1 /* $Id: s_nvfragprog.c,v 1.13 2003/04/05 00:38:10 brianp Exp $ */
4 * Mesa 3-D graphics library
7 * Copyright (C) 1999-2003 Brian Paul All Rights Reserved.
9 * Permission is hereby granted, free of charge, to any person obtaining a
10 * copy of this software and associated documentation files (the "Software"),
11 * to deal in the Software without restriction, including without limitation
12 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
13 * and/or sell copies of the Software, and to permit persons to whom the
14 * Software is furnished to do so, subject to the following conditions:
16 * The above copyright notice and this permission notice shall be included
17 * in all copies or substantial portions of the Software.
19 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
20 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
21 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
22 * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
23 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
24 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
32 #include "nvfragprog.h"
35 #include "s_nvfragprog.h"
37 #include "s_texture.h"
44 fetch_texel( GLcontext
*ctx
, const GLfloat texcoord
[4], GLuint unit
,
47 const GLfloat
*lambda
= NULL
;
49 SWcontext
*swrast
= SWRAST_CONTEXT(ctx
);
51 swrast
->TextureSample
[unit
](ctx
, unit
, ctx
->Texture
.Unit
[unit
]._Current
,
52 1, (const GLfloat (*)[4]) texcoord
,
54 color
[0] = CHAN_TO_FLOAT(rgba
[0]);
55 color
[1] = CHAN_TO_FLOAT(rgba
[1]);
56 color
[2] = CHAN_TO_FLOAT(rgba
[2]);
57 color
[3] = CHAN_TO_FLOAT(rgba
[3]);
62 * Fetch a texel with the given partial derivatives to compute a level
63 * of detail in the mipmap.
66 fetch_texel_deriv( GLcontext
*ctx
, const GLfloat texcoord
[4],
67 const GLfloat texdx
[4], const GLfloat texdy
[4],
68 GLuint unit
, GLfloat color
[4] )
70 SWcontext
*swrast
= SWRAST_CONTEXT(ctx
);
71 const struct gl_texture_object
*texObj
= ctx
->Texture
.Unit
[unit
]._Current
;
72 const struct gl_texture_image
*texImg
= texObj
->Image
[texObj
->BaseLevel
];
73 const GLfloat texW
= (GLfloat
) texImg
->WidthScale
;
74 const GLfloat texH
= (GLfloat
) texImg
->HeightScale
;
77 GLfloat lambda
= _swrast_compute_lambda(texdx
[0], texdy
[0], /* ds/dx, ds/dy */
78 texdx
[1], texdy
[1], /* dt/dx, dt/dy */
79 texdx
[3], texdy
[2], /* dq/dx, dq/dy */
81 texcoord
[0], texcoord
[1], texcoord
[3],
84 swrast
->TextureSample
[unit
](ctx
, unit
, ctx
->Texture
.Unit
[unit
]._Current
,
85 1, (const GLfloat (*)[4]) texcoord
,
87 color
[0] = CHAN_TO_FLOAT(rgba
[0]);
88 color
[1] = CHAN_TO_FLOAT(rgba
[1]);
89 color
[2] = CHAN_TO_FLOAT(rgba
[2]);
90 color
[3] = CHAN_TO_FLOAT(rgba
[3]);
96 * Fetch a 4-element float vector from the given source register.
97 * Apply swizzling and negating as needed.
100 fetch_vector4( const struct fp_src_register
*source
,
101 const struct fp_machine
*machine
,
102 const struct fragment_program
*program
,
107 if (source
->IsParameter
) {
108 src
= program
->Parameters
[source
->Register
].Values
;
111 src
= machine
->Registers
[source
->Register
];
114 result
[0] = src
[source
->Swizzle
[0]];
115 result
[1] = src
[source
->Swizzle
[1]];
116 result
[2] = src
[source
->Swizzle
[2]];
117 result
[3] = src
[source
->Swizzle
[3]];
119 if (source
->NegateBase
) {
120 result
[0] = -result
[0];
121 result
[1] = -result
[1];
122 result
[2] = -result
[2];
123 result
[3] = -result
[3];
126 result
[0] = FABSF(result
[0]);
127 result
[1] = FABSF(result
[1]);
128 result
[2] = FABSF(result
[2]);
129 result
[3] = FABSF(result
[3]);
131 if (source
->NegateAbs
) {
132 result
[0] = -result
[0];
133 result
[1] = -result
[1];
134 result
[2] = -result
[2];
135 result
[3] = -result
[3];
141 * Fetch the derivative with respect to X for the given register.
142 * \return GL_TRUE if it was easily computed or GL_FALSE if we
143 * need to execute another instance of the program (ugh)!
