0.0F
) - (a
[3] < 0.0F
));
1499 store_vector4(inst
, machine
, result
);
1502 case OPCODE_STR
: /* set true, operands ignored */
1504 static const GLfloat result
[4] = { 1.0F
, 1.0F
, 1.0F
, 1.0F
};
1505 store_vector4(inst
, machine
, result
);
1510 GLfloat a
[4], b
[4], result
[4];
1511 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1512 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1513 result
[0] = a
[0] - b
[0];
1514 result
[1] = a
[1] - b
[1];
1515 result
[2] = a
[2] - b
[2];
1516 result
[3] = a
[3] - b
[3];
1517 store_vector4(inst
, machine
, result
);
1519 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1520 result
[0], result
[1], result
[2], result
[3],
1521 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
1525 case OPCODE_SWZ
: /* extended swizzle */
1527 const struct prog_src_register
*source
= &inst
->SrcReg
[0];
1528 const GLfloat
*src
= get_src_register_pointer(source
, machine
);
1531 for (i
= 0; i
< 4; i
++) {
1532 const GLuint swz
= GET_SWZ(source
->Swizzle
, i
);
1533 if (swz
== SWIZZLE_ZERO
)
1535 else if (swz
== SWIZZLE_ONE
)
1540 result
[i
] = src
[swz
];
1542 if (source
->Negate
& (1 << i
))
1543 result
[i
] = -result
[i
];
1545 store_vector4(inst
, machine
, result
);
1548 case OPCODE_TEX
: /* Both ARB and NV frag prog */
1549 /* Simple texel lookup */
1551 GLfloat texcoord
[4], color
[4];
1552 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1554 fetch_texel(ctx
, machine
, inst
, texcoord
, 0.0, color
);
1557 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1558 color
[0], color
[1], color
[2], color
[3],
1560 texcoord
[0], texcoord
[1], texcoord
[2], texcoord
[3]);
1562 store_vector4(inst
, machine
, color
);
1565 case OPCODE_TXB
: /* GL_ARB_fragment_program only */
1566 /* Texel lookup with LOD bias */
1568 GLfloat texcoord
[4], color
[4], lodBias
;
1570 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1572 /* texcoord[3] is the bias to add to lambda */
1573 lodBias
= texcoord
[3];
1575 fetch_texel(ctx
, machine
, inst
, texcoord
, lodBias
, color
);
1577 store_vector4(inst
, machine
, color
);
1580 case OPCODE_TXD
: /* GL_NV_fragment_program only */
1581 /* Texture lookup w/ partial derivatives for LOD */
1583 GLfloat texcoord
[4], dtdx
[4], dtdy
[4], color
[4];
1584 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1585 fetch_vector4(&inst
->SrcReg
[1], machine
, dtdx
);
1586 fetch_vector4(&inst
->SrcReg
[2], machine
, dtdy
);
1587 machine
->FetchTexelDeriv(ctx
, texcoord
, dtdx
, dtdy
,
1589 inst
->TexSrcUnit
, color
);
1590 store_vector4(inst
, machine
, color
);
1593 case OPCODE_TXP
: /* GL_ARB_fragment_program only */
1594 /* Texture lookup w/ projective divide */
1596 GLfloat texcoord
[4], color
[4];
1598 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1599 /* Not so sure about this test - if texcoord[3] is
1600 * zero, we'd probably be fine except for an ASSERT in
1601 * IROUND_POS() which gets triggered by the inf values created.
1603 if (texcoord
[3] != 0.0) {
1604 texcoord
[0] /= texcoord
[3];
1605 texcoord
[1] /= texcoord
[3];
1606 texcoord
[2] /= texcoord
[3];
1609 fetch_texel(ctx
, machine
, inst
, texcoord
, 0.0, color
);
1611 store_vector4(inst
, machine
, color
);
1614 case OPCODE_TXP_NV
: /* GL_NV_fragment_program only */
1615 /* Texture lookup w/ projective divide, as above, but do not
1616 * do the divide by w if sampling from a cube map.
1619 GLfloat texcoord
[4], color
[4];
1621 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1622 if (inst
->TexSrcTarget
!= TEXTURE_CUBE_INDEX
&&
1623 texcoord
[3] != 0.0) {
1624 texcoord
[0] /= texcoord
[3];
1625 texcoord
[1] /= texcoord
[3];
1626 texcoord
[2] /= texcoord
[3];
1629 fetch_texel(ctx
, machine
, inst
, texcoord
, 0.0, color
);
1631 store_vector4(inst
, machine
, color
);
1634 case OPCODE_TRUNC
: /* truncate toward zero */
1636 GLfloat a
[4], result
[4];
1637 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1638 result
[0] = (GLfloat
) (GLint
) a
[0];
1639 result
[1] = (GLfloat
) (GLint
) a
[1];
1640 result
[2] = (GLfloat
) (GLint
) a
[2];
1641 result
[3] = (GLfloat
) (GLint
) a
[3];
1642 store_vector4(inst
, machine
, result
);
1645 case OPCODE_UP2H
: /* unpack two 16-bit floats */
1647 GLfloat a
[4], result
[4];
1650 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1654 result
[0] = result
[2] = _mesa_half_to_float(hx
);
1655 result
[1] = result
[3] = _mesa_half_to_float(hy
);
1656 store_vector4(inst
, machine
, result
);
1659 case OPCODE_UP2US
: /* unpack two GLushorts */
1661 GLfloat a
[4], result
[4];
1664 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1666 usx
= fi
.i
& 0xffff;
1668 result
[0] = result
[2] = usx
* (1.0f
/ 65535.0f
);
1669 result
[1] = result
[3] = usy
* (1.0f
/ 65535.0f
);
1670 store_vector4(inst
, machine
, result
);
1673 case OPCODE_UP4B
: /* unpack four GLbytes */
1675 GLfloat a
[4], result
[4];
1677 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1679 result
[0] = (((fi
.i
>> 0) & 0xff) - 128) / 127.0F
;
1680 result
[1] = (((fi
.i
>> 8) & 0xff) - 128) / 127.0F
;
1681 result
[2] = (((fi
.i
>> 16) & 0xff) - 128) / 127.0F
;
1682 result
[3] = (((fi
.i
>> 24) & 0xff) - 128) / 127.0F
;
1683 store_vector4(inst
, machine
, result
);
1686 case OPCODE_UP4UB
: /* unpack four GLubytes */
1688 GLfloat a
[4], result
[4];
1690 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1692 result
[0] = ((fi
.i
>> 0) & 0xff) / 255.0F
;
1693 result
[1] = ((fi
.i
>> 8) & 0xff) / 255.0F
;
1694 result
[2] = ((fi
.i
>> 16) & 0xff) / 255.0F
;
1695 result
[3] = ((fi
.i
>> 24) & 0xff) / 255.0F
;
1696 store_vector4(inst
, machine
, result
);
1699 case OPCODE_XOR
: /* bitwise XOR */
1701 GLuint a
[4], b
[4], result
[4];
1702 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
1703 fetch_vector4ui(&inst
->SrcReg
[1], machine
, b
);
1704 result
[0] = a
[0] ^ b
[0];
1705 result
[1] = a
[1] ^ b
[1];
1706 result
[2] = a
[2] ^ b
[2];
1707 result
[3] = a
[3] ^ b
[3];
1708 store_vector4ui(inst
, machine
, result
);
1711 case OPCODE_XPD
: /* cross product */
1713 GLfloat a
[4], b
[4], result
[4];
1714 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1715 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1716 result
[0] = a
[1] * b
[2] - a
[2] * b
[1];
1717 result
[1] = a
[2] * b
[0] - a
[0] * b
[2];
1718 result
[2] = a
[0] * b
[1] - a
[1] * b
[0];
1720 store_vector4(inst
, machine
, result
);
1722 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1723 result
[0], result
[1], result
[2], result
[3],
1724 a
[0], a
[1], a
[2], b
[0], b
[1], b
[2]);
1728 case OPCODE_X2D
: /* 2-D matrix transform */
1730 GLfloat a
[4], b
[4], c
[4], result
[4];
1731 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1732 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1733 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
1734 result
[0] = a
[0] + b
[0] * c
[0] + b
[1] * c
[1];
1735 result
[1] = a
[1] + b
[0] * c
[2] + b
[1] * c
[3];
1736 result
[2] = a
[2] + b
[0] * c
[0] + b
[1] * c
[1];
1737 result
[3] = a
[3] + b
[0] * c
[2] + b
[1] * c
[3];
1738 store_vector4(inst
, machine
, result
);
1743 if (inst
->SrcReg
[0].File
!= -1) {
1745 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1746 _mesa_printf("%s%g, %g, %g, %g\n", (const char *) inst
->Data
,
1747 a
[0], a
[1], a
[2], a
[3]);
1750 _mesa_printf("%s\n", (const char *) inst
->Data
);
1757 _mesa_problem(ctx
, "Bad opcode %d in _mesa_execute_program",
1759 return GL_TRUE
; /* return value doesn't matter */
1763 if (numExec
> maxExec
) {
1764 _mesa_problem(ctx
, "Infinite loop detected in fragment program");