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) ( *((GLuint *) (void *)&x) = 0x7F800000 )
58 #define SET_NEG_INFINITY(x) ( *((GLuint *) (void *)&x) = 0xFF800000 )
60 #define SET_POS_INFINITY(x) x = __MAXFLOAT
61 #define SET_NEG_INFINITY(x) x = -__MAXFLOAT
63 #define SET_POS_INFINITY(x) x = (GLfloat) HUGE_VAL
64 #define SET_NEG_INFINITY(x) x = (GLfloat) -HUGE_VAL
67 #define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
70 static const GLfloat ZeroVec
[4] = { 0.0F
, 0.0F
, 0.0F
, 0.0F
};
75 * Return a pointer to the 4-element float vector specified by the given
78 static INLINE
const GLfloat
*
79 get_src_register_pointer(const struct prog_src_register
*source
,
80 const struct gl_program_machine
*machine
)
82 const struct gl_program
*prog
= machine
->CurProgram
;
83 GLint reg
= source
->Index
;
85 if (source
->RelAddr
) {
86 /* add address register value to src index/offset */
87 reg
+= machine
->AddressReg
[0][0];
93 switch (source
->File
) {
94 case PROGRAM_TEMPORARY
:
95 if (reg
>= MAX_PROGRAM_TEMPS
)
97 return machine
->Temporaries
[reg
];
100 if (prog
->Target
== GL_VERTEX_PROGRAM_ARB
) {
101 if (reg
>= VERT_ATTRIB_MAX
)
103 return machine
->VertAttribs
[reg
];
106 if (reg
>= FRAG_ATTRIB_MAX
)
108 return machine
->Attribs
[reg
][machine
->CurElement
];
112 if (reg
>= MAX_PROGRAM_OUTPUTS
)
114 return machine
->Outputs
[reg
];
116 case PROGRAM_LOCAL_PARAM
:
117 if (reg
>= MAX_PROGRAM_LOCAL_PARAMS
)
119 return machine
->CurProgram
->LocalParams
[reg
];
121 case PROGRAM_ENV_PARAM
:
122 if (reg
>= MAX_PROGRAM_ENV_PARAMS
)
124 return machine
->EnvParams
[reg
];
126 case PROGRAM_STATE_VAR
:
128 case PROGRAM_CONSTANT
:
130 case PROGRAM_UNIFORM
:
132 case PROGRAM_NAMED_PARAM
:
133 if (reg
>= (GLint
) prog
->Parameters
->NumParameters
)
135 return prog
->Parameters
->ParameterValues
[reg
];
139 "Invalid src register file %d in get_src_register_pointer()",
147 * Return a pointer to the 4-element float vector specified by the given
148 * destination register.
150 static INLINE GLfloat
*
151 get_dst_register_pointer(const struct prog_dst_register
*dest
,
152 struct gl_program_machine
*machine
)
154 static GLfloat dummyReg
[4];
155 GLint reg
= dest
->Index
;
158 /* add address register value to src index/offset */
159 reg
+= machine
->AddressReg
[0][0];
165 switch (dest
->File
) {
166 case PROGRAM_TEMPORARY
:
167 if (reg
>= MAX_PROGRAM_TEMPS
)
169 return machine
->Temporaries
[reg
];
172 if (reg
>= MAX_PROGRAM_OUTPUTS
)
174 return machine
->Outputs
[reg
];
176 case PROGRAM_WRITE_ONLY
:
181 "Invalid dest register file %d in get_dst_register_pointer()",
189 #if FEATURE_MESA_program_debug
190 static struct gl_program_machine
*CurrentMachine
= NULL
;
193 * For GL_MESA_program_debug.
194 * Return current value (4*GLfloat) of a program register.
195 * Called via ctx->Driver.GetProgramRegister().
198 _mesa_get_program_register(GLcontext
*ctx
, enum register_file file
,
199 GLuint index
, GLfloat val
[4])
201 if (CurrentMachine
) {
202 struct prog_src_register srcReg
;
205 srcReg
.Index
= index
;
206 src
= get_src_register_pointer(&srcReg
, CurrentMachine
);
210 #endif /* FEATURE_MESA_program_debug */
214 * Fetch a 4-element float vector from the given source register.
215 * Apply swizzling and negating as needed.
218 fetch_vector4(const struct prog_src_register
*source
,
219 const struct gl_program_machine
*machine
, GLfloat result
[4])
221 const GLfloat
*src
= get_src_register_pointer(source
, machine
);
224 if (source
->Swizzle
== SWIZZLE_NOOP
) {
226 COPY_4V(result
, src
);
229 ASSERT(GET_SWZ(source
->Swizzle
, 0) <= 3);
230 ASSERT(GET_SWZ(source
->Swizzle
, 1) <= 3);
231 ASSERT(GET_SWZ(source
->Swizzle
, 2) <= 3);
232 ASSERT(GET_SWZ(source
->Swizzle
, 3) <= 3);
233 result
[0] = src
[GET_SWZ(source
->Swizzle
, 0)];
234 result
[1] = src
[GET_SWZ(source
->Swizzle
, 1)];
235 result
[2] = src
[GET_SWZ(source
->Swizzle
, 2)];
236 result
[3] = src
[GET_SWZ(source
->Swizzle
, 3)];
239 if (source
->NegateBase
) {
240 result
[0] = -result
[0];
241 result
[1] = -result
[1];
242 result
[2] = -result
[2];
243 result
[3] = -result
[3];
246 result
[0] = FABSF(result
[0]);
247 result
[1] = FABSF(result
[1]);
248 result
[2] = FABSF(result
[2]);
249 result
[3] = FABSF(result
[3]);
251 if (source
->NegateAbs
) {
252 result
[0] = -result
[0];
253 result
[1] = -result
[1];
254 result
[2] = -result
[2];
255 result
[3] = -result
[3];
261 * Fetch a 4-element uint vector from the given source register.
262 * Apply swizzling but not negation/abs.
265 fetch_vector4ui(const struct prog_src_register
*source
,
266 const struct gl_program_machine
*machine
, GLuint result
[4])
268 const GLuint
*src
= (GLuint
*) get_src_register_pointer(source
, machine
);
271 if (source
->Swizzle
== SWIZZLE_NOOP
) {
273 COPY_4V(result
, src
);
276 ASSERT(GET_SWZ(source
->Swizzle
, 0) <= 3);
277 ASSERT(GET_SWZ(source
->Swizzle
, 1) <= 3);
278 ASSERT(GET_SWZ(source
->Swizzle
, 2) <= 3);
279 ASSERT(GET_SWZ(source
->Swizzle
, 3) <= 3);
280 result
[0] = src
[GET_SWZ(source
->Swizzle
, 0)];
281 result
[1] = src
[GET_SWZ(source
->Swizzle
, 1)];
282 result
[2] = src
[GET_SWZ(source
->Swizzle
, 2)];
283 result
[3] = src
[GET_SWZ(source
->Swizzle
, 3)];
286 /* Note: no NegateBase, Abs, NegateAbs here */
292 * Fetch the derivative with respect to X or Y for the given register.
293 * XXX this currently only works for fragment program input attribs.
