2 * Mesa 3-D graphics library
5 * Copyright (C) 1999-2004 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 * Regarding GL_NV_fragment_program:
28 * Portions of this software may use or implement intellectual
29 * property owned and licensed by NVIDIA Corporation. NVIDIA disclaims
30 * any and all warranties with respect to such intellectual property,
31 * including any use thereof or modifications thereto.
37 #include "nvfragprog.h"
41 #include "s_nvfragprog.h"
43 #include "s_texture.h"
46 /* if 1, print some debugging info */
53 fetch_texel( GLcontext
*ctx
, const GLfloat texcoord
[4], GLfloat lambda
,
54 GLuint unit
, GLfloat color
[4] )
57 SWcontext
*swrast
= SWRAST_CONTEXT(ctx
);
59 /* XXX use a float-valued TextureSample routine here!!! */
60 swrast
->TextureSample
[unit
](ctx
, unit
, ctx
->Texture
.Unit
[unit
]._Current
,
61 1, (const GLfloat (*)[4]) texcoord
,
63 color
[0] = CHAN_TO_FLOAT(rgba
[0]);
64 color
[1] = CHAN_TO_FLOAT(rgba
[1]);
65 color
[2] = CHAN_TO_FLOAT(rgba
[2]);
66 color
[3] = CHAN_TO_FLOAT(rgba
[3]);
71 * Fetch a texel with the given partial derivatives to compute a level
72 * of detail in the mipmap.
75 fetch_texel_deriv( GLcontext
*ctx
, const GLfloat texcoord
[4],
76 const GLfloat texdx
[4], const GLfloat texdy
[4],
77 GLuint unit
, GLfloat color
[4] )
79 SWcontext
*swrast
= SWRAST_CONTEXT(ctx
);
80 const struct gl_texture_object
*texObj
= ctx
->Texture
.Unit
[unit
]._Current
;
81 const struct gl_texture_image
*texImg
= texObj
->Image
[0][texObj
->BaseLevel
];
82 const GLfloat texW
= (GLfloat
) texImg
->WidthScale
;
83 const GLfloat texH
= (GLfloat
) texImg
->HeightScale
;
86 GLfloat lambda
= _swrast_compute_lambda(texdx
[0], texdy
[0], /* ds/dx, ds/dy */
87 texdx
[1], texdy
[1], /* dt/dx, dt/dy */
88 texdx
[3], texdy
[2], /* dq/dx, dq/dy */
90 texcoord
[0], texcoord
[1], texcoord
[3],
93 swrast
->TextureSample
[unit
](ctx
, unit
, ctx
->Texture
.Unit
[unit
]._Current
,
94 1, (const GLfloat (*)[4]) texcoord
,
96 color
[0] = CHAN_TO_FLOAT(rgba
[0]);
97 color
[1] = CHAN_TO_FLOAT(rgba
[1]);
98 color
[2] = CHAN_TO_FLOAT(rgba
[2]);
99 color
[3] = CHAN_TO_FLOAT(rgba
[3]);
104 * Return a pointer to the 4-element float vector specified by the given
107 static INLINE
const GLfloat
*
108 get_register_pointer( GLcontext
*ctx
,
109 const struct fp_src_register
*source
,
110 const struct fp_machine
*machine
,
111 const struct fragment_program
*program
)
114 switch (source
->File
) {
115 case PROGRAM_TEMPORARY
:
116 ASSERT(source
->Index
< MAX_NV_FRAGMENT_PROGRAM_TEMPS
);
117 src
= machine
->Temporaries
[source
->Index
];
120 ASSERT(source
->Index
< MAX_NV_FRAGMENT_PROGRAM_INPUTS
);
121 src
= machine
->Inputs
[source
->Index
];
123 case PROGRAM_LOCAL_PARAM
:
124 ASSERT(source
->Index
< MAX_PROGRAM_LOCAL_PARAMS
);
125 src
= program
->Base
.LocalParams
[source
->Index
];
127 case PROGRAM_ENV_PARAM
:
128 ASSERT(source
->Index
< MAX_NV_FRAGMENT_PROGRAM_PARAMS
);
129 src
= ctx
->FragmentProgram
.Parameters
[source
->Index
];
132 case PROGRAM_STATE_VAR
:
135 case PROGRAM_NAMED_PARAM
:
136 ASSERT(source
->Index
< (GLint
) program
->Parameters
->NumParameters
);
137 src
= program
->Parameters
->Parameters
[source
->Index
].Values
;
140 _mesa_problem(ctx
, "Invalid input register file in fetch_vector4");
148 * Fetch a 4-element float vector from the given source register.
149 * Apply swizzling and negating as needed.
152 fetch_vector4( GLcontext
*ctx
,
153 const struct fp_src_register
*source
,
154 const struct fp_machine
*machine
,
155 const struct fragment_program
*program
,
158 const GLfloat
*src
= get_register_pointer(ctx
, source
, machine
, program
);
161 result
[0] = src
[source
->Swizzle
[0]];
162 result
[1] = src
[source
->Swizzle
[1]];
163 result
[2] = src
[source
->Swizzle
[2]];
164 result
[3] = src
[source
->Swizzle
[3]];
166 if (source
->NegateBase
) {
167 result
[0] = -result
[0];
168 result
[1] = -result
[1];
169 result
[2] = -result
[2];
170 result
[3] = -result
[3];
173 result
[0] = FABSF(result
[0]);
174 result
[1] = FABSF(result
[1]);
175 result
[2] = FABSF(result
[2]);
176 result
[3] = FABSF(result
[3]);
178 if (source
->NegateAbs
) {
179 result
[0] = -result
[0];
180 result
[1] = -result
[1];
181 result
[2] = -result
[2];
182 result
[3] = -result
[3];
188 * Fetch the derivative with respect to X for the given register.
189 * \return GL_TRUE if it was easily computed or GL_FALSE if we
190 * need to execute another instance of the program (ugh)!
193 fetch_vector4_deriv( GLcontext
*ctx
,
194 const struct fp_src_register
*source
,
195 const struct sw_span
*span
,
196 char xOrY
, GLint column
, GLfloat result
[4] )
200 ASSERT(xOrY
== 'X' || xOrY
== 'Y');
202 switch (source
->Index
) {
203 case FRAG_ATTRIB_WPOS
:
207 src
[2] = span
->dzdx
/ ctx
->DepthMaxF
;
213 src
[2] = span
->dzdy
/ ctx
->DepthMaxF
;
217 case FRAG_ATTRIB_COL0
:
219 src
[0] = span
->drdx
* (1.0F
/ CHAN_MAXF
);
220 src
[1] = span
->dgdx
* (1.0F
/ CHAN_MAXF
);
221 src
[2] = span
->dbdx
* (1.0F
/ CHAN_MAXF
);
222 src
[3] = span
->dadx
* (1.0F
/ CHAN_MAXF
);
225 src
[0] = span
->drdy
* (1.0F
/ CHAN_MAXF
);
226 src
[1] = span
->dgdy
* (1.0F
/ CHAN_MAXF
);
227 src
[2] = span
->dbdy
* (1.0F
/ CHAN_MAXF
);
228 src
[3] = span
->dady
* (1.0F
/ CHAN_MAXF
);
231 case FRAG_ATTRIB_COL1
:
233 src
[0] = span
->dsrdx
* (1.0F
/ CHAN_MAXF
);
234 src
[1] = span
->dsgdx
* (1.0F
/ CHAN_MAXF
);
235 src
[2] = span
->dsbdx
* (1.0F
/ CHAN_MAXF
);
236 src
[3] = 0.0; /* XXX need this */
239 src
[0] = span
->dsrdy
* (1.0F
/ CHAN_MAXF
);
240 src
[1] = span
->dsgdy
* (1.0F
/ CHAN_MAXF
);
241 src
[2] = span
->dsbdy
* (1.0F
/ CHAN_MAXF
);
242 src
[3] = 0.0; /* XXX need this */
245 case FRAG_ATTRIB_FOGC
:
247 src
[0] = span
->dfogdx
;
253 src
[0] = span
->dfogdy
;
259 case FRAG_ATTRIB_TEX0
:
260 case FRAG_ATTRIB_TEX1
:
261 case FRAG_ATTRIB_TEX2
:
262 case FRAG_ATTRIB_TEX3
:
263 case FRAG_ATTRIB_TEX4
:
264 case FRAG_ATTRIB_TEX5
:
265 case FRAG_ATTRIB_TEX6
:
266 case FRAG_ATTRIB_TEX7
:
268 const GLuint u
= source
->Index
- FRAG_ATTRIB_TEX0
;
269 /* this is a little tricky - I think I've got it right */
270 const GLfloat invQ
= 1.0f
/ (span
->tex
[u
][3]
271 + span
->texStepX
[u
][3] * column
);
272 src
[0] = span
->texStepX
[u
][0] * invQ
;
273 src
[1] = span
->texStepX
[u
][1] * invQ
;
274 src
[2] = span
->texStepX
[u
][2] * invQ
;
275 src
[3] = span
->texStepX
[u
][3] * invQ
;
278 const GLuint u
= source
->Index
- FRAG_ATTRIB_TEX0
;
279 /* Tricky, as above, but in Y direction */
280 const GLfloat invQ
= 1.0f
/ (span
->tex
[u
][3] + span
->texStepY
[u
][3]);
281 src
[0] = span
->texStepY
[u
][0] * invQ
;
282 src
[1] = span
->texStepY
[u
][1] * invQ
;
283 src
[2] = span
->texStepY
[u
][2] * invQ
;
284 src
[3] = span
->texStepY
[u
][3] * invQ
;
291 result
[0] = src
[source
->Swizzle
[0]];
292 result
[1] = src
[source
->Swizzle
[1]];
293 result
[2] = src
[source
->Swizzle
[2]];
294 result
[3] = src
[source
->Swizzle
[3]];
296 if (source
->NegateBase
) {
297 result
[0] = -result
[0];
298 result
[1] = -result
[1];
299 result
[2] = -result
[2];
300 result
[3] = -result
[3];
303 result
[0] = FABSF(result
[0]);
304 result
[1] = FABSF(result
[1]);
305 result
[2] = FABSF(result
[2]);
306 result
[3] = FABSF(result
[3]);
308 if (source
->NegateAbs
) {
309 result
[0] = -result
[0];
310 result
[1] = -result
[1];
311 result
[2] = -result
[2];
312 result
[3] = -result
[3];
319 * As above, but only return result[0] element.
322 fetch_vector1( GLcontext
*ctx
,
323 const struct fp_src_register
*source
,
324 const struct fp_machine
*machine
,
325 const struct fragment_program
*program
,
328 const GLfloat
*src
= get_register_pointer(ctx
, source
, machine
, program
);
331 result
[0] = src
[source
->Swizzle
[0]];
333 if (source
->NegateBase
) {
334 result
[0] = -result
[0];
337 result
[0] = FABSF(result
[0]);
339 if (source
->NegateAbs
) {
340 result
[0] = -result
[0];
346 * Test value against zero and return GT, LT, EQ or UN if NaN.
349 generate_cc( float value
)
352 return COND_UN
; /* NaN */
361 * Test if the ccMaskRule is satisfied by the given condition code.
362 * Used to mask destination writes according to the current condition codee.
364 static INLINE GLboolean
365 test_cc(GLuint condCode
, GLuint ccMaskRule
)
367 switch (ccMaskRule
) {
368 case COND_EQ
: return (condCode
== COND_EQ
);
369 case COND_NE
: return (condCode
!= COND_EQ
);
370 case COND_LT
: return (condCode
== COND_LT
);
371 case COND_GE
: return (condCode
== COND_GT
|| condCode
== COND_EQ
);
372 case COND_LE
: return (condCode
== COND_LT
|| condCode
== COND_EQ
);
373 case COND_GT
: return (condCode
== COND_GT
);
374 case COND_TR
: return GL_TRUE
;
375 case COND_FL
: return GL_FALSE
;
376 default: return GL_TRUE
;
382 * Store 4 floats into a register. Observe the instructions saturate and
383 * set-condition-code flags.
386 store_vector4( const struct fp_instruction
*inst
,
387 struct fp_machine
*machine
,
388 const GLfloat value
[4] )
390 const struct fp_dst_register
*dest
= &(inst
->DstReg
);
391 const GLboolean clamp
= inst
->Saturate
;
392 const GLboolean updateCC
= inst
->UpdateCondRegister
;
394 GLfloat clampedValue
[4];
395 const GLboolean
*writeMask
= dest
->WriteMask
;
396 GLboolean condWriteMask
[4];
398 switch (dest
->File
) {
400 dstReg
= machine
->Outputs
[dest
->Index
];
402 case PROGRAM_TEMPORARY
:
403 dstReg
= machine
->Temporaries
[dest
->Index
];
406 _mesa_problem(NULL
, "bad register file in store_vector4(fp)");
411 if (value
[0] > 1.0e10
||
412 IS_INF_OR_NAN(value
[0]) ||
413 IS_INF_OR_NAN(value
[1]) ||
414 IS_INF_OR_NAN(value
[2]) ||
415 IS_INF_OR_NAN(value
[3]) )
416 printf("store %g %g %g %g\n", value
[0], value
[1], value
[2], value
[3]);
420 clampedValue
[0] = CLAMP(value
[0], 0.0F
, 1.0F
);
421 clampedValue
[1] = CLAMP(value
[1], 0.0F
, 1.0F
);
422 clampedValue
[2] = CLAMP(value
[2], 0.0F
, 1.0F
);
423 clampedValue
[3] = CLAMP(value
[3], 0.0F
, 1.0F
);
424 value
= clampedValue
;
427 if (dest
->CondMask
!= COND_TR
) {
428 condWriteMask
[0] = writeMask
[0]
429 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[0]], dest
->CondMask
);
430 condWriteMask
[1] = writeMask
[1]
431 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[1]], dest
->CondMask
);
432 condWriteMask
[2] = writeMask
[2]
433 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[2]], dest
->CondMask
);
434 condWriteMask
[3] = writeMask
[3]
435 && test_cc(machine
->CondCodes
[dest
->CondSwizzle
[3]], dest
->CondMask
);
436 writeMask
= condWriteMask
;
440 dstReg
[0] = value
[0];
442 machine
->CondCodes
[0] = generate_cc(value
[0]);
445 dstReg
[1] = value
[1];
447 machine
->CondCodes
[1] = generate_cc(value
[1]);
450 dstReg
[2] = value
[2];
452 machine
->CondCodes
[2] = generate_cc(value
[2]);
455 dstReg
[3] = value
[3];
457 machine
->CondCodes
[3] = generate_cc(value
[3]);
463 * Initialize a new machine state instance from an existing one, adding
464 * the partial derivatives onto the input registers.
465 * Used to implement DDX and DDY instructions in non-trivial cases.
468 init_machine_deriv( GLcontext
*ctx
,
469 const struct fp_machine
*machine
,
470 const struct fragment_program
*program
,
471 const struct sw_span
*span
, char xOrY
,
472 struct fp_machine
*dMachine
)
476 ASSERT(xOrY
== 'X' || xOrY
== 'Y');
478 /* copy existing machine */
479 _mesa_memcpy(dMachine
, machine
, sizeof(struct fp_machine
));
481 /* Clear temporary registers */
482 _mesa_bzero( (void*) machine
->Temporaries
,
483 MAX_NV_FRAGMENT_PROGRAM_TEMPS
* 4 * sizeof(GLfloat
));
485 /* Add derivatives */
486 if (program
->InputsRead
& (1 << FRAG_ATTRIB_WPOS
)) {
487 GLfloat
*wpos
= (GLfloat
*) machine
->Inputs
[FRAG_ATTRIB_WPOS
];
491 wpos
[2] += span
->dzdx
;
492 wpos
[3] += span
->dwdx
;
497 wpos
[2] += span
->dzdy
;
498 wpos
[3] += span
->dwdy
;
501 if (program
->InputsRead
& (1 << FRAG_ATTRIB_COL0
)) {
502 GLfloat
*col0
= (GLfloat
*) machine
->Inputs
[FRAG_ATTRIB_COL0
];
504 col0
[0] += span
->drdx
* (1.0F
/ CHAN_MAXF
);
505 col0
[1] += span
->dgdx
* (1.0F
/ CHAN_MAXF
);
506 col0
[2] += span
->dbdx
* (1.0F
/ CHAN_MAXF
);
507 col0
[3] += span
->dadx
* (1.0F
/ CHAN_MAXF
);
510 col0
[0] += span
->drdy
* (1.0F
/ CHAN_MAXF
);
511 col0
[1] += span
->dgdy
* (1.0F
/ CHAN_MAXF
);
512 col0
[2] += span
->dbdy
* (1.0F
/ CHAN_MAXF
);
513 col0
[3] += span
->dady
* (1.0F
/ CHAN_MAXF
);
516 if (program
->InputsRead
& (1 << FRAG_ATTRIB_COL1
)) {
517 GLfloat
*col1
= (GLfloat
*) machine
->Inputs
[FRAG_ATTRIB_COL1
];
519 col1
[0] += span
->dsrdx
* (1.0F
/ CHAN_MAXF
);
520 col1
[1] += span
->dsgdx
* (1.0F
/ CHAN_MAXF
);
521 col1
[2] += span
->dsbdx
* (1.0F
/ CHAN_MAXF
);
522 col1
[3] += 0.0; /*XXX fix */
525 col1
[0] += span
->dsrdy
* (1.0F
/ CHAN_MAXF
);
526 col1
[1] += span
->dsgdy
* (1.0F
/ CHAN_MAXF
);
527 col1
[2] += span
->dsbdy
* (1.0F
/ CHAN_MAXF
);
528 col1
[3] += 0.0; /*XXX fix */
531 if (program
->InputsRead
& (1 << FRAG_ATTRIB_FOGC
)) {
532 GLfloat
*fogc
= (GLfloat
*) machine
->Inputs
[FRAG_ATTRIB_FOGC
];
534 fogc
[0] += span
->dfogdx
;
537 fogc
[0] += span
->dfogdy
;
540 for (u
= 0; u
< ctx
->Const
.MaxTextureCoordUnits
; u
++) {
541 if (program
->InputsRead
& (1 << (FRAG_ATTRIB_TEX0
+ u
))) {
542 GLfloat
*tex
= (GLfloat
*) machine
->Inputs
[FRAG_ATTRIB_TEX0
+ u
];
543 /* XXX perspective-correct interpolation */
545 tex
[0] += span
->texStepX
[u
][0];
546 tex
[1] += span
->texStepX
[u
][1];
547 tex
[2] += span
->texStepX
[u
][2];
548 tex
[3] += span
->texStepX
[u
][3];
551 tex
[0] += span
->texStepY
[u
][0];
552 tex
[1] += span
->texStepY
[u
][1];
553 tex
[2] += span
->texStepY
[u
][2];
554 tex
[3] += span
->texStepY
[u
][3];
559 /* init condition codes */
560 dMachine
->CondCodes
[0] = COND_EQ
;
561 dMachine
->CondCodes
[1] = COND_EQ
;
562 dMachine
->CondCodes
[2] = COND_EQ
;
563 dMachine
->CondCodes
[3] = COND_EQ
;
568 * Execute the given vertex program.
569 * NOTE: we do everything in single-precision floating point; we don't
570 * currently observe the single/half/fixed-precision qualifiers.
571 * \param ctx - rendering context
572 * \param program - the fragment program to execute
573 * \param machine - machine state (register file)
574 * \param maxInst - max number of instructions to execute
575 * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
578 execute_program( GLcontext
*ctx
,
579 const struct fragment_program
*program
, GLuint maxInst
,
580 struct fp_machine
*machine
, const struct sw_span
*span
,
586 printf("execute fragment program --------------------\n");
589 for (pc
= 0; pc
< maxInst
; pc
++) {
590 const struct fp_instruction
*inst
= program
->Instructions
+ pc
;
592 if (ctx
->FragmentProgram
.CallbackEnabled
&&
593 ctx
->FragmentProgram
.Callback
) {
594 ctx
->FragmentProgram
.CurrentPosition
= inst
->StringPos
;
595 ctx
->FragmentProgram
.Callback(program
->Base
.Target
,
596 ctx
->FragmentProgram
.CallbackData
);
599 switch (inst
->Opcode
) {
602 GLfloat a
[4], result
[4];
603 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
604 result
[0] = FABSF(a
[0]);
605 result
[1] = FABSF(a
[1]);
606 result
[2] = FABSF(a
[2]);
607 result
[3] = FABSF(a
[3]);
608 store_vector4( inst
, machine
, result
);
613 GLfloat a
[4], b
[4], result
[4];
614 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
615 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
616 result
[0] = a
[0] + b
[0];
617 result
[1] = a
[1] + b
[1];
618 result
[2] = a
[2] + b
[2];
619 result
[3] = a
[3] + b
[3];
620 store_vector4( inst
, machine
, result
);
625 GLfloat a
[4], b
[4], c
[4], result
[4];
626 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
627 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
628 fetch_vector4( ctx
, &inst
->SrcReg
[2], machine
, program
, c
);
629 result
[0] = a
[0] < 0.0F
? b
[0] : c
[0];
630 result
[1] = a
[1] < 0.0F
? b
[1] : c
[1];
631 result
[2] = a
[2] < 0.0F
? b
[2] : c
[2];
632 result
[3] = a
[3] < 0.0F
? b
[3] : c
[3];
633 store_vector4( inst
, machine
, result
);
638 GLfloat a
[4], result
[4];
639 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
640 result
[0] = result
[1] = result
[2] = result
[3] = (GLfloat
)_mesa_cos(a
[0]);
641 store_vector4( inst
, machine
, result
);
644 case FP_OPCODE_DDX
: /* Partial derivative with respect to X */
646 GLfloat a
[4], aNext
[4], result
[4];
647 struct fp_machine dMachine
;
648 if (!fetch_vector4_deriv(ctx
, &inst
->SrcReg
[0], span
, 'X',
650 /* This is tricky. Make a copy of the current machine state,
651 * increment the input registers by the dx or dy partial
652 * derivatives, then re-execute the program up to the
653 * preceeding instruction, then fetch the source register.
654 * Finally, find the difference in the register values for
655 * the original and derivative runs.
657 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
658 init_machine_deriv(ctx
, machine
, program
, span
,
660 execute_program(ctx
, program
, pc
, &dMachine
, span
, column
);
661 fetch_vector4( ctx
, &inst
->SrcReg
[0], &dMachine
, program
, aNext
);
662 result
[0] = aNext
[0] - a
[0];
663 result
[1] = aNext
[1] - a
[1];
664 result
[2] = aNext
[2] - a
[2];
665 result
[3] = aNext
[3] - a
[3];
667 store_vector4( inst
, machine
, result
);
670 case FP_OPCODE_DDY
: /* Partial derivative with respect to Y */
672 GLfloat a
[4], aNext
[4], result
[4];
673 struct fp_machine dMachine
;
674 if (!fetch_vector4_deriv(ctx
, &inst
->SrcReg
[0], span
, 'Y',
676 init_machine_deriv(ctx
, machine
, program
, span
,
678 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
679 execute_program(ctx
, program
, pc
, &dMachine
, span
, column
);
680 fetch_vector4( ctx
, &inst
->SrcReg
[0], &dMachine
, program
, aNext
);
681 result
[0] = aNext
[0] - a
[0];
682 result
[1] = aNext
[1] - a
[1];
683 result
[2] = aNext
[2] - a
[2];
684 result
[3] = aNext
[3] - a
[3];
686 store_vector4( inst
, machine
, result
);
691 GLfloat a
[4], b
[4], result
[4];
692 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
693 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
694 result
[0] = result
[1] = result
[2] = result
[3] =
695 a
[0] * b
[0] + a
[1] * b
[1] + a
[2] * b
[2];
696 store_vector4( inst
, machine
, result
);
698 printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
699 result
[0], a
[0], a
[1], a
[2], b
[0], b
[1], b
[2]);
705 GLfloat a
[4], b
[4], result
[4];
706 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
707 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
708 result
[0] = result
[1] = result
[2] = result
[3] =
709 a
[0] * b
[0] + a
[1] * b
[1] + a
[2] * b
[2] + a
[3] * b
[3];
710 store_vector4( inst
, machine
, result
);
715 GLfloat a
[4], b
[4], result
[4];
716 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
717 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
718 result
[0] = result
[1] = result
[2] = result
[3] =
719 a
[0] * b
[0] + a
[1] * b
[1] + a
[2] * b
[2] + b
[3];
720 store_vector4( inst
, machine
, result
);
723 case FP_OPCODE_DST
: /* Distance vector */
725 GLfloat a
[4], b
[4], result
[4];
726 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
727 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
729 result
[1] = a
[1] * b
[1];
732 store_vector4( inst
, machine
, result
);
735 case FP_OPCODE_EX2
: /* Exponential base 2 */
737 GLfloat a
[4], result
[4];
738 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
739 result
[0] = result
[1] = result
[2] = result
[3] =
740 (GLfloat
) _mesa_pow(2.0, a
[0]);
741 store_vector4( inst
, machine
, result
);
746 GLfloat a
[4], result
[4];
747 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
748 result
[0] = FLOORF(a
[0]);
749 result
[1] = FLOORF(a
[1]);
750 result
[2] = FLOORF(a
[2]);
751 result
[3] = FLOORF(a
[3]);
752 store_vector4( inst
, machine
, result
);
757 GLfloat a
[4], result
[4];
758 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
759 result
[0] = a
[0] - FLOORF(a
[0]);
760 result
[1] = a
[1] - FLOORF(a
[1]);
761 result
[2] = a
[2] - FLOORF(a
[2]);
762 result
[3] = a
[3] - FLOORF(a
[3]);
763 store_vector4( inst
, machine
, result
);
768 const GLuint
*swizzle
= inst
->DstReg
.CondSwizzle
;
769 const GLuint condMask
= inst
->DstReg
.CondMask
;
770 if (test_cc(machine
->CondCodes
[swizzle
[0]], condMask
) ||
771 test_cc(machine
->CondCodes
[swizzle
[1]], condMask
) ||
772 test_cc(machine
->CondCodes
[swizzle
[2]], condMask
) ||
773 test_cc(machine
->CondCodes
[swizzle
[3]], condMask
)) {
778 case FP_OPCODE_LG2
: /* log base 2 */
780 GLfloat a
[4], result
[4];
781 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
782 result
[0] = result
[1] = result
[2] = result
[3]
784 store_vector4( inst
, machine
, result
);
789 GLfloat a
[4], result
[4];
790 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
797 result
[2] = (a
[0] > 0.0F
) ? (GLfloat
) exp(a
[3] * log(a
[1])) : 0.0F
;
799 store_vector4( inst
, machine
, result
);
804 GLfloat a
[4], b
[4], c
[4], result
[4];
805 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
806 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
807 fetch_vector4( ctx
, &inst
->SrcReg
[2], machine
, program
, c
);
808 result
[0] = a
[0] * b
[0] + (1.0F
- a
[0]) * c
[0];
809 result
[1] = a
[1] * b
[1] + (1.0F
- a
[1]) * c
[1];
810 result
[2] = a
[2] * b
[2] + (1.0F
- a
[2]) * c
[2];
811 result
[3] = a
[3] * b
[3] + (1.0F
- a
[3]) * c
[3];
812 store_vector4( inst
, machine
, result
);
817 GLfloat a
[4], b
[4], c
[4], result
[4];
818 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
819 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
820 fetch_vector4( ctx
, &inst
->SrcReg
[2], machine
, program
, c
);
821 result
[0] = a
[0] * b
[0] + c
[0];
822 result
[1] = a
[1] * b
[1] + c
[1];
823 result
[2] = a
[2] * b
[2] + c
[2];
824 result
[3] = a
[3] * b
[3] + c
[3];
825 store_vector4( inst
, machine
, result
);
830 GLfloat a
[4], b
[4], result
[4];
831 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
832 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
833 result
[0] = MAX2(a
[0], b
[0]);
834 result
[1] = MAX2(a
[1], b
[1]);
835 result
[2] = MAX2(a
[2], b
[2]);
836 result
[3] = MAX2(a
[3], b
[3]);
837 store_vector4( inst
, machine
, result
);
842 GLfloat a
[4], b
[4], result
[4];
843 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
844 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
845 result
[0] = MIN2(a
[0], b
[0]);
846 result
[1] = MIN2(a
[1], b
[1]);
847 result
[2] = MIN2(a
[2], b
[2]);
848 result
[3] = MIN2(a
[3], b
[3]);
849 store_vector4( inst
, machine
, result
);
855 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, result
);
856 store_vector4( inst
, machine
, result
);
861 GLfloat a
[4], b
[4], result
[4];
862 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
863 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
864 result
[0] = a
[0] * b
[0];
865 result
[1] = a
[1] * b
[1];
866 result
[2] = a
[2] * b
[2];
867 result
[3] = a
[3] * b
[3];
868 store_vector4( inst
, machine
, result
);
870 printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
871 result
[0], result
[1], result
[2], result
[3],
872 a
[0], a
[1], a
[2], a
[3],
873 b
[0], b
[1], b
[2], b
[3]);
877 case FP_OPCODE_PK2H
: /* pack two 16-bit floats in one 32-bit float */
879 GLfloat a
[4], result
[4];
881 GLuint
*rawResult
= (GLuint
*) result
;
883 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
884 hx
= _mesa_float_to_half(a
[0]);
885 hy
= _mesa_float_to_half(a
[1]);
886 twoHalves
= hx
| (hy
<< 16);
887 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
889 store_vector4( inst
, machine
, result
);
892 case FP_OPCODE_PK2US
: /* pack two GLushorts into one 32-bit float */
894 GLfloat a
[4], result
[4];
895 GLuint usx
, usy
, *rawResult
= (GLuint
*) result
;
896 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
897 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
898 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
899 usx
= IROUND(a
[0] * 65535.0F
);
900 usy
= IROUND(a
[1] * 65535.0F
);
901 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
903 store_vector4( inst
, machine
, result
);
906 case FP_OPCODE_PK4B
: /* pack four GLbytes into one 32-bit float */
908 GLfloat a
[4], result
[4];
909 GLuint ubx
, uby
, ubz
, ubw
, *rawResult
= (GLuint
*) result
;
910 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
911 a
[0] = CLAMP(a
[0], -128.0F
/ 127.0F
, 1.0F
);
912 a
[1] = CLAMP(a
[1], -128.0F
/ 127.0F
, 1.0F
);
913 a
[2] = CLAMP(a
[2], -128.0F
/ 127.0F
, 1.0F
);
914 a
[3] = CLAMP(a
[3], -128.0F
/ 127.0F
, 1.0F
);
915 ubx
= IROUND(127.0F
* a
[0] + 128.0F
);
916 uby
= IROUND(127.0F
* a
[1] + 128.0F
);
917 ubz
= IROUND(127.0F
* a
[2] + 128.0F
);
918 ubw
= IROUND(127.0F
* a
[3] + 128.0F
);
919 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
920 = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
921 store_vector4( inst
, machine
, result
);
924 case FP_OPCODE_PK4UB
: /* pack four GLubytes into one 32-bit float */
926 GLfloat a
[4], result
[4];
927 GLuint ubx
, uby
, ubz
, ubw
, *rawResult
= (GLuint
*) result
;
928 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
929 a
[0] = CLAMP(a
[0], 0.0F
, 1.0F
);
930 a
[1] = CLAMP(a
[1], 0.0F
, 1.0F
);
931 a
[2] = CLAMP(a
[2], 0.0F
, 1.0F
);
932 a
[3] = CLAMP(a
[3], 0.0F
, 1.0F
);
933 ubx
= IROUND(255.0F
* a
[0]);
934 uby
= IROUND(255.0F
* a
[1]);
935 ubz
= IROUND(255.0F
* a
[2]);
936 ubw
= IROUND(255.0F
* a
[3]);
937 rawResult
[0] = rawResult
[1] = rawResult
[2] = rawResult
[3]
938 = ubx
| (uby
<< 8) | (ubz
<< 16) | (ubw
<< 24);
939 store_vector4( inst
, machine
, result
);
944 GLfloat a
[4], b
[4], result
[4];
945 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
946 fetch_vector1( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
947 result
[0] = result
[1] = result
[2] = result
[3]
948 = (GLfloat
)_mesa_pow(a
[0], b
[0]);
949 store_vector4( inst
, machine
, result
);
954 GLfloat a
[4], result
[4];
955 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
959 else if (IS_INF_OR_NAN(a
[0]))
960 printf("RCP(inf)\n");
962 result
[0] = result
[1] = result
[2] = result
[3]
964 store_vector4( inst
, machine
, result
);
969 GLfloat axis
[4], dir
[4], result
[4], tmp
[4];
970 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, axis
);
971 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, dir
);
972 tmp
[3] = axis
[0] * axis
[0]
975 tmp
[0] = (2.0F
* (axis
[0] * dir
[0] +
977 axis
[2] * dir
[2])) / tmp
[3];
978 result
[0] = tmp
[0] * axis
[0] - dir
[0];
979 result
[1] = tmp
[0] * axis
[1] - dir
[1];
980 result
[2] = tmp
[0] * axis
[2] - dir
[2];
981 /* result[3] is never written! XXX enforce in parser! */
982 store_vector4( inst
, machine
, result
);
985 case FP_OPCODE_RSQ
: /* 1 / sqrt() */
987 GLfloat a
[4], result
[4];
988 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
989 result
[0] = result
[1] = result
[2] = result
[3] = INV_SQRTF(a
[0]);
990 store_vector4( inst
, machine
, result
);
992 printf("RSQ %g = 1/sqrt(%g)\n", result
[0], a
[0]);
996 case FP_OPCODE_SCS
: /* sine and cos */
998 GLfloat a
[4], result
[4];
999 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1000 result
[0] = (GLfloat
)cos(a
[0]);
1001 result
[1] = (GLfloat
)sin(a
[0]);
1002 result
[2] = 0.0; /* undefined! */
1003 result
[3] = 0.0; /* undefined! */
1004 store_vector4( inst
, machine
, result
);
1007 case FP_OPCODE_SEQ
: /* set on equal */
1009 GLfloat a
[4], b
[4], result
[4];
1010 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1011 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1012 result
[0] = (a
[0] == b
[0]) ? 1.0F
: 0.0F
;
1013 result
[1] = (a
[1] == b
[1]) ? 1.0F
: 0.0F
;
1014 result
[2] = (a
[2] == b
[2]) ? 1.0F
: 0.0F
;
1015 result
[3] = (a
[3] == b
[3]) ? 1.0F
: 0.0F
;
1016 store_vector4( inst
, machine
, result
);
1019 case FP_OPCODE_SFL
: /* set false, operands ignored */
1021 static const GLfloat result
[4] = { 0.0F
, 0.0F
, 0.0F
, 0.0F
};
1022 store_vector4( inst
, machine
, result
);
1025 case FP_OPCODE_SGE
: /* set on greater or equal */
1027 GLfloat a
[4], b
[4], result
[4];
1028 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1029 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1030 result
[0] = (a
[0] >= b
[0]) ? 1.0F
: 0.0F
;
1031 result
[1] = (a
[1] >= b
[1]) ? 1.0F
: 0.0F
;
1032 result
[2] = (a
[2] >= b
[2]) ? 1.0F
: 0.0F
;
1033 result
[3] = (a
[3] >= b
[3]) ? 1.0F
: 0.0F
;
1034 store_vector4( inst
, machine
, result
);
1037 case FP_OPCODE_SGT
: /* set on greater */
1039 GLfloat a
[4], b
[4], result
[4];
1040 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1041 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1042 result
[0] = (a
[0] > b
[0]) ? 1.0F
: 0.0F
;
1043 result
[1] = (a
[1] > b
[1]) ? 1.0F
: 0.0F
;
1044 result
[2] = (a
[2] > b
[2]) ? 1.0F
: 0.0F
;
1045 result
[3] = (a
[3] > b
[3]) ? 1.0F
: 0.0F
;
1046 store_vector4( inst
, machine
, result
);
1051 GLfloat a
[4], result
[4];
1052 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1053 result
[0] = result
[1] = result
[2] =
1054 result
[3] = (GLfloat
)_mesa_sin(a
[0]);
1055 store_vector4( inst
, machine
, result
);
1058 case FP_OPCODE_SLE
: /* set on less or equal */
1060 GLfloat a
[4], b
[4], result
[4];
1061 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1062 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1063 result
[0] = (a
[0] <= b
[0]) ? 1.0F
: 0.0F
;
1064 result
[1] = (a
[1] <= b
[1]) ? 1.0F
: 0.0F
;
1065 result
[2] = (a
[2] <= b
[2]) ? 1.0F
: 0.0F
;
1066 result
[3] = (a
[3] <= b
[3]) ? 1.0F
: 0.0F
;
1067 store_vector4( inst
, machine
, result
);
1070 case FP_OPCODE_SLT
: /* set on less */
1072 GLfloat a
[4], b
[4], result
[4];
1073 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1074 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1075 result
[0] = (a
[0] < b
[0]) ? 1.0F
: 0.0F
;
1076 result
[1] = (a
[1] < b
[1]) ? 1.0F
: 0.0F
;
1077 result
[2] = (a
[2] < b
[2]) ? 1.0F
: 0.0F
;
1078 result
[3] = (a
[3] < b
[3]) ? 1.0F
: 0.0F
;
1079 store_vector4( inst
, machine
, result
);
1082 case FP_OPCODE_SNE
: /* set on not equal */
1084 GLfloat a
[4], b
[4], result
[4];
1085 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1086 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1087 result
[0] = (a
[0] != b
[0]) ? 1.0F
: 0.0F
;
1088 result
[1] = (a
[1] != b
[1]) ? 1.0F
: 0.0F
;
1089 result
[2] = (a
[2] != b
[2]) ? 1.0F
: 0.0F
;
1090 result
[3] = (a
[3] != b
[3]) ? 1.0F
: 0.0F
;
1091 store_vector4( inst
, machine
, result
);
1094 case FP_OPCODE_STR
: /* set true, operands ignored */
1096 static const GLfloat result
[4] = { 1.0F
, 1.0F
, 1.0F
, 1.0F
};
1097 store_vector4( inst
, machine
, result
);
1102 GLfloat a
[4], b
[4], result
[4];
1103 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1104 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1105 result
[0] = a
[0] - b
[0];
1106 result
[1] = a
[1] - b
[1];
1107 result
[2] = a
[2] - b
[2];
1108 result
[3] = a
[3] - b
[3];
1109 store_vector4( inst
, machine
, result
);
1114 const struct fp_src_register
*source
= &inst
->SrcReg
[0];
1115 const GLfloat
*src
= get_register_pointer(ctx
, source
,
1120 /* do extended swizzling here */
1121 for (i
= 0; i
< 3; i
++) {
1122 if (source
->Swizzle
[i
] == SWIZZLE_ZERO
)
1124 else if (source
->Swizzle
[i
] == SWIZZLE_ONE
)
1127 result
[i
] = -src
[source
->Swizzle
[i
]];
1128 if (source
->NegateBase
)
1129 result
[i
] = -result
[i
];
1131 store_vector4( inst
, machine
, result
);
1137 GLfloat texcoord
[4], color
[4];
1138 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, texcoord
);
1139 /* XXX: Undo perspective divide from interpolate_texcoords() */
1140 fetch_texel( ctx
, texcoord
,
1141 span
->array
->lambda
[inst
->TexSrcUnit
][column
],
1142 inst
->TexSrcUnit
, color
);
1143 store_vector4( inst
, machine
, color
);
1147 /* Texel lookup with LOD bias */
1149 GLfloat texcoord
[4], color
[4], bias
, lambda
;
1151 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, texcoord
);
1152 /* texcoord[3] is the bias to add to lambda */
1153 bias
= ctx
->Texture
.Unit
[inst
->TexSrcUnit
].LodBias
1154 + ctx
->Texture
.Unit
[inst
->TexSrcUnit
]._Current
->LodBias
1156 lambda
= span
->array
->lambda
[inst
->TexSrcUnit
][column
] + bias
;
1157 fetch_texel( ctx
, texcoord
, lambda
,
1158 inst
->TexSrcUnit
, color
);
1159 store_vector4( inst
, machine
, color
);
1163 /* Texture lookup w/ partial derivatives for LOD */
1165 GLfloat texcoord
[4], dtdx
[4], dtdy
[4], color
[4];
1166 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, texcoord
);
1167 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, dtdx
);
1168 fetch_vector4( ctx
, &inst
->SrcReg
[2], machine
, program
, dtdy
);
1169 fetch_texel_deriv( ctx
, texcoord
, dtdx
, dtdy
, inst
->TexSrcUnit
,
1171 store_vector4( inst
, machine
, color
);
1175 /* Texture lookup w/ perspective divide */
1177 GLfloat texcoord
[4], color
[4];
1178 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, texcoord
);
1179 /* Already did perspective divide in interpolate_texcoords() */
1180 fetch_texel( ctx
, texcoord
,
1181 span
->array
->lambda
[inst
->TexSrcUnit
][column
],
1182 inst
->TexSrcUnit
, color
);
1183 store_vector4( inst
, machine
, color
);
1186 case FP_OPCODE_UP2H
: /* unpack two 16-bit floats */
1188 GLfloat a
[4], result
[4];
1189 const GLuint
*rawBits
= (const GLuint
*) a
;
1191 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1192 hx
= rawBits
[0] & 0xffff;
1193 hy
= rawBits
[0] >> 16;
1194 result
[0] = result
[2] = _mesa_half_to_float(hx
);
1195 result
[1] = result
[3] = _mesa_half_to_float(hy
);
1196 store_vector4( inst
, machine
, result
);
1199 case FP_OPCODE_UP2US
: /* unpack two GLushorts */
1201 GLfloat a
[4], result
[4];
1202 const GLuint
*rawBits
= (const GLuint
*) a
;
1204 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1205 usx
= rawBits
[0] & 0xffff;
1206 usy
= rawBits
[0] >> 16;
1207 result
[0] = result
[2] = usx
* (1.0f
/ 65535.0f
);
1208 result
[1] = result
[3] = usy
* (1.0f
/ 65535.0f
);
1209 store_vector4( inst
, machine
, result
);
1212 case FP_OPCODE_UP4B
: /* unpack four GLbytes */
1214 GLfloat a
[4], result
[4];
1215 const GLuint
*rawBits
= (const GLuint
*) a
;
1216 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1217 result
[0] = (((rawBits
[0] >> 0) & 0xff) - 128) / 127.0F
;
1218 result
[0] = (((rawBits
[0] >> 8) & 0xff) - 128) / 127.0F
;
1219 result
[0] = (((rawBits
[0] >> 16) & 0xff) - 128) / 127.0F
;
1220 result
[0] = (((rawBits
[0] >> 24) & 0xff) - 128) / 127.0F
;
1221 store_vector4( inst
, machine
, result
);
1224 case FP_OPCODE_UP4UB
: /* unpack four GLubytes */
1226 GLfloat a
[4], result
[4];
1227 const GLuint
*rawBits
= (const GLuint
*) a
;
1228 fetch_vector1( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1229 result
[0] = ((rawBits
[0] >> 0) & 0xff) / 255.0F
;
1230 result
[0] = ((rawBits
[0] >> 8) & 0xff) / 255.0F
;
1231 result
[0] = ((rawBits
[0] >> 16) & 0xff) / 255.0F
;
1232 result
[0] = ((rawBits
[0] >> 24) & 0xff) / 255.0F
;
1233 store_vector4( inst
, machine
, result
);
1236 case FP_OPCODE_X2D
: /* 2-D matrix transform */
1238 GLfloat a
[4], b
[4], c
[4], result
[4];
1239 fetch_vector4( ctx
, &inst
->SrcReg
[0], machine
, program
, a
);
1240 fetch_vector4( ctx
, &inst
->SrcReg
[1], machine
, program
, b
);
1241 fetch_vector4( ctx
, &inst
->SrcReg
[2], machine
, program
, c
);
1242 result
[0] = a
[0] + b
[0] * c
[0] + b
[1] * c
[1];
1243 result
[1] = a
[1] + b
[0] * c
[2] + b
[1] * c
[3];
1244 result
[2] = a
[2] + b
[0] * c
[0] + b
[1] * c
[1];
1245 result
[3] = a
[3] + b
[0] * c
[2] + b
[1] * c
[3];
1246 store_vector4( inst
, machine
, result
);
1252 _mesa_problem(ctx
, "Bad opcode %d in _mesa_exec_fragment_program",
1254 return GL_TRUE
; /* return value doesn't matter */
1262 init_machine( GLcontext
*ctx
, struct fp_machine
*machine
,
1263 const struct fragment_program
*program
,
1264 const struct sw_span
*span
, GLuint col
)
1266 GLuint inputsRead
= program
->InputsRead
;
1269 if (ctx
->FragmentProgram
.CallbackEnabled
)
1272 /* Clear temporary registers */
1273 _mesa_bzero(machine
->Temporaries
,
1274 MAX_NV_FRAGMENT_PROGRAM_TEMPS
* 4 * sizeof(GLfloat
));
1276 /* Load input registers */
1277 if (inputsRead
& (1 << FRAG_ATTRIB_WPOS
)) {
1278 GLfloat
*wpos
= machine
->Inputs
[FRAG_ATTRIB_WPOS
];
1279 wpos
[0] = (GLfloat
) span
->x
+ col
;
1280 wpos
[1] = (GLfloat
) span
->y
;
1281 wpos
[2] = (GLfloat
) span
->array
->z
[col
] / ctx
->DepthMaxF
;
1282 wpos
[3] = span
->w
+ col
* span
->dwdx
;
1284 if (inputsRead
& (1 << FRAG_ATTRIB_COL0
)) {
1285 GLfloat
*col0
= machine
->Inputs
[FRAG_ATTRIB_COL0
];
1286 col0
[0] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][RCOMP
]);
1287 col0
[1] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][GCOMP
]);
1288 col0
[2] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][BCOMP
]);
1289 col0
[3] = CHAN_TO_FLOAT(span
->array
->rgba
[col
][ACOMP
]);
1291 if (inputsRead
& (1 << FRAG_ATTRIB_COL1
)) {
1292 GLfloat
*col1
= machine
->Inputs
[FRAG_ATTRIB_COL1
];
1293 col1
[0] = CHAN_TO_FLOAT(span
->array
->spec
[col
][RCOMP
]);
1294 col1
[1] = CHAN_TO_FLOAT(span
->array
->spec
[col
][GCOMP
]);
1295 col1
[2] = CHAN_TO_FLOAT(span
->array
->spec
[col
][BCOMP
]);
1296 col1
[3] = CHAN_TO_FLOAT(span
->array
->spec
[col
][ACOMP
]);
1298 if (inputsRead
& (1 << FRAG_ATTRIB_FOGC
)) {
1299 GLfloat
*fogc
= machine
->Inputs
[FRAG_ATTRIB_FOGC
];
1300 fogc
[0] = span
->array
->fog
[col
];
1305 for (u
= 0; u
< ctx
->Const
.MaxTextureCoordUnits
; u
++) {
1306 if (inputsRead
& (1 << (FRAG_ATTRIB_TEX0
+ u
))) {
1307 GLfloat
*tex
= machine
->Inputs
[FRAG_ATTRIB_TEX0
+ u
];
1308 /*ASSERT(ctx->Texture._EnabledCoordUnits & (1 << u));*/
1309 COPY_4V(tex
, span
->array
->texcoords
[u
][col
]);
1310 /*ASSERT(tex[0] != 0 || tex[1] != 0 || tex[2] != 0);*/
1314 /* init condition codes */
1315 machine
->CondCodes
[0] = COND_EQ
;
1316 machine
->CondCodes
[1] = COND_EQ
;
1317 machine
->CondCodes
[2] = COND_EQ
;
1318 machine
->CondCodes
[3] = COND_EQ
;
1323 _swrast_exec_nv_fragment_program( GLcontext
*ctx
, struct sw_span
*span
)
1325 const struct fragment_program
*program
= ctx
->FragmentProgram
.Current
;
1328 ctx
->_CurrentProgram
= GL_FRAGMENT_PROGRAM_ARB
; /* or NV, doesn't matter */
1330 for (i
= 0; i
< span
->end
; i
++) {
1331 if (span
->array
->mask
[i
]) {
1332 init_machine(ctx
, &ctx
->FragmentProgram
.Machine
,
1333 ctx
->FragmentProgram
.Current
, span
, i
);
1335 if (!execute_program(ctx
, program
, ~0,
1336 &ctx
->FragmentProgram
.Machine
, span
, i
)) {
1337 span
->array
->mask
[i
] = GL_FALSE
; /* killed fragment */
1340 /* Store output registers */
1342 const GLfloat
*colOut
1343 = ctx
->FragmentProgram
.Machine
.Outputs
[FRAG_OUTPUT_COLR
];
1344 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][RCOMP
], colOut
[0]);
1345 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][GCOMP
], colOut
[1]);
1346 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][BCOMP
], colOut
[2]);
1347 UNCLAMPED_FLOAT_TO_CHAN(span
->array
->rgba
[i
][ACOMP
], colOut
[3]);
1350 if (program
->OutputsWritten
& (1 << FRAG_OUTPUT_DEPR
))
1351 span
->array
->z
[i
] = IROUND(ctx
->FragmentProgram
.Machine
.Outputs
[FRAG_OUTPUT_DEPR
][0] * ctx
->DepthMaxF
);
1355 ctx
->_CurrentProgram
= 0;