146 fetch_vector4_deriv( const struct fp_src_register
*source
,
147 const struct sw_span
*span
,
148 char xOrY
, GLfloat result
[4] )
152 ASSERT(xOrY
== 'X' || xOrY
== 'Y');
154 switch (source
->Register
) {
155 case FRAG_ATTRIB_WPOS
:
169 case FRAG_ATTRIB_COL0
:
171 src
[0] = span
->drdx
* (1.0F
/ CHAN_MAXF
);
172 src
[1] = span
->dgdx
* (1.0F
/ CHAN_MAXF
);
173 src
[2] = span
->dbdx
* (1.0F
/ CHAN_MAXF
);
174 src
[3] = span
->dadx
* (1.0F
/ CHAN_MAXF
);
177 src
[0] = span
->drdy
* (1.0F
/ CHAN_MAXF
);
178 src
[1] = span
->dgdy
* (1.0F
/ CHAN_MAXF
);
179 src
[2] = span
->dbdy
* (1.0F
/ CHAN_MAXF
);
180 src
[3] = span
->dady
* (1.0F
/ CHAN_MAXF
);
183 case FRAG_ATTRIB_COL1
:
185 src
[0] = span
->dsrdx
* (1.0F
/ CHAN_MAXF
);
186 src
[1] = span
->dsgdx
* (1.0F
/ CHAN_MAXF
);
187 src
[2] = span
->dsbdx
* (1.0F
/ CHAN_MAXF
);
188 src
[3] = 0.0; /* XXX need this */
191 src
[0] = span
->dsrdy
* (1.0F
/ CHAN_MAXF
);
192 src
[1] = span
->dsgdy
* (1.0F
/ CHAN_MAXF
);
193 src
[2] = span
->dsbdy
* (1.0F
/ CHAN_MAXF
);
194 src
[3] = 0.0; /* XXX need this */
197 case FRAG_ATTRIB_FOGC
:
199 src
[0] = span
->dfogdx
;
205 src
[0] = span
->dfogdy
;
211 case FRAG_ATTRIB_TEX0
:
212 case FRAG_ATTRIB_TEX1
:
213 case FRAG_ATTRIB_TEX2
:
214 case FRAG_ATTRIB_TEX3
:
215 case FRAG_ATTRIB_TEX4
:
216 case FRAG_ATTRIB_TEX5
:
217 case FRAG_ATTRIB_TEX6
:
218 case FRAG_ATTRIB_TEX7
:
220 const GLuint u
= source
->Register
- FRAG_ATTRIB_TEX0
;
221 src
[0] = span
->texStepX
[u
][0] * (1.0F
/ CHAN_MAXF
);
222 src
[1] = span
->texStepX
[u
][1] * (1.0F
/ CHAN_MAXF
);
223 src
[2] = span
->texStepX
[u
][2] * (1.0F
/ CHAN_MAXF
);
224 src
[3] = span
->texStepX
[u
][3] * (1.0F
/ CHAN_MAXF
);
227 const GLuint u
= source
->Register
- FRAG_ATTRIB_TEX0
;
228 src
[0] = span
->texStepY
[u
][0] * (1.0F
/ CHAN_MAXF
);
229 src
[1] = span
->texStepY
[u
][1] * (1.0F
/ CHAN_MAXF
);
230 src
[2] = span
->texStepY
[u
][2] * (1.0F
/ CHAN_MAXF
);
231 src
[3] = span
->texStepY
[u
][3] * (1.0F
/ CHAN_MAXF
);
238 result
[0] = src
[source
->Swizzle
[0]];
239 result
[1] = src
[source
->Swizzle
[1]];
240 result
[2] = src
[source
->Swizzle
[2]];
241 result
[3] = src
[source
->Swizzle
[3]];
243 if (source
->NegateBase
) {
244 result
[0] = -result
[0];
245 result
[1] = -result
[1];
246 result
[2] = -result
[2];
247 result
[3] = -result
[3];
250 result
[0] = FABSF(result
[0]);
251 result
[1] = FABSF(result
[1]);
252 result
[2] = FABSF(result
[2]);
253 result
[3] = FABSF(result
[3]);
255 if (source
->NegateAbs
) {
256 result
[0] = -result
[0];
257 result
[1] = -result
[1];
258 result
[2] = -result
[2];
259 result
[3] = -result
[3];
266 * As above, but only return result[0] element.
269 fetch_vector1( const struct fp_src_register
*source
,
270 const struct fp_machine
*machine
,
271 const struct fragment_program
*program
,
276 if (source
->IsParameter
) {
277 src
= program
->Parameters
[source
->Register
].Values
;
280 src
= machine
->Registers
[source
->Register
];
283 result
[0] = src
[source
->Swizzle
[0]];
285 if (source
->NegateBase
) {
286 result
[0] = -result
[0];
289 result
[0] = FABSF(result
[0]);
291 if (source
->NegateAbs
) {
292 result
[0] = -result
[0];
298 * Test value against zero and return GT, LT, EQ or UN if NaN.
301 generate_cc( float value
)
304 return COND_UN
; /* NaN */
313 * Test if the ccMaskRule is satisfied by the given condition code.
314 * Used to mask destination writes according to the current condition codee.
316 static INLINE GLboolean
317 test_cc(GLuint condCode
, GLuint ccMaskRule
)
319 switch (ccMaskRule
) {
320 case COND_EQ
: return (condCode
== COND_EQ
);
321 case COND_NE
: return (condCode
!= COND_EQ
);
322 case COND_LT
: return (condCode
== COND_LT
);
323 case COND_GE
: return (condCode
== COND_GT
|| condCode
== COND_EQ
);
324 case COND_LE
: return (condCode
== COND_LT
|| condCode
== COND_EQ
);
325 case COND_GT
: return (condCode
== COND_GT
);
326 case COND_TR
: return GL_TRUE
;
327 case COND_FL
: return GL_FALSE
;
328 default: return GL_TRUE
;
334 * Store 4 floats into a register. Observe the instructions saturate and
335 * set-condition-code flags.
338 store_vector4( const struct fp_instruction
*inst
,
339 struct fp_machine
*machine
,
340 const GLfloat value
[4] )
342 const struct fp_dst_register
*dest
= &(inst
->DstReg
);
343 const GLboolean clamp
= inst
->Saturate
;
344 const GLboolean updateCC
= inst
->UpdateCondRegister
;
345 GLfloat
*dstReg
= machine
->Registers
[dest
->Register
];
346 GLfloat clampedValue
[4];
347 const GLboolean
*writeMask
= dest
->WriteMask
;
348 GLboolean condWriteMask
[4];
351 clampedValue
[0] = CLAMP(value
[0], 0.0F
, 1.0F
);
352 clampedValue
[1] = CLAMP(value
[1], 0.0F
, 1.0F
);
353 clampedValue
[2] = CLAMP(value
[2], 0.0F
, 1.0F
);
354 clampedValue
[3] = CLAMP(value
[3], 0.0F
, 1.0F
);
355 value
= clampedValue
;
358 if (dest
->CondMask
!= COND_TR
) {
359 condWriteMask
[0] = writeMask
[0]
360 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[0]], dest
->CondMask
);
361 condWriteMask
[1] = writeMask
[1]
362 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[1]], dest
->CondMask
);
363 condWriteMask
[2] = writeMask
[2]
364 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[2]], dest
->CondMask
);
365 condWriteMask
[3] = writeMask
[3]
366 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[3]], dest
->CondMask
);
367 writeMask
= condWriteMask
;
371 dstReg
[0] = value
[0];
373 machine
->CondCodes
[0] = generate_cc(value
[0]);
376 dstReg
[1] = value
[1];
378 machine
->CondCodes
[1] = generate_cc(value
[1]);
381 dstReg
[2] = value
[2];
383 machine
->CondCodes
[2] = generate_cc(value
[2]);
386 dstReg
[3] = value
[3];
388 machine
->CondCodes
[3] = generate_cc(value
[3]);
394 * Initialize a new machine state instance from an existing one, adding
395 * the partial derivatives onto the input registers.
396 * Used to implement DDX and DDY instructions in non-trivial cases.
399 init_machine_deriv( GLcontext
*ctx
,
400 const struct fp_machine
*machine
,
401 const struct fragment_program
*program
,
402 const struct sw_span
*span
, char xOrY
,
403 struct fp_machine
*dMachine
)
407 ASSERT(xOrY
== 'X' || xOrY
== 'Y');
409 /* copy existing machine */
410 _mesa_memcpy(dMachine
, machine
, sizeof(struct fp_machine
));
412 /* Clear temporary registers */
413 _mesa_bzero((GLfloat
*) (machine
->Registers
+ FP_TEMP_REG_START
) ,
414 MAX_NV_FRAGMENT_PROGRAM_TEMPS
* 4 * sizeof(GLfloat
));
416 /* Add derivatives */
417 if (program
->InputsRead
& (1 << FRAG_ATTRIB_WPOS
)) {
418 GLfloat
*wpos
= (GLfloat
*) machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_WPOS
];
422 wpos
[2] += span
->dzdx
;
423 wpos
[3] += span
->dwdx
;
428 wpos
[2] += span
->dzdy
;
429 wpos
[3] += span
->dwdy
;
432 if (program
->InputsRead
& (1 << FRAG_ATTRIB_COL0
)) {
433 GLfloat
*col0
= (GLfloat
*) machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_COL0
];
435 col0
[0] += span
->drdx
* (1.0F
/ CHAN_MAXF
);
436 col0
[1] += span
->dgdx
* (1.0F
/ CHAN_MAXF
);
437 col0
[2] += span
->dbdx
* (1.0F
/ CHAN_MAXF
);
438 col0
[3] += span
->dadx
* (1.0F
/ CHAN_MAXF
);
441 col0
[0] += span
->drdy
* (1.0F
/ CHAN_MAXF
);
442 col0
[1] += span
->dgdy
* (1.0F
/ CHAN_MAXF
);
443 col0
[2] += span
->dbdy
* (1.0F
/ CHAN_MAXF
);
444 col0
[3] += span
->dady
* (1.0F
/ CHAN_MAXF
);
447 if (program
->InputsRead
& (1 << FRAG_ATTRIB_COL1
)) {
448 GLfloat
*col1
= (GLfloat
*) machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_COL1
];
450 col1
[0] += span
->dsrdx
* (1.0F
/ CHAN_MAXF
);
451 col1
[1] += span
->dsgdx
* (1.0F
/ CHAN_MAXF
);
452 col1
[2] += span
->dsbdx
* (1.0F
/ CHAN_MAXF
);
453 col1
[3] += 0.0; /*XXX fix */
456 col1
[0] += span
->dsrdy
* (1.0F
/ CHAN_MAXF
);
457 col1
[1] += span
->dsgdy
* (1.0F
/ CHAN_MAXF
);
458 col1
[2] += span
->dsbdy
* (1.0F
/ CHAN_MAXF
);
459 col1
[3] += 0.0; /*XXX fix */
462 if (program
->InputsRead
& (1 << FRAG_ATTRIB_FOGC
)) {
463 GLfloat
*fogc
= (GLfloat
*) machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_FOGC
];
465 fogc
[0] += span
->dfogdx
;
468 fogc
[0] += span
->dfogdy
;
471 for (u
= 0; u
< ctx
->Const
.MaxTextureCoordUnits
; u
++) {
472 if (program
->InputsRead
& (1 << (FRAG_ATTRIB_TEX0
+ u
))) {
473 GLfloat
*tex
= (GLfloat
*) machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_TEX0
+u
];
475 tex
[0] += span
->texStepX
[u
][0];
476 tex
[1] += span
->texStepX
[u
][1];
477 tex
[2] += span
->texStepX
[u
][2];
478 tex
[3] += span
->texStepX
[u
][3];
481 tex
[0] += span
->texStepY
[u
][0];
482 tex
[1] += span
->texStepY
[u
][1];
483 tex
[2] += span
->texStepY
[u
][2];
484 tex
[3] += span
->texStepY
[u
][3];
492 * Execute the given vertex program.
493 * NOTE: we do everything in single-precision floating point; we don't
494 * currently observe the single/half/fixed-precision qualifiers.
495 * \param ctx - rendering context
496 * \param program - the fragment program to execute
497 * \param machine - machine state (register file)
498 * \param maxInst - max number of instructions to execute
499 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
502 execute_program( GLcontext
*ctx
,
503 const struct fragment_program
*program
, GLuint maxInst
,
504 struct fp_machine
*machine
, const struct sw_span
*span
)
508 for (pc
= 0; pc
< maxInst
; pc
++) {
509 const struct fp_instruction
*inst
= program
->Instructions
+ pc
;
510 switch (inst
->Opcode
) {
513 GLfloat a
[4], b
[4], result
[4];
514 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
515 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
516 result
[0] = a
[0] + b
[0];
517 result
[1] = a
[1] + b
[1];
518 result
[2] = a
[2] + b
[2];
519 result
[3] = a
[3] + b
[3];
520 store_vector4( inst
, machine
, result
);
525 GLfloat a
[4], result
[4];
526 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
527 result
[0] = result
[1] = result
[2] = result
[3] = _mesa_cos(a
[0]);
528 store_vector4( inst
, machine
, result
);
531 case FP_OPCODE_DDX
: /* Partial derivative with respect to X */
533 GLfloat a
[4], aNext
[4], result
[4];
534 struct fp_machine dMachine
;
535 if (!fetch_vector4_deriv(&inst
->SrcReg
[0], span
, 'X', result
)) {
536 /* This is tricky. Make a copy of the current machine state,
537 * increment the input registers by the dx or dy partial
538 * derivatives, then re-execute the program up to the
539 * preceeding instruction, then fetch the source register.
540 * Finally, find the difference in the register values for
541 * the original and derivative runs.
543 init_machine_deriv(ctx
, machine
, program
, span
,
545 execute_program(ctx
, program
, pc
, &dMachine
, span
);
546 fetch_vector4( &inst
->SrcReg
[0], &dMachine
, program
, aNext
);
547 result
[0] = aNext
[0] - a
[0];
548 result
[1] = aNext
[1] - a
[1];
549 result
[2] = aNext
[2] - a
[2];
550 result
[3] = aNext
[3] - a
[3];
552 store_vector4( inst
, machine
, result
);
555 case FP_OPCODE_DDY
: /* Partial derivative with respect to Y */
557 GLfloat a
[4], aNext
[4], result
[4];
558 struct fp_machine dMachine
;
559 if (!fetch_vector4_deriv(&inst
->SrcReg
[0], span
, 'Y', result
)) {
560 init_machine_deriv(ctx
, machine
, program
, span
,
562 execute_program(ctx
, program
, pc
, &dMachine
, span
);
563 fetch_vector4( &inst
->SrcReg
[0], &dMachine
, program
, aNext
);
564 result
[0] = aNext
[0] - a
[0];
565 result
[1] = aNext
[1] - a
[1];
566 result
[2] = aNext
[2] - a
[2];
567 result
[3] = aNext
[3] - a
[3];
569 store_vector4( inst
, machine
, result
);
574 GLfloat a
[4], b
[4], result
[4];
575 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
576 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
577 result
[0] = result
[1] = result
[2] = result
[3] =
578 a
[0] + b
[0] + a
[1] * b
[1] + a
[2] * b
[2];
579 store_vector4( inst
, machine
, result
);
584 GLfloat a
[4], b
[4], result
[4];
585 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
586 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
587 result
[0] = result
[1] = result
[2] = result
[3] =
588 a
[0] + b
[0] + a
[1] * b
[1] + a
[2] * b
[2] + a
[3] * b
[3];
589 store_vector4( inst
, machine
, result
);
592 case FP_OPCODE_DST
: /* Distance vector */
594 GLfloat a
[4], b
[4], result
[4];
595 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
596 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
598 result
[1] = a
[1] * b
[1];
601 store_vector4( inst
, machine
, result
);
604 case FP_OPCODE_EX2
: /* Exponential base 2 */
606 GLfloat a
[4], result
[4];
607 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
608 result
[0] = result
[1] = result
[2] = result
[3] =
609 (GLfloat
) _mesa_pow(2.0, a
[0]);
610 store_vector4( inst
, machine
, result
);
615 GLfloat a
[4], result
[4];
616 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
617 result
[0] = FLOORF(a
[0]);
618 result
[1] = FLOORF(a
[1]);
619 result
[2] = FLOORF(a
[2]);
620 result
[3] = FLOORF(a
[3]);
621 store_vector4( inst
, machine
, result
);
626 GLfloat a
[4], result
[4];
627 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
628 result
[0] = a
[0] - FLOORF(a
[0]);
629 result
[1] = a
[1] - FLOORF(a
[1]);
630 result
[2] = a
[2] - FLOORF(a
[2]);
631 result
[3] = a
[3] - FLOORF(a
[3]);
632 store_vector4( inst
, machine
, result
);
637 const GLuint
*swizzle
= inst
->DstReg
.CondSwizzle
;
638 const GLuint condMask
= inst
->DstReg
.CondMask
;
639 if (test_cc(machine
->CondCodes
[swizzle
[0]], condMask
) ||
640 test_cc(machine
->CondCodes
[swizzle
[1]], condMask
) ||
641 test_cc(machine
->CondCodes
[swizzle
[2]], condMask
) ||
642 test_cc(machine
->CondCodes
[swizzle
[3]], condMask
))
646 case FP_OPCODE_LG2
: /* log base 2 */
648 GLfloat a
[4], result
[4];
649 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
650 result
[0] = result
[1] = result
[2] = result
[3]
652 store_vector4( inst
, machine
, result
);
657 GLfloat a
[4], result
[4];
658 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
665 result
[2] = (a
[0] > 0.0) ? _mesa_pow(2.0, a
[3]) : 0.0F
;
667 store_vector4( inst
, machine
, result
);
672 GLfloat a
[4], b
[4], c
[4], result
[4];
673 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
674 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
675 fetch_vector4( &inst
->SrcReg
[2], machine
, program
, c
);
676 result
[0] = a
[0] * b
[0] + (1.0F
- a
[0]) * c
[0];
677 result
[1] = a
[1] * b
[1] + (1.0F
- a
[1]) * c
[1];
678 result
[2] = a
[2] * b
[2] + (1.0F
- a
[2]) * c
[2];
679 result
[3] = a
[3] * b
[3] + (1.0F
- a
[3]) * c
[3];
680 store_vector4( inst
, machine
, result
);
685 GLfloat a
[4], b
[4], c
[4], result
[4];
686 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
687 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
688 fetch_vector4( &inst
->SrcReg
[2], machine
, program
, c
);
689 result
[0] = a
[0] * b
[0] + c
[0];
690 result
[1] = a
[1] * b
[1] + c
[1];
691 result
[2] = a
[2] * b
[2] + c
[2];
692 result
[3] = a
[3] * b
[3] + c
[3];
693 store_vector4( inst
, machine
, result
);
698 GLfloat a
[4], b
[4], result
[4];
699 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
700 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
701 result
[0] = MAX2(a
[0], b
[0]);
702 result
[1] = MAX2(a
[1], b
[1]);
703 result
[2] = MAX2(a
[2], b
[2]);
704 result
[3] = MAX2(a
[3], b
[3]);
705 store_vector4( inst
, machine
, result
);
710 GLfloat a
[4], b
[4], result
[4];
711 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
712 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
713 result
[0] = MIN2(a
[0], b
[0]);
714 result
[1] = MIN2(a
[1], b
[1]);
715 result
[2] = MIN2(a
[2], b
[2]);
716 result
[3] = MIN2(a
[3], b
[3]);
717 store_vector4( inst
, machine
, result
);
723 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, result
);
724 store_vector4( inst
, machine
, result
);
729 GLfloat a
[4], b
[4], result
[4];
730 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
731 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
732 result
[0] = a
[0] * b
[0];
733 result
[1] = a
[1] * b
[1];
734 result
[2] = a
[2] * b
[2];
735 result
[3] = a
[3] * b
[3];
736 store_vector4( inst
, machine
, result
);
739 case FP_OPCODE_PK2H
: /* pack two 16-bit floats */
740 /* XXX this is probably wrong */
742 GLfloat a
[4], result
[4];
743 const GLuint
*rawBits
= (const GLuint
*) a
;
744 GLuint
*rawResult
= (GLuint
*) result
;
745 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
746 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
747 = rawBits
[0] | (rawBits
[1] << 16);
748 store_vector4( inst
, machine
, result
);
751 case FP_OPCODE_PK2US
: /* pack two GLushorts */
753 GLfloat a
[4], result
[4];
754 GLuint usx
, usy
, *rawResult
= (GLuint
*) result
;
755 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
756 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
757 a
[1] = CLAMP(a
[0], 0.0F
, 1.0F
);
758 usx
= IROUND(a
[0] * 65535.0F
);
759 usy
= IROUND(a
[1] * 65535.0F
);
760 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
762 store_vector4( inst
, machine
, result
);
765 case FP_OPCODE_PK4B
: /* pack four GLbytes */
767 GLfloat a
[4], result
[4];
768 GLuint ubx
, uby
, ubz
, ubw
, *rawResult
= (GLuint
*) result
;
769 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
770 a
[0] = CLAMP(a
[0], -128.0F
/ 127.0F
, 1.0F
);
771 a
[1] = CLAMP(a
[1], -128.0F
/ 127.0F
, 1.0F
);
772 a
[2] = CLAMP(a
[2], -128.0F
/ 127.0F
, 1.0F
);
773 a
[3] = CLAMP(a
[3], -128.0F
/ 127.0F
, 1.0F
);
774 ubx
= IROUND(127.0F
* a
[0] + 128.0F
);
775 uby
= IROUND(127.0F
* a
[1] + 128.0F
);
776 ubz
= IROUND(127.0F
* a
[2] + 128.0F
);
777 ubw
= IROUND(127.0F
* a
[3] + 128.0F
);
778 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
779 = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
780 store_vector4( inst
, machine
, result
);
783 case FP_OPCODE_PK4UB
: /* pack four GLubytes */
785 GLfloat a
[4], result
[4];
786 GLuint ubx
, uby
, ubz
, ubw
, *rawResult
= (GLuint
*) result
;
787 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
788 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
789 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
790 a
[2] = CLAMP(a
[2], 0.0F
, 1.0F
);
791 a
[3] = CLAMP(a
[3], 0.0F
, 1.0F
);
792 ubx
= IROUND(255.0F
* a
[0]);
793 uby
= IROUND(255.0F
* a
[1]);
794 ubz
= IROUND(255.0F
* a
[2]);
795 ubw
= IROUND(255.0F
* a
[3]);
796 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
797 = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
798 store_vector4( inst
, machine
, result
);
803 GLfloat a
[4], b
[4], result
[4];
804 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
805 fetch_vector1( &inst
->SrcReg
[1], machine
, program
, b
);
806 result
[0] = result
[1] = result
[2] = result
[3]
807 = _mesa_pow(a
[0], b
[0]);
808 store_vector4( inst
, machine
, result
);
813 GLfloat a
[4], result
[4];
814 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
815 result
[0] = result
[1] = result
[2] = result
[3]
817 store_vector4( inst
, machine
, result
);
822 GLfloat axis
[4], dir
[4], result
[4], tmp
[4];
823 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, axis
);
824 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, dir
);
825 tmp
[3] = axis
[0] * axis
[0]
828 tmp
[0] = (2.0F
* (axis
[0] * dir
[0] +
830 axis
[2] * dir
[2])) / tmp
[3];
831 result
[0] = tmp
[0] * axis
[0] - dir
[0];
832 result
[1] = tmp
[0] * axis
[1] - dir
[1];
833 result
[2] = tmp
[0] * axis
[2] - dir
[2];
834 /* result[3] is never written! XXX enforce in parser! */
835 store_vector4( inst
, machine
, result
);
838 case FP_OPCODE_RSQ
: /* 1 / sqrt() */
840 GLfloat a
[4], result
[4];
841 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
842 result
[0] = result
[1] = result
[2] = result
[3] = INV_SQRTF(a
[0]);
843 store_vector4( inst
, machine
, result
);
846 case FP_OPCODE_SEQ
: /* set on equal */
848 GLfloat a
[4], b
[4], result
[4];
849 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
850 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
851 result
[0] = (a
[0] == b
[0]) ? 1.0F
: 0.0F
;
852 result
[1] = (a
[1] == b
[1]) ? 1.0F
: 0.0F
;
853 result
[2] = (a
[2] == b
[2]) ? 1.0F
: 0.0F
;
854 result
[3] = (a
[3] == b
[3]) ? 1.0F
: 0.0F
;
855 store_vector4( inst
, machine
, result
);
858 case FP_OPCODE_SFL
: /* set false, operands ignored */
860 static const GLfloat result
[4] = { 0.0F
, 0.0F
, 0.0F
, 0.0F
};
861 store_vector4( inst
, machine
, result
);
864 case FP_OPCODE_SGE
: /* set on greater or equal */
866 GLfloat a
[4], b
[4], result
[4];
867 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
868 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
869 result
[0] = (a
[0] >= b
[0]) ? 1.0F
: 0.0F
;
870 result
[1] = (a
[1] >= b
[1]) ? 1.0F
: 0.0F
;
871 result
[2] = (a
[2] >= b
[2]) ? 1.0F
: 0.0F
;
872 result
[3] = (a
[3] >= b
[3]) ? 1.0F
: 0.0F
;
873 store_vector4( inst
, machine
, result
);
876 case FP_OPCODE_SGT
: /* set on greater */
878 GLfloat a
[4], b
[4], result
[4];
879 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
880 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
881 result
[0] = (a
[0] > b
[0]) ? 1.0F
: 0.0F
;
882 result
[1] = (a
[1] > b
[1]) ? 1.0F
: 0.0F
;
883 result
[2] = (a
[2] > b
[2]) ? 1.0F
: 0.0F
;
884 result
[3] = (a
[3] > b
[3]) ? 1.0F
: 0.0F
;
885 store_vector4( inst
, machine
, result
);
890 GLfloat a
[4], result
[4];
891 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
892 result
[0] = result
[1] = result
[2] = result
[3] = _mesa_sin(a
[0]);
893 store_vector4( inst
, machine
, result
);
896 case FP_OPCODE_SLE
: /* set on less or equal */
898 GLfloat a
[4], b
[4], result
[4];
899 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
900 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
901 result
[0] = (a
[0] <= b
[0]) ? 1.0F
: 0.0F
;
902 result
[1] = (a
[1] <= b
[1]) ? 1.0F
: 0.0F
;
903 result
[2] = (a
[2] <= b
[2]) ? 1.0F
: 0.0F
;
904 result
[3] = (a
[3] <= b
[3]) ? 1.0F
: 0.0F
;
905 store_vector4( inst
, machine
, result
);
908 case FP_OPCODE_SLT
: /* set on less */
910 GLfloat a
[4], b
[4], result
[4];
911 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
912 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
913 result
[0] = (a
[0] < b
[0]) ? 1.0F
: 0.0F
;
914 result
[1] = (a
[1] < b
[1]) ? 1.0F
: 0.0F
;
915 result
[2] = (a
[2] < b
[2]) ? 1.0F
: 0.0F
;
916 result
[3] = (a
[3] < b
[3]) ? 1.0F
: 0.0F
;
917 store_vector4( inst
, machine
, result
);
920 case FP_OPCODE_SNE
: /* set on not equal */
922 GLfloat a
[4], b
[4], result
[4];
923 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
924 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
925 result
[0] = (a
[0] != b
[0]) ? 1.0F
: 0.0F
;
926 result
[1] = (a
[1] != b
[1]) ? 1.0F
: 0.0F
;
927 result
[2] = (a
[2] != b
[2]) ? 1.0F
: 0.0F
;
928 result
[3] = (a
[3] != b
[3]) ? 1.0F
: 0.0F
;
929 store_vector4( inst
, machine
, result
);
932 case FP_OPCODE_STR
: /* set true, operands ignored */
934 static const GLfloat result
[4] = { 1.0F
, 1.0F
, 1.0F
, 1.0F
};
935 store_vector4( inst
, machine
, result
);
940 GLfloat a
[4], b
[4], result
[4];
941 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
942 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
943 result
[0] = a
[0] - b
[0];
944 result
[1] = a
[1] - b
[1];
945 result
[2] = a
[2] - b
[2];
946 result
[3] = a
[3] - b
[3];
947 store_vector4( inst
, machine
, result
);
953 GLfloat texcoord
[4], color
[4];
954 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, texcoord
);
955 /* XXX: Undo perspective divide from interpolate_texcoords() */
956 fetch_texel( ctx
, texcoord
, inst
->TexSrcUnit
, color
);
957 store_vector4( inst
, machine
, color
);
961 /* Texture lookup w/ partial derivatives for LOD */
963 GLfloat texcoord
[4], dtdx
[4], dtdy
[4], color
[4];
964 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, texcoord
);
965 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, dtdx
);
966 fetch_vector4( &inst
->SrcReg
[2], machine
, program
, dtdy
);
967 fetch_texel_deriv( ctx
, texcoord
, dtdx
, dtdy
, inst
->TexSrcUnit
,
969 store_vector4( inst
, machine
, color
);
973 /* Texture lookup w/ perspective divide */
975 GLfloat texcoord
[4], color
[4];
976 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, texcoord
);
977 /* Already did perspective divide in interpolate_texcoords() */
978 fetch_texel( ctx
, texcoord
, inst
->TexSrcUnit
, color
);
979 store_vector4( inst
, machine
, color
);
982 case FP_OPCODE_UP2H
: /* unpack two 16-bit floats */
983 /* XXX this is probably wrong */
985 GLfloat a
[4], result
[4];
986 const GLuint
*rawBits
= (const GLuint
*) a
;
987 GLuint
*rawResult
= (GLuint
*) result
;
988 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
989 rawResult
[0] = rawBits
[0] & 0xffff;
990 rawResult
[1] = (rawBits
[0] >> 16) & 0xffff;
991 rawResult
[2] = rawBits
[0] & 0xffff;
992 rawResult
[3] = (rawBits
[0] >> 16) & 0xffff;
993 store_vector4( inst
, machine
, result
);
996 case FP_OPCODE_UP2US
: /* unpack two GLushorts */
998 GLfloat a
[4], result
[4];
999 const GLuint
*rawBits
= (const GLuint
*) a
;
1000 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
1001 result
[0] = (GLfloat
) ((rawBits
[0] >> 0) & 0xffff) / 65535.0F
;
1002 result
[1] = (GLfloat
) ((rawBits
[0] >> 16) & 0xffff) / 65535.0F
;
1003 result
[2] = result
[0];
1004 result
[3] = result
[1];
1005 store_vector4( inst
, machine
, result
);
1008 case FP_OPCODE_UP4B
: /* unpack four GLbytes */
1010 GLfloat a
[4], result
[4];
1011 const GLuint
*rawBits
= (const GLuint
*) a
;
1012 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
1013 result
[0] = (((rawBits
[0] >> 0) & 0xff) - 128) / 127.0F
;
1014 result
[0] = (((rawBits
[0] >> 8) & 0xff) - 128) / 127.0F
;
1015 result
[0] = (((rawBits
[0] >> 16) & 0xff) - 128) / 127.0F
;
1016 result
[0] = (((rawBits
[0] >> 24) & 0xff) - 128) / 127.0F
;
1017 store_vector4( inst
, machine
, result
);
1020 case FP_OPCODE_UP4UB
: /* unpack four GLubytes */
1022 GLfloat a
[4], result
[4];
1023 const GLuint
*rawBits
= (const GLuint
*) a
;
1024 fetch_vector1( &inst
->SrcReg
[0], machine
, program
, a
);
1025 result
[0] = ((rawBits
[0] >> 0) & 0xff) / 255.0F
;
1026 result
[0] = ((rawBits
[0] >> 8) & 0xff) / 255.0F
;
1027 result
[0] = ((rawBits
[0] >> 16) & 0xff) / 255.0F
;
1028 result
[0] = ((rawBits
[0] >> 24) & 0xff) / 255.0F
;
1029 store_vector4( inst
, machine
, result
);
1032 case FP_OPCODE_X2D
: /* 2-D matrix transform */
1034 GLfloat a
[4], b
[4], c
[4], result
[4];
1035 fetch_vector4( &inst
->SrcReg
[0], machine
, program
, a
);
1036 fetch_vector4( &inst
->SrcReg
[1], machine
, program
, b
);
1037 fetch_vector4( &inst
->SrcReg
[2], machine
, program
, c
);
1038 result
[0] = a
[0] + b
[0] * c
[0] + b
[1] * c
[1];
1039 result
[1] = a
[1] + b
[0] * c
[2] + b
[1] * c
[3];
1040 result
[2] = a
[2] + b
[0] * c
[0] + b
[1] * c
[1];
1041 result
[3] = a
[3] + b
[0] * c
[2] + b
[1] * c
[3];
1042 store_vector4( inst
, machine
, result
);
1048 _mesa_problem(ctx
, "Bad opcode %d in _mesa_exec_fragment_program",
1050 return GL_TRUE
; /* return value doesn't matter */
1058 init_machine( GLcontext
*ctx
, struct fp_machine
*machine
,
1059 const struct fragment_program
*program
,
1060 const struct sw_span
*span
, GLuint col
)
1064 /* Clear temporary registers */
1065 _mesa_bzero(machine
->Registers
+ FP_TEMP_REG_START
,
1066 MAX_NV_FRAGMENT_PROGRAM_TEMPS
* 4 * sizeof(GLfloat
));
1068 /* Load program local parameters */
1069 for (j
= 0; j
< MAX_NV_FRAGMENT_PROGRAM_PARAMS
; j
++) {
1070 COPY_4V(machine
->Registers
[FP_PROG_REG_START
+ j
],
1071 program
->LocalParams
[j
]);
1074 /* Load input registers */
1075 if (program
->InputsRead
& (1 << FRAG_ATTRIB_WPOS
)) {
1076 GLfloat
*wpos
= machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_WPOS
];
1077 wpos
[0] = span
->x
+ col
;
1079 wpos
[2] = (GLfloat
) span
->array
->z
[col
] / ctx
->DepthMaxF
;
1080 wpos
[3] = span
->w
+ col
* span
->dwdx
;
1082 if (program
->InputsRead
& (1 << FRAG_ATTRIB_COL0
)) {
1083 GLfloat
*col0
= machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_COL0
];
1084 col0
[0] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][RCOMP
]);
1085 col0
[1] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][GCOMP
]);
1086 col0
[2] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][BCOMP
]);
1087 col0
[3] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][ACOMP
]);
1089 if (program
->InputsRead
& (1 << FRAG_ATTRIB_COL1
)) {
1090 GLfloat
*col1
= machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_COL1
];
1091 col1
[0] = CHAN_TO_FLOAT(span
->array
->spec
[col
][RCOMP
]);
1092 col1
[1] = CHAN_TO_FLOAT(span
->array
->spec
[col
][GCOMP
]);
1093 col1
[2] = CHAN_TO_FLOAT(span
->array
->spec
[col
][BCOMP
]);
1094 col1
[3] = CHAN_TO_FLOAT(span
->array
->spec
[col
][ACOMP
]);
1096 if (program
->InputsRead
& (1 << FRAG_ATTRIB_FOGC
)) {
1097 GLfloat
*fogc
= machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_FOGC
];
1098 fogc
[0] = span
->array
->fog
[col
];
1103 for (u
= 0; u
< ctx
->Const
.MaxTextureCoordUnits
; u
++) {
1104 if (program
->InputsRead
& (1 << (FRAG_ATTRIB_TEX0
+ u
))) {
1105 GLfloat
*tex
= machine
->Registers
[FP_INPUT_REG_START
+FRAG_ATTRIB_TEX0
+u
];
1106 if (ctx
->Texture
.Unit
[u
]._ReallyEnabled
) {
1107 COPY_4V(tex
, span
->array
->texcoords
[u
][col
]);
1110 COPY_4V(tex
, ctx
->Current
.Attrib
[VERT_ATTRIB_TEX0
+ u
]);
1118 _swrast_exec_nv_fragment_program( GLcontext
*ctx
, struct sw_span
*span
)
1120 const struct fragment_program
*program
= ctx
->FragmentProgram
.Current
;
1123 for (i
= 0; i
< span
->end
; i
++) {
1124 if (span
->array
->mask
[i
]) {
1125 init_machine(ctx
, &ctx
->FragmentProgram
.Machine
,
1126 ctx
->FragmentProgram
.Current
, span
, i
);
1128 if (!execute_program(ctx
, program
, ~0,
1129 &ctx
->FragmentProgram
.Machine
, span
))
1130 span
->array
->mask
[i
] = GL_FALSE
; /* killed fragment */
1132 /* Store output registers */
1134 const GLfloat
*colOut
1135 = ctx
->FragmentProgram
.Machine
.Registers
[FP_OUTPUT_REG_START
];
1136 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][RCOMP
], colOut
[0]);
1137 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][GCOMP
], colOut
[1]);
1138 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][BCOMP
], colOut
[2]);
1139 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][ACOMP
], colOut
[3]);
1142 if (ctx
->FragmentProgram
.Current
->OutputsWritten
& 2)
1143 span
->array
->z
[i
] = IROUND(ctx
->FragmentProgram
.Machine
.Registers
[FP_OUTPUT_REG_START
+ 2][0] * ctx
->DepthMaxF
);