296 fetch_vector4_deriv(GLcontext
* ctx
,
297 const struct prog_src_register
*source
,
298 const struct gl_program_machine
*machine
,
299 char xOrY
, GLfloat result
[4])
301 if (source
->File
== PROGRAM_INPUT
&&
302 source
->Index
< (GLint
) machine
->NumDeriv
) {
303 const GLint col
= machine
->CurElement
;
304 const GLfloat w
= machine
->Attribs
[FRAG_ATTRIB_WPOS
][col
][3];
305 const GLfloat invQ
= 1.0f
/ w
;
309 deriv
[0] = machine
->DerivX
[source
->Index
][0] * invQ
;
310 deriv
[1] = machine
->DerivX
[source
->Index
][1] * invQ
;
311 deriv
[2] = machine
->DerivX
[source
->Index
][2] * invQ
;
312 deriv
[3] = machine
->DerivX
[source
->Index
][3] * invQ
;
315 deriv
[0] = machine
->DerivY
[source
->Index
][0] * invQ
;
316 deriv
[1] = machine
->DerivY
[source
->Index
][1] * invQ
;
317 deriv
[2] = machine
->DerivY
[source
->Index
][2] * invQ
;
318 deriv
[3] = machine
->DerivY
[source
->Index
][3] * invQ
;
321 result
[0] = deriv
[GET_SWZ(source
->Swizzle
, 0)];
322 result
[1] = deriv
[GET_SWZ(source
->Swizzle
, 1)];
323 result
[2] = deriv
[GET_SWZ(source
->Swizzle
, 2)];
324 result
[3] = deriv
[GET_SWZ(source
->Swizzle
, 3)];
326 if (source
->NegateBase
) {
327 result
[0] = -result
[0];
328 result
[1] = -result
[1];
329 result
[2] = -result
[2];
330 result
[3] = -result
[3];
333 result
[0] = FABSF(result
[0]);
334 result
[1] = FABSF(result
[1]);
335 result
[2] = FABSF(result
[2]);
336 result
[3] = FABSF(result
[3]);
338 if (source
->NegateAbs
) {
339 result
[0] = -result
[0];
340 result
[1] = -result
[1];
341 result
[2] = -result
[2];
342 result
[3] = -result
[3];
346 ASSIGN_4V(result
, 0.0, 0.0, 0.0, 0.0);
352 * As above, but only return result[0] element.
355 fetch_vector1(const struct prog_src_register
*source
,
356 const struct gl_program_machine
*machine
, GLfloat result
[4])
358 const GLfloat
*src
= get_src_register_pointer(source
, machine
);
361 result
[0] = src
[GET_SWZ(source
->Swizzle
, 0)];
363 if (source
->NegateBase
) {
364 result
[0] = -result
[0];
367 result
[0] = FABSF(result
[0]);
369 if (source
->NegateAbs
) {
370 result
[0] = -result
[0];
376 * Fetch texel from texture. Use partial derivatives when possible.
379 fetch_texel(GLcontext
*ctx
,
380 const struct gl_program_machine
*machine
,
381 const struct prog_instruction
*inst
,
382 const GLfloat texcoord
[4], GLfloat lodBias
,
385 const GLuint unit
= machine
->Samplers
[inst
->TexSrcUnit
];
387 /* Note: we only have the right derivatives for fragment input attribs.
389 if (machine
->NumDeriv
> 0 &&
390 inst
->SrcReg
[0].File
== PROGRAM_INPUT
&&
391 inst
->SrcReg
[0].Index
== FRAG_ATTRIB_TEX0
+ inst
->TexSrcUnit
) {
392 /* simple texture fetch for which we should have derivatives */
393 GLuint attr
= inst
->SrcReg
[0].Index
;
394 machine
->FetchTexelDeriv(ctx
, texcoord
,
395 machine
->DerivX
[attr
],
396 machine
->DerivY
[attr
],
397 lodBias
, unit
, color
);
400 machine
->FetchTexelLod(ctx
, texcoord
, lodBias
, unit
, color
);
406 * Test value against zero and return GT, LT, EQ or UN if NaN.
409 generate_cc(float value
)
412 return COND_UN
; /* NaN */
422 * Test if the ccMaskRule is satisfied by the given condition code.
423 * Used to mask destination writes according to the current condition code.
425 static INLINE GLboolean
426 test_cc(GLuint condCode
, GLuint ccMaskRule
)
428 switch (ccMaskRule
) {
429 case COND_EQ
: return (condCode
== COND_EQ
);
430 case COND_NE
: return (condCode
!= COND_EQ
);
431 case COND_LT
: return (condCode
== COND_LT
);
432 case COND_GE
: return (condCode
== COND_GT
|| condCode
== COND_EQ
);
433 case COND_LE
: return (condCode
== COND_LT
|| condCode
== COND_EQ
);
434 case COND_GT
: return (condCode
== COND_GT
);
435 case COND_TR
: return GL_TRUE
;
436 case COND_FL
: return GL_FALSE
;
437 default: return GL_TRUE
;
443 * Evaluate the 4 condition codes against a predicate and return GL_TRUE
444 * or GL_FALSE to indicate result.
446 static INLINE GLboolean
447 eval_condition(const struct gl_program_machine
*machine
,
448 const struct prog_instruction
*inst
)
450 const GLuint swizzle
= inst
->DstReg
.CondSwizzle
;
451 const GLuint condMask
= inst
->DstReg
.CondMask
;
452 if (test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 0)], condMask
) ||
453 test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 1)], condMask
) ||
454 test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 2)], condMask
) ||
455 test_cc(machine
->CondCodes
[GET_SWZ(swizzle
, 3)], condMask
)) {
466 * Store 4 floats into a register. Observe the instructions saturate and
467 * set-condition-code flags.
470 store_vector4(const struct prog_instruction
*inst
,
471 struct gl_program_machine
*machine
, const GLfloat value
[4])
473 const struct prog_dst_register
*dstReg
= &(inst
->DstReg
);
474 const GLboolean clamp
= inst
->SaturateMode
== SATURATE_ZERO_ONE
;
475 GLuint writeMask
= dstReg
->WriteMask
;
476 GLfloat clampedValue
[4];
477 GLfloat
*dst
= get_dst_register_pointer(dstReg
, machine
);
480 if (value
[0] > 1.0e10
||
481 IS_INF_OR_NAN(value
[0]) ||
482 IS_INF_OR_NAN(value
[1]) ||
483 IS_INF_OR_NAN(value
[2]) || IS_INF_OR_NAN(value
[3]))
484 printf("store %g %g %g %g\n", value
[0], value
[1], value
[2], value
[3]);
488 clampedValue
[0] = CLAMP(value
[0], 0.0F
, 1.0F
);
489 clampedValue
[1] = CLAMP(value
[1], 0.0F
, 1.0F
);
490 clampedValue
[2] = CLAMP(value
[2], 0.0F
, 1.0F
);
491 clampedValue
[3] = CLAMP(value
[3], 0.0F
, 1.0F
);
492 value
= clampedValue
;
495 if (dstReg
->CondMask
!= COND_TR
) {
496 /* condition codes may turn off some writes */
497 if (writeMask
& WRITEMASK_X
) {
498 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 0)],
500 writeMask
&= ~WRITEMASK_X
;
502 if (writeMask
& WRITEMASK_Y
) {
503 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 1)],
505 writeMask
&= ~WRITEMASK_Y
;
507 if (writeMask
& WRITEMASK_Z
) {
508 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 2)],
510 writeMask
&= ~WRITEMASK_Z
;
512 if (writeMask
& WRITEMASK_W
) {
513 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 3)],
515 writeMask
&= ~WRITEMASK_W
;
519 if (writeMask
& WRITEMASK_X
)
521 if (writeMask
& WRITEMASK_Y
)
523 if (writeMask
& WRITEMASK_Z
)
525 if (writeMask
& WRITEMASK_W
)
528 if (inst
->CondUpdate
) {
529 if (writeMask
& WRITEMASK_X
)
530 machine
->CondCodes
[0] = generate_cc(value
[0]);
531 if (writeMask
& WRITEMASK_Y
)
532 machine
->CondCodes
[1] = generate_cc(value
[1]);
533 if (writeMask
& WRITEMASK_Z
)
534 machine
->CondCodes
[2] = generate_cc(value
[2]);
535 if (writeMask
& WRITEMASK_W
)
536 machine
->CondCodes
[3] = generate_cc(value
[3]);
538 printf("CondCodes=(%s,%s,%s,%s) for:\n",
539 _mesa_condcode_string(machine
->CondCodes
[0]),
540 _mesa_condcode_string(machine
->CondCodes
[1]),
541 _mesa_condcode_string(machine
->CondCodes
[2]),
542 _mesa_condcode_string(machine
->CondCodes
[3]));
549 * Store 4 uints into a register. Observe the set-condition-code flags.
552 store_vector4ui(const struct prog_instruction
*inst
,
553 struct gl_program_machine
*machine
, const GLuint value
[4])
555 const struct prog_dst_register
*dstReg
= &(inst
->DstReg
);
556 GLuint writeMask
= dstReg
->WriteMask
;
557 GLuint
*dst
= (GLuint
*) get_dst_register_pointer(dstReg
, machine
);
559 if (dstReg
->CondMask
!= COND_TR
) {
560 /* condition codes may turn off some writes */
561 if (writeMask
& WRITEMASK_X
) {
562 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 0)],
564 writeMask
&= ~WRITEMASK_X
;
566 if (writeMask
& WRITEMASK_Y
) {
567 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 1)],
569 writeMask
&= ~WRITEMASK_Y
;
571 if (writeMask
& WRITEMASK_Z
) {
572 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 2)],
574 writeMask
&= ~WRITEMASK_Z
;
576 if (writeMask
& WRITEMASK_W
) {
577 if (!test_cc(machine
->CondCodes
[GET_SWZ(dstReg
->CondSwizzle
, 3)],
579 writeMask
&= ~WRITEMASK_W
;
583 if (writeMask
& WRITEMASK_X
)
585 if (writeMask
& WRITEMASK_Y
)
587 if (writeMask
& WRITEMASK_Z
)
589 if (writeMask
& WRITEMASK_W
)
592 if (inst
->CondUpdate
) {
593 if (writeMask
& WRITEMASK_X
)
594 machine
->CondCodes
[0] = generate_cc(value
[0]);
595 if (writeMask
& WRITEMASK_Y
)
596 machine
->CondCodes
[1] = generate_cc(value
[1]);
597 if (writeMask
& WRITEMASK_Z
)
598 machine
->CondCodes
[2] = generate_cc(value
[2]);
599 if (writeMask
& WRITEMASK_W
)
600 machine
->CondCodes
[3] = generate_cc(value
[3]);
602 printf("CondCodes=(%s,%s,%s,%s) for:\n",
603 _mesa_condcode_string(machine
->CondCodes
[0]),
604 _mesa_condcode_string(machine
->CondCodes
[1]),
605 _mesa_condcode_string(machine
->CondCodes
[2]),
606 _mesa_condcode_string(machine
->CondCodes
[3]));
614 * Execute the given vertex/fragment program.
616 * \param ctx rendering context
617 * \param program the program to execute
618 * \param machine machine state (must be initialized)
619 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
622 _mesa_execute_program(GLcontext
* ctx
,
623 const struct gl_program
*program
,
624 struct gl_program_machine
*machine
)
626 const GLuint numInst
= program
->NumInstructions
;
627 const GLuint maxExec
= 10000;
628 GLuint pc
, numExec
= 0;
630 machine
->CurProgram
= program
;
633 printf("execute program %u --------------------\n", program
->Id
);
636 #if FEATURE_MESA_program_debug
637 CurrentMachine
= machine
;
640 if (program
->Target
== GL_VERTEX_PROGRAM_ARB
) {
641 machine
->EnvParams
= ctx
->VertexProgram
.Parameters
;
644 machine
->EnvParams
= ctx
->FragmentProgram
.Parameters
;
647 for (pc
= 0; pc
< numInst
; pc
++) {
648 const struct prog_instruction
*inst
= program
->Instructions
+ pc
;
650 #if FEATURE_MESA_program_debug
651 if (ctx
->FragmentProgram
.CallbackEnabled
&&
652 ctx
->FragmentProgram
.Callback
) {
653 ctx
->FragmentProgram
.CurrentPosition
= inst
->StringPos
;
654 ctx
->FragmentProgram
.Callback(program
->Target
,
655 ctx
->FragmentProgram
.CallbackData
);
660 _mesa_print_instruction(inst
);
663 switch (inst
->Opcode
) {
666 GLfloat a
[4], result
[4];
667 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
668 result
[0] = FABSF(a
[0]);
669 result
[1] = FABSF(a
[1]);
670 result
[2] = FABSF(a
[2]);
671 result
[3] = FABSF(a
[3]);
672 store_vector4(inst
, machine
, result
);
677 GLfloat a
[4], b
[4], result
[4];
678 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
679 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
680 result
[0] = a
[0] + b
[0];
681 result
[1] = a
[1] + b
[1];
682 result
[2] = a
[2] + b
[2];
683 result
[3] = a
[3] + b
[3];
684 store_vector4(inst
, machine
, result
);
686 printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
687 result
[0], result
[1], result
[2], result
[3],
688 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
692 case OPCODE_AND
: /* bitwise AND */
694 GLuint a
[4], b
[4], result
[4];
695 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
696 fetch_vector4ui(&inst
->SrcReg
[1], machine
, b
);
697 result
[0] = a
[0] & b
[0];
698 result
[1] = a
[1] & b
[1];
699 result
[2] = a
[2] & b
[2];
700 result
[3] = a
[3] & b
[3];
701 store_vector4ui(inst
, machine
, result
);
707 fetch_vector4(&inst
->SrcReg
[0], machine
, t
);
708 machine
->AddressReg
[0][0] = IFLOOR(t
[0]);
715 /* subtract 1 here since pc is incremented by for(pc) loop */
716 pc
= inst
->BranchTarget
- 1; /* go to matching BNGLOOP */
718 case OPCODE_BGNSUB
: /* begin subroutine */
720 case OPCODE_ENDSUB
: /* end subroutine */
722 case OPCODE_BRA
: /* branch (conditional) */
724 case OPCODE_BRK
: /* break out of loop (conditional) */
726 case OPCODE_CONT
: /* continue loop (conditional) */
727 if (eval_condition(machine
, inst
)) {
729 /* Subtract 1 here since we'll do pc++ at end of for-loop */
730 pc
= inst
->BranchTarget
- 1;
733 case OPCODE_CAL
: /* Call subroutine (conditional) */
734 if (eval_condition(machine
, inst
)) {
735 /* call the subroutine */
736 if (machine
->StackDepth
>= MAX_PROGRAM_CALL_DEPTH
) {
737 return GL_TRUE
; /* Per GL_NV_vertex_program2 spec */
739 machine
->CallStack
[machine
->StackDepth
++] = pc
+ 1; /* next inst */
740 /* Subtract 1 here since we'll do pc++ at end of for-loop */
741 pc
= inst
->BranchTarget
- 1;
746 GLfloat a
[4], b
[4], c
[4], result
[4];
747 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
748 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
749 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
750 result
[0] = a
[0] < 0.0F
? b
[0] : c
[0];
751 result
[1] = a
[1] < 0.0F
? b
[1] : c
[1];
752 result
[2] = a
[2] < 0.0F
? b
[2] : c
[2];
753 result
[3] = a
[3] < 0.0F
? b
[3] : c
[3];
754 store_vector4(inst
, machine
, result
);
759 GLfloat a
[4], result
[4];
760 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
761 result
[0] = result
[1] = result
[2] = result
[3]
762 = (GLfloat
) _mesa_cos(a
[0]);
763 store_vector4(inst
, machine
, result
);
766 case OPCODE_DDX
: /* Partial derivative with respect to X */
769 fetch_vector4_deriv(ctx
, &inst
->SrcReg
[0], machine
,
771 store_vector4(inst
, machine
, result
);
774 case OPCODE_DDY
: /* Partial derivative with respect to Y */
777 fetch_vector4_deriv(ctx
, &inst
->SrcReg
[0], machine
,
779 store_vector4(inst
, machine
, result
);
784 GLfloat a
[4], b
[4], result
[4];
785 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
786 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
787 result
[0] = result
[1] = result
[2] = result
[3] = DOT2(a
, b
);
788 store_vector4(inst
, machine
, result
);
790 printf("DP2 %g = (%g %g) . (%g %g)\n",
791 result
[0], a
[0], a
[1], b
[0], b
[1]);
797 GLfloat a
[4], b
[4], c
, result
[4];
798 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
799 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
800 fetch_vector1(&inst
->SrcReg
[1], machine
, &c
);
801 result
[0] = result
[1] = result
[2] = result
[3] = DOT2(a
, b
) + c
;
802 store_vector4(inst
, machine
, result
);
804 printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
805 result
[0], a
[0], a
[1], b
[0], b
[1], c
);
811 GLfloat a
[4], b
[4], result
[4];
812 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
813 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
814 result
[0] = result
[1] = result
[2] = result
[3] = DOT3(a
, b
);
815 store_vector4(inst
, machine
, result
);
817 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
818 result
[0], a
[0], a
[1], a
[2], b
[0], b
[1], b
[2]);
824 GLfloat a
[4], b
[4], result
[4];
825 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
826 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
827 result
[0] = result
[1] = result
[2] = result
[3] = DOT4(a
, b
);
828 store_vector4(inst
, machine
, result
);
830 printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
831 result
[0], a
[0], a
[1], a
[2], a
[3],
832 b
[0], b
[1], b
[2], b
[3]);
838 GLfloat a
[4], b
[4], result
[4];
839 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
840 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
841 result
[0] = result
[1] = result
[2] = result
[3] = DOT3(a
, b
) + b
[3];
842 store_vector4(inst
, machine
, result
);
845 case OPCODE_DST
: /* Distance vector */
847 GLfloat a
[4], b
[4], result
[4];
848 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
849 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
851 result
[1] = a
[1] * b
[1];
854 store_vector4(inst
, machine
, result
);
859 GLfloat t
[4], q
[4], floor_t0
;
860 fetch_vector1(&inst
->SrcReg
[0], machine
, t
);
861 floor_t0
= FLOORF(t
[0]);
862 if (floor_t0
> FLT_MAX_EXP
) {
863 SET_POS_INFINITY(q
[0]);
864 SET_POS_INFINITY(q
[2]);
866 else if (floor_t0
< FLT_MIN_EXP
) {
871 q
[0] = LDEXPF(1.0, (int) floor_t0
);
872 /* Note: GL_NV_vertex_program expects
873 * result.z = result.x * APPX(result.y)
874 * We do what the ARB extension says.
876 q
[2] = (GLfloat
) pow(2.0, t
[0]);
878 q
[1] = t
[0] - floor_t0
;
880 store_vector4( inst
, machine
, q
);
883 case OPCODE_EX2
: /* Exponential base 2 */
885 GLfloat a
[4], result
[4];
886 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
887 result
[0] = result
[1] = result
[2] = result
[3] =
888 (GLfloat
) _mesa_pow(2.0, a
[0]);
889 store_vector4(inst
, machine
, result
);
894 GLfloat a
[4], result
[4];
895 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
896 result
[0] = FLOORF(a
[0]);
897 result
[1] = FLOORF(a
[1]);
898 result
[2] = FLOORF(a
[2]);
899 result
[3] = FLOORF(a
[3]);
900 store_vector4(inst
, machine
, result
);
905 GLfloat a
[4], result
[4];
906 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
907 result
[0] = a
[0] - FLOORF(a
[0]);
908 result
[1] = a
[1] - FLOORF(a
[1]);
909 result
[2] = a
[2] - FLOORF(a
[2]);
910 result
[3] = a
[3] - FLOORF(a
[3]);
911 store_vector4(inst
, machine
, result
);
918 if (inst
->SrcReg
[0].File
!= PROGRAM_UNDEFINED
) {
920 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
921 cond
= (a
[0] != 0.0);
924 cond
= eval_condition(machine
, inst
);
927 printf("IF: %d\n", cond
);
931 /* do if-clause (just continue execution) */
934 /* go to the instruction after ELSE or ENDIF */
935 assert(inst
->BranchTarget
>= 0);
936 pc
= inst
->BranchTarget
- 1;
942 assert(inst
->BranchTarget
>= 0);
943 pc
= inst
->BranchTarget
- 1;
948 case OPCODE_KIL_NV
: /* NV_f_p only (conditional) */
949 if (eval_condition(machine
, inst
)) {
953 case OPCODE_KIL
: /* ARB_f_p only */
956 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
957 if (a
[0] < 0.0F
|| a
[1] < 0.0F
|| a
[2] < 0.0F
|| a
[3] < 0.0F
) {
962 case OPCODE_LG2
: /* log base 2 */
964 GLfloat a
[4], result
[4];
965 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
966 /* The fast LOG2 macro doesn't meet the precision requirements.
968 result
[0] = result
[1] = result
[2] = result
[3] =
969 (log(a
[0]) * 1.442695F
);
970 store_vector4(inst
, machine
, result
);
975 const GLfloat epsilon
= 1.0F
/ 256.0F
; /* from NV VP spec */
976 GLfloat a
[4], result
[4];
977 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
978 a
[0] = MAX2(a
[0], 0.0F
);
979 a
[1] = MAX2(a
[1], 0.0F
);
980 /* XXX ARB version clamps a[3], NV version doesn't */
981 a
[3] = CLAMP(a
[3], -(128.0F
- epsilon
), (128.0F
- epsilon
));
984 /* XXX we could probably just use pow() here */
986 if (a
[1] == 0.0 && a
[3] == 0.0)
989 result
[2] = EXPF(a
[3] * LOGF(a
[1]));
995 store_vector4(inst
, machine
, result
);
997 printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
998 result
[0], result
[1], result
[2], result
[3],
999 a
[0], a
[1], a
[2], a
[3]);
1005 GLfloat t
[4], q
[4], abs_t0
;
1006 fetch_vector1(&inst
->SrcReg
[0], machine
, t
);
1007 abs_t0
= FABSF(t
[0]);
1008 if (abs_t0
!= 0.0F
) {
1009 /* Since we really can't handle infinite values on VMS
1010 * like other OSes we'll use __MAXFLOAT to represent
1011 * infinity. This may need some tweaking.
1014 if (abs_t0
== __MAXFLOAT
)
1016 if (IS_INF_OR_NAN(abs_t0
))
1019 SET_POS_INFINITY(q
[0]);
1021 SET_POS_INFINITY(q
[2]);
1025 GLfloat mantissa
= FREXPF(t
[0], &exponent
);
1026 q
[0] = (GLfloat
) (exponent
- 1);
1027 q
[1] = (GLfloat
) (2.0 * mantissa
); /* map [.5, 1) -> [1, 2) */
1029 /* The fast LOG2 macro doesn't meet the precision
1032 q
[2] = (log(t
[0]) * 1.442695F
);
1036 SET_NEG_INFINITY(q
[0]);
1038 SET_NEG_INFINITY(q
[2]);
1041 store_vector4(inst
, machine
, q
);
1046 GLfloat a
[4], b
[4], c
[4], result
[4];
1047 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1048 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1049 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
1050 result
[0] = a
[0] * b
[0] + (1.0F
- a
[0]) * c
[0];
1051 result
[1] = a
[1] * b
[1] + (1.0F
- a
[1]) * c
[1];
1052 result
[2] = a
[2] * b
[2] + (1.0F
- a
[2]) * c
[2];
1053 result
[3] = a
[3] * b
[3] + (1.0F
- a
[3]) * c
[3];
1054 store_vector4(inst
, machine
, result
);
1056 printf("LRP (%g %g %g %g) = (%g %g %g %g), "
1057 "(%g %g %g %g), (%g %g %g %g)\n",
1058 result
[0], result
[1], result
[2], result
[3],
1059 a
[0], a
[1], a
[2], a
[3],
1060 b
[0], b
[1], b
[2], b
[3], c
[0], c
[1], c
[2], c
[3]);
1066 GLfloat a
[4], b
[4], c
[4], result
[4];
1067 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1068 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1069 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
1070 result
[0] = a
[0] * b
[0] + c
[0];
1071 result
[1] = a
[1] * b
[1] + c
[1];
1072 result
[2] = a
[2] * b
[2] + c
[2];
1073 result
[3] = a
[3] * b
[3] + c
[3];
1074 store_vector4(inst
, machine
, result
);
1076 printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
1077 "(%g %g %g %g) + (%g %g %g %g)\n",
1078 result
[0], result
[1], result
[2], result
[3],
1079 a
[0], a
[1], a
[2], a
[3],
1080 b
[0], b
[1], b
[2], b
[3], c
[0], c
[1], c
[2], c
[3]);
1086 GLfloat a
[4], b
[4], result
[4];
1087 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1088 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1089 result
[0] = MAX2(a
[0], b
[0]);
1090 result
[1] = MAX2(a
[1], b
[1]);
1091 result
[2] = MAX2(a
[2], b
[2]);
1092 result
[3] = MAX2(a
[3], b
[3]);
1093 store_vector4(inst
, machine
, result
);
1095 printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
1096 result
[0], result
[1], result
[2], result
[3],
1097 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
1103 GLfloat a
[4], b
[4], result
[4];
1104 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1105 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1106 result
[0] = MIN2(a
[0], b
[0]);
1107 result
[1] = MIN2(a
[1], b
[1]);
1108 result
[2] = MIN2(a
[2], b
[2]);
1109 result
[3] = MIN2(a
[3], b
[3]);
1110 store_vector4(inst
, machine
, result
);
1116 fetch_vector4(&inst
->SrcReg
[0], machine
, result
);
1117 store_vector4(inst
, machine
, result
);
1119 printf("MOV (%g %g %g %g)\n",
1120 result
[0], result
[1], result
[2], result
[3]);
1126 GLfloat a
[4], b
[4], result
[4];
1127 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1128 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1129 result
[0] = a
[0] * b
[0];
1130 result
[1] = a
[1] * b
[1];
1131 result
[2] = a
[2] * b
[2];
1132 result
[3] = a
[3] * b
[3];
1133 store_vector4(inst
, machine
, result
);
1135 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
1136 result
[0], result
[1], result
[2], result
[3],
1137 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
1143 GLfloat a
[4], result
[4];
1144 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1148 result
[3] = _mesa_noise1(a
[0]);
1149 store_vector4(inst
, machine
, result
);
1154 GLfloat a
[4], result
[4];
1155 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1158 result
[2] = result
[3] = _mesa_noise2(a
[0], a
[1]);
1159 store_vector4(inst
, machine
, result
);
1164 GLfloat a
[4], result
[4];
1165 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1169 result
[3] = _mesa_noise3(a
[0], a
[1], a
[2]);
1170 store_vector4(inst
, machine
, result
);
1175 GLfloat a
[4], result
[4];
1176 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1180 result
[3] = _mesa_noise4(a
[0], a
[1], a
[2], a
[3]);
1181 store_vector4(inst
, machine
, result
);
1186 case OPCODE_NOT
: /* bitwise NOT */
1188 GLuint a
[4], result
[4];
1189 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
1194 store_vector4ui(inst
, machine
, result
);
1197 case OPCODE_NRM3
: /* 3-component normalization */
1199 GLfloat a
[4], result
[4];
1201 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1202 tmp
= a
[0] * a
[0] + a
[1] * a
[1] + a
[2] * a
[2];
1204 tmp
= INV_SQRTF(tmp
);
1205 result
[0] = tmp
* a
[0];
1206 result
[1] = tmp
* a
[1];
1207 result
[2] = tmp
* a
[2];
1208 result
[3] = 0.0; /* undefined, but prevent valgrind warnings */
1209 store_vector4(inst
, machine
, result
);
1212 case OPCODE_NRM4
: /* 4-component normalization */
1214 GLfloat a
[4], result
[4];
1216 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1217 tmp
= a
[0] * a
[0] + a
[1] * a
[1] + a
[2] * a
[2] + a
[3] * a
[3];
1219 tmp
= INV_SQRTF(tmp
);
1220 result
[0] = tmp
* a
[0];
1221 result
[1] = tmp
* a
[1];
1222 result
[2] = tmp
* a
[2];
1223 result
[3] = tmp
* a
[3];
1224 store_vector4(inst
, machine
, result
);
1227 case OPCODE_OR
: /* bitwise OR */
1229 GLuint a
[4], b
[4], result
[4];
1230 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
1231 fetch_vector4ui(&inst
->SrcReg
[1], machine
, b
);
1232 result
[0] = a
[0] | b
[0];
1233 result
[1] = a
[1] | b
[1];
1234 result
[2] = a
[2] | b
[2];
1235 result
[3] = a
[3] | b
[3];
1236 store_vector4ui(inst
, machine
, result
);
1239 case OPCODE_PK2H
: /* pack two 16-bit floats in one 32-bit float */
1244 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1245 hx
= _mesa_float_to_half(a
[0]);
1246 hy
= _mesa_float_to_half(a
[1]);
1250 result
[3] = hx
| (hy
<< 16);
1251 store_vector4ui(inst
, machine
, result
);
1254 case OPCODE_PK2US
: /* pack two GLushorts into one 32-bit float */
1257 GLuint result
[4], usx
, usy
;
1258 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1259 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
1260 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
1261 usx
= IROUND(a
[0] * 65535.0F
);
1262 usy
= IROUND(a
[1] * 65535.0F
);
1266 result
[3] = usx
| (usy
<< 16);
1267 store_vector4ui(inst
, machine
, result
);
1270 case OPCODE_PK4B
: /* pack four GLbytes into one 32-bit float */
1273 GLuint result
[4], ubx
, uby
, ubz
, ubw
;
1274 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1275 a
[0] = CLAMP(a
[0], -128.0F
/ 127.0F
, 1.0F
);
1276 a
[1] = CLAMP(a
[1], -128.0F
/ 127.0F
, 1.0F
);
1277 a
[2] = CLAMP(a
[2], -128.0F
/ 127.0F
, 1.0F
);
1278 a
[3] = CLAMP(a
[3], -128.0F
/ 127.0F
, 1.0F
);
1279 ubx
= IROUND(127.0F
* a
[0] + 128.0F
);
1280 uby
= IROUND(127.0F
* a
[1] + 128.0F
);
1281 ubz
= IROUND(127.0F
* a
[2] + 128.0F
);
1282 ubw
= IROUND(127.0F
* a
[3] + 128.0F
);
1286 result
[3] = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
1287 store_vector4ui(inst
, machine
, result
);
1290 case OPCODE_PK4UB
: /* pack four GLubytes into one 32-bit float */
1293 GLuint result
[4], ubx
, uby
, ubz
, ubw
;
1294 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1295 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
1296 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
1297 a
[2] = CLAMP(a
[2], 0.0F
, 1.0F
);
1298 a
[3] = CLAMP(a
[3], 0.0F
, 1.0F
);
1299 ubx
= IROUND(255.0F
* a
[0]);
1300 uby
= IROUND(255.0F
* a
[1]);
1301 ubz
= IROUND(255.0F
* a
[2]);
1302 ubw
= IROUND(255.0F
* a
[3]);
1306 result
[3] = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
1307 store_vector4ui(inst
, machine
, result
);
1312 GLfloat a
[4], b
[4], result
[4];
1313 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1314 fetch_vector1(&inst
->SrcReg
[1], machine
, b
);
1315 result
[0] = result
[1] = result
[2] = result
[3]
1316 = (GLfloat
) _mesa_pow(a
[0], b
[0]);
1317 store_vector4(inst
, machine
, result
);
1322 GLfloat a
[4], result
[4];
1323 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1327 else if (IS_INF_OR_NAN(a
[0]))
1328 printf("RCP(inf)\n");
1330 result
[0] = result
[1] = result
[2] = result
[3] = 1.0F
/ a
[0];
1331 store_vector4(inst
, machine
, result
);
1334 case OPCODE_RET
: /* return from subroutine (conditional) */
1335 if (eval_condition(machine
, inst
)) {
1336 if (machine
->StackDepth
== 0) {
1337 return GL_TRUE
; /* Per GL_NV_vertex_program2 spec */
1339 /* subtract one because of pc++ in the for loop */
1340 pc
= machine
->CallStack
[--machine
->StackDepth
] - 1;
1343 case OPCODE_RFL
: /* reflection vector */
1345 GLfloat axis
[4], dir
[4], result
[4], tmpX
, tmpW
;
1346 fetch_vector4(&inst
->SrcReg
[0], machine
, axis
);
1347 fetch_vector4(&inst
->SrcReg
[1], machine
, dir
);
1348 tmpW
= DOT3(axis
, axis
);
1349 tmpX
= (2.0F
* DOT3(axis
, dir
)) / tmpW
;
1350 result
[0] = tmpX
* axis
[0] - dir
[0];
1351 result
[1] = tmpX
* axis
[1] - dir
[1];
1352 result
[2] = tmpX
* axis
[2] - dir
[2];
1353 /* result[3] is never written! XXX enforce in parser! */
1354 store_vector4(inst
, machine
, result
);
1357 case OPCODE_RSQ
: /* 1 / sqrt() */
1359 GLfloat a
[4], result
[4];
1360 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1362 result
[0] = result
[1] = result
[2] = result
[3] = INV_SQRTF(a
[0]);
1363 store_vector4(inst
, machine
, result
);
1365 printf("RSQ %g = 1/sqrt(|%g|)\n", result
[0], a
[0]);
1369 case OPCODE_SCS
: /* sine and cos */
1371 GLfloat a
[4], result
[4];
1372 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1373 result
[0] = (GLfloat
) _mesa_cos(a
[0]);
1374 result
[1] = (GLfloat
) _mesa_sin(a
[0]);
1375 result
[2] = 0.0; /* undefined! */
1376 result
[3] = 0.0; /* undefined! */
1377 store_vector4(inst
, machine
, result
);
1380 case OPCODE_SEQ
: /* set on equal */
1382 GLfloat a
[4], b
[4], result
[4];
1383 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1384 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1385 result
[0] = (a
[0] == b
[0]) ? 1.0F
: 0.0F
;
1386 result
[1] = (a
[1] == b
[1]) ? 1.0F
: 0.0F
;
1387 result
[2] = (a
[2] == b
[2]) ? 1.0F
: 0.0F
;
1388 result
[3] = (a
[3] == b
[3]) ? 1.0F
: 0.0F
;
1389 store_vector4(inst
, machine
, result
);
1391 printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
1392 result
[0], result
[1], result
[2], result
[3],
1393 a
[0], a
[1], a
[2], a
[3],
1394 b
[0], b
[1], b
[2], b
[3]);
1398 case OPCODE_SFL
: /* set false, operands ignored */
1400 static const GLfloat result
[4] = { 0.0F
, 0.0F
, 0.0F
, 0.0F
};
1401 store_vector4(inst
, machine
, result
);
1404 case OPCODE_SGE
: /* set on greater or equal */
1406 GLfloat a
[4], b
[4], result
[4];
1407 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1408 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1409 result
[0] = (a
[0] >= b
[0]) ? 1.0F
: 0.0F
;
1410 result
[1] = (a
[1] >= b
[1]) ? 1.0F
: 0.0F
;
1411 result
[2] = (a
[2] >= b
[2]) ? 1.0F
: 0.0F
;
1412 result
[3] = (a
[3] >= b
[3]) ? 1.0F
: 0.0F
;
1413 store_vector4(inst
, machine
, result
);
1415 printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
1416 result
[0], result
[1], result
[2], result
[3],
1417 a
[0], a
[1], a
[2], a
[3],
1418 b
[0], b
[1], b
[2], b
[3]);
1422 case OPCODE_SGT
: /* set on greater */
1424 GLfloat a
[4], b
[4], result
[4];
1425 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1426 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1427 result
[0] = (a
[0] > b
[0]) ? 1.0F
: 0.0F
;
1428 result
[1] = (a
[1] > b
[1]) ? 1.0F
: 0.0F
;
1429 result
[2] = (a
[2] > b
[2]) ? 1.0F
: 0.0F
;
1430 result
[3] = (a
[3] > b
[3]) ? 1.0F
: 0.0F
;
1431 store_vector4(inst
, machine
, result
);
1433 printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
1434 result
[0], result
[1], result
[2], result
[3],
1435 a
[0], a
[1], a
[2], a
[3],
1436 b
[0], b
[1], b
[2], b
[3]);
1442 GLfloat a
[4], result
[4];
1443 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1444 result
[0] = result
[1] = result
[2] = result
[3]
1445 = (GLfloat
) _mesa_sin(a
[0]);
1446 store_vector4(inst
, machine
, result
);
1449 case OPCODE_SLE
: /* set on less or equal */
1451 GLfloat a
[4], b
[4], result
[4];
1452 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1453 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1454 result
[0] = (a
[0] <= b
[0]) ? 1.0F
: 0.0F
;
1455 result
[1] = (a
[1] <= b
[1]) ? 1.0F
: 0.0F
;
1456 result
[2] = (a
[2] <= b
[2]) ? 1.0F
: 0.0F
;
1457 result
[3] = (a
[3] <= b
[3]) ? 1.0F
: 0.0F
;
1458 store_vector4(inst
, machine
, result
);
1460 printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
1461 result
[0], result
[1], result
[2], result
[3],
1462 a
[0], a
[1], a
[2], a
[3],
1463 b
[0], b
[1], b
[2], b
[3]);
1467 case OPCODE_SLT
: /* set on less */
1469 GLfloat a
[4], b
[4], result
[4];
1470 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1471 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1472 result
[0] = (a
[0] < b
[0]) ? 1.0F
: 0.0F
;
1473 result
[1] = (a
[1] < b
[1]) ? 1.0F
: 0.0F
;
1474 result
[2] = (a
[2] < b
[2]) ? 1.0F
: 0.0F
;
1475 result
[3] = (a
[3] < b
[3]) ? 1.0F
: 0.0F
;
1476 store_vector4(inst
, machine
, result
);
1478 printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
1479 result
[0], result
[1], result
[2], result
[3],
1480 a
[0], a
[1], a
[2], a
[3],
1481 b
[0], b
[1], b
[2], b
[3]);
1485 case OPCODE_SNE
: /* set on not equal */
1487 GLfloat a
[4], b
[4], result
[4];
1488 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1489 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1490 result
[0] = (a
[0] != b
[0]) ? 1.0F
: 0.0F
;
1491 result
[1] = (a
[1] != b
[1]) ? 1.0F
: 0.0F
;
1492 result
[2] = (a
[2] != b
[2]) ? 1.0F
: 0.0F
;
1493 result
[3] = (a
[3] != b
[3]) ? 1.0F
: 0.0F
;
1494 store_vector4(inst
, machine
, result
);
1496 printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
1497 result
[0], result
[1], result
[2], result
[3],
1498 a
[0], a
[1], a
[2], a
[3],
1499 b
[0], b
[1], b
[2], b
[3]);
1503 case OPCODE_SSG
: /* set sign (-1, 0 or +1) */
1505 GLfloat a
[4], result
[4];
1506 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1507 result
[0] = (GLfloat
) ((a
[0] > 0.0F
) - (a
[0] < 0.0F
));
1508 result
[1] = (GLfloat
) ((a
[1] > 0.0F
) - (a
[1] < 0.0F
));
1509 result
[2] = (GLfloat
) ((a
[2] > 0.0F
) - (a
[2] < 0.0F
));
1510 result
[3] = (GLfloat
) ((a
[3] > 0.0F
) - (a
[3] < 0.0F
));
1511 store_vector4(inst
, machine
, result
);
1514 case OPCODE_STR
: /* set true, operands ignored */
1516 static const GLfloat result
[4] = { 1.0F
, 1.0F
, 1.0F
, 1.0F
};
1517 store_vector4(inst
, machine
, result
);
1522 GLfloat a
[4], b
[4], result
[4];
1523 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1524 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1525 result
[0] = a
[0] - b
[0];
1526 result
[1] = a
[1] - b
[1];
1527 result
[2] = a
[2] - b
[2];
1528 result
[3] = a
[3] - b
[3];
1529 store_vector4(inst
, machine
, result
);
1531 printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
1532 result
[0], result
[1], result
[2], result
[3],
1533 a
[0], a
[1], a
[2], a
[3], b
[0], b
[1], b
[2], b
[3]);
1537 case OPCODE_SWZ
: /* extended swizzle */
1539 const struct prog_src_register
*source
= &inst
->SrcReg
[0];
1540 const GLfloat
*src
= get_src_register_pointer(source
, machine
);
1543 for (i
= 0; i
< 4; i
++) {
1544 const GLuint swz
= GET_SWZ(source
->Swizzle
, i
);
1545 if (swz
== SWIZZLE_ZERO
)
1547 else if (swz
== SWIZZLE_ONE
)
1552 result
[i
] = src
[swz
];
1554 if (source
->NegateBase
& (1 << i
))
1555 result
[i
] = -result
[i
];
1557 store_vector4(inst
, machine
, result
);
1560 case OPCODE_TEX
: /* Both ARB and NV frag prog */
1561 /* Simple texel lookup */
1563 GLfloat texcoord
[4], color
[4];
1564 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1566 fetch_texel(ctx
, machine
, inst
, texcoord
, 0.0, color
);
1569 printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
1570 color
[0], color
[1], color
[2], color
[3],
1572 texcoord
[0], texcoord
[1], texcoord
[2], texcoord
[3]);
1574 store_vector4(inst
, machine
, color
);
1577 case OPCODE_TXB
: /* GL_ARB_fragment_program only */
1578 /* Texel lookup with LOD bias */
1580 const struct gl_texture_unit
*texUnit
1581 = &ctx
->Texture
.Unit
[inst
->TexSrcUnit
];
1582 GLfloat texcoord
[4], color
[4], lodBias
;
1584 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1586 /* texcoord[3] is the bias to add to lambda */
1587 lodBias
= texUnit
->LodBias
+ texcoord
[3];
1588 if (texUnit
->_Current
) {
1589 lodBias
+= texUnit
->_Current
->LodBias
;
1592 fetch_texel(ctx
, machine
, inst
, texcoord
, lodBias
, color
);
1594 store_vector4(inst
, machine
, color
);
1597 case OPCODE_TXD
: /* GL_NV_fragment_program only */
1598 /* Texture lookup w/ partial derivatives for LOD */
1600 GLfloat texcoord
[4], dtdx
[4], dtdy
[4], color
[4];
1601 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1602 fetch_vector4(&inst
->SrcReg
[1], machine
, dtdx
);
1603 fetch_vector4(&inst
->SrcReg
[2], machine
, dtdy
);
1604 machine
->FetchTexelDeriv(ctx
, texcoord
, dtdx
, dtdy
,
1606 inst
->TexSrcUnit
, color
);
1607 store_vector4(inst
, machine
, color
);
1610 case OPCODE_TXP
: /* GL_ARB_fragment_program only */
1611 /* Texture lookup w/ projective divide */
1613 GLfloat texcoord
[4], color
[4];
1615 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1616 /* Not so sure about this test - if texcoord[3] is
1617 * zero, we'd probably be fine except for an ASSERT in
1618 * IROUND_POS() which gets triggered by the inf values created.
1620 if (texcoord
[3] != 0.0) {
1621 texcoord
[0] /= texcoord
[3];
1622 texcoord
[1] /= texcoord
[3];
1623 texcoord
[2] /= texcoord
[3];
1626 fetch_texel(ctx
, machine
, inst
, texcoord
, 0.0, color
);
1628 store_vector4(inst
, machine
, color
);
1631 case OPCODE_TXP_NV
: /* GL_NV_fragment_program only */
1632 /* Texture lookup w/ projective divide, as above, but do not
1633 * do the divide by w if sampling from a cube map.
1636 GLfloat texcoord
[4], color
[4];
1638 fetch_vector4(&inst
->SrcReg
[0], machine
, texcoord
);
1639 if (inst
->TexSrcTarget
!= TEXTURE_CUBE_INDEX
&&
1640 texcoord
[3] != 0.0) {
1641 texcoord
[0] /= texcoord
[3];
1642 texcoord
[1] /= texcoord
[3];
1643 texcoord
[2] /= texcoord
[3];
1646 fetch_texel(ctx
, machine
, inst
, texcoord
, 0.0, color
);
1648 store_vector4(inst
, machine
, color
);
1651 case OPCODE_TRUNC
: /* truncate toward zero */
1653 GLfloat a
[4], result
[4];
1654 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1655 result
[0] = (GLfloat
) (GLint
) a
[0];
1656 result
[1] = (GLfloat
) (GLint
) a
[1];
1657 result
[2] = (GLfloat
) (GLint
) a
[2];
1658 result
[3] = (GLfloat
) (GLint
) a
[3];
1659 store_vector4(inst
, machine
, result
);
1662 case OPCODE_UP2H
: /* unpack two 16-bit floats */
1664 GLfloat a
[4], result
[4];
1665 const GLuint
*rawBits
= (const GLuint
*) a
;
1667 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1668 hx
= rawBits
[0] & 0xffff;
1669 hy
= rawBits
[0] >> 16;
1670 result
[0] = result
[2] = _mesa_half_to_float(hx
);
1671 result
[1] = result
[3] = _mesa_half_to_float(hy
);
1672 store_vector4(inst
, machine
, result
);
1675 case OPCODE_UP2US
: /* unpack two GLushorts */
1677 GLfloat a
[4], result
[4];
1678 const GLuint
*rawBits
= (const GLuint
*) a
;
1680 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1681 usx
= rawBits
[0] & 0xffff;
1682 usy
= rawBits
[0] >> 16;
1683 result
[0] = result
[2] = usx
* (1.0f
/ 65535.0f
);
1684 result
[1] = result
[3] = usy
* (1.0f
/ 65535.0f
);
1685 store_vector4(inst
, machine
, result
);
1688 case OPCODE_UP4B
: /* unpack four GLbytes */
1690 GLfloat a
[4], result
[4];
1691 const GLuint
*rawBits
= (const GLuint
*) a
;
1692 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1693 result
[0] = (((rawBits
[0] >> 0) & 0xff) - 128) / 127.0F
;
1694 result
[1] = (((rawBits
[0] >> 8) & 0xff) - 128) / 127.0F
;
1695 result
[2] = (((rawBits
[0] >> 16) & 0xff) - 128) / 127.0F
;
1696 result
[3] = (((rawBits
[0] >> 24) & 0xff) - 128) / 127.0F
;
1697 store_vector4(inst
, machine
, result
);
1700 case OPCODE_UP4UB
: /* unpack four GLubytes */
1702 GLfloat a
[4], result
[4];
1703 const GLuint
*rawBits
= (const GLuint
*) a
;
1704 fetch_vector1(&inst
->SrcReg
[0], machine
, a
);
1705 result
[0] = ((rawBits
[0] >> 0) & 0xff) / 255.0F
;
1706 result
[1] = ((rawBits
[0] >> 8) & 0xff) / 255.0F
;
1707 result
[2] = ((rawBits
[0] >> 16) & 0xff) / 255.0F
;
1708 result
[3] = ((rawBits
[0] >> 24) & 0xff) / 255.0F
;
1709 store_vector4(inst
, machine
, result
);
1712 case OPCODE_XOR
: /* bitwise XOR */
1714 GLuint a
[4], b
[4], result
[4];
1715 fetch_vector4ui(&inst
->SrcReg
[0], machine
, a
);
1716 fetch_vector4ui(&inst
->SrcReg
[1], machine
, b
);
1717 result
[0] = a
[0] ^ b
[0];
1718 result
[1] = a
[1] ^ b
[1];
1719 result
[2] = a
[2] ^ b
[2];
1720 result
[3] = a
[3] ^ b
[3];
1721 store_vector4ui(inst
, machine
, result
);
1724 case OPCODE_XPD
: /* cross product */
1726 GLfloat a
[4], b
[4], result
[4];
1727 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1728 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1729 result
[0] = a
[1] * b
[2] - a
[2] * b
[1];
1730 result
[1] = a
[2] * b
[0] - a
[0] * b
[2];
1731 result
[2] = a
[0] * b
[1] - a
[1] * b
[0];
1733 store_vector4(inst
, machine
, result
);
1735 printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
1736 result
[0], result
[1], result
[2], result
[3],
1737 a
[0], a
[1], a
[2], b
[0], b
[1], b
[2]);
1741 case OPCODE_X2D
: /* 2-D matrix transform */
1743 GLfloat a
[4], b
[4], c
[4], result
[4];
1744 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1745 fetch_vector4(&inst
->SrcReg
[1], machine
, b
);
1746 fetch_vector4(&inst
->SrcReg
[2], machine
, c
);
1747 result
[0] = a
[0] + b
[0] * c
[0] + b
[1] * c
[1];
1748 result
[1] = a
[1] + b
[0] * c
[2] + b
[1] * c
[3];
1749 result
[2] = a
[2] + b
[0] * c
[0] + b
[1] * c
[1];
1750 result
[3] = a
[3] + b
[0] * c
[2] + b
[1] * c
[3];
1751 store_vector4(inst
, machine
, result
);
1756 if (inst
->SrcReg
[0].File
!= -1) {
1758 fetch_vector4(&inst
->SrcReg
[0], machine
, a
);
1759 _mesa_printf("%s%g, %g, %g, %g\n", (const char *) inst
->Data
,
1760 a
[0], a
[1], a
[2], a
[3]);
1763 _mesa_printf("%s\n", (const char *) inst
->Data
);
1770 _mesa_problem(ctx
, "Bad opcode %d in _mesa_execute_program",
1772 return GL_TRUE
; /* return value doesn't matter */
1776 if (numExec
> maxExec
) {
1777 _mesa_problem(ctx
, "Infinite loop detected in fragment program");
1783 #if FEATURE_MESA_program_debug
1784 CurrentMachine
= NULL
;