1 /**************************************************************************
3 * Copyright 2003 VMware, Inc.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
28 #include "main/glheader.h"
29 #include "main/macros.h"
30 #include "main/enums.h"
32 #include "program/prog_instruction.h"
33 #include "program/prog_parameter.h"
34 #include "program/program.h"
35 #include "program/programopt.h"
36 #include "program/prog_print.h"
39 #include "tnl/t_context.h"
41 #include "intel_batchbuffer.h"
44 #include "i915_context.h"
45 #include "i915_program.h"
47 static const GLfloat sin_quad_constants
[2][4] = {
62 static const GLfloat sin_constants
[4] = { 1.0,
64 1.0 / (5 * 4 * 3 * 2 * 1),
65 -1.0 / (7 * 6 * 5 * 4 * 3 * 2 * 1)
68 /* 1, -1/2!, 1/4!, -1/6! */
69 static const GLfloat cos_constants
[4] = { 1.0,
71 1.0 / (4 * 3 * 2 * 1),
72 -1.0 / (6 * 5 * 4 * 3 * 2 * 1)
75 /* texcoord_mapping[unit] = index | TEXCOORD_{TEX,VAR} */
76 #define TEXCOORD_TEX (0<<7)
77 #define TEXCOORD_VAR (1<<7)
80 get_texcoord_mapping(struct i915_fragment_program
*p
, uint8_t texcoord
)
82 for (unsigned i
= 0; i
< p
->ctx
->Const
.MaxTextureCoordUnits
; i
++) {
83 if (p
->texcoord_mapping
[i
] == texcoord
)
88 return p
->ctx
->Const
.MaxTextureCoordUnits
- 1;
92 * Retrieve a ureg for the given source register. Will emit
93 * constants, apply swizzling and negation as needed.
96 src_vector(struct i915_fragment_program
*p
,
97 const struct prog_src_register
*source
,
98 const struct gl_fragment_program
*program
)
103 switch (source
->File
) {
107 case PROGRAM_TEMPORARY
:
108 if (source
->Index
>= I915_MAX_TEMPORARY
) {
109 i915_program_error(p
, "Exceeded max temporary reg: %d/%d",
110 source
->Index
, I915_MAX_TEMPORARY
);
113 src
= UREG(REG_TYPE_R
, source
->Index
);
116 switch (source
->Index
) {
117 case VARYING_SLOT_POS
:
118 src
= i915_emit_decl(p
, REG_TYPE_T
, p
->wpos_tex
, D0_CHANNEL_ALL
);
120 case VARYING_SLOT_COL0
:
121 src
= i915_emit_decl(p
, REG_TYPE_T
, T_DIFFUSE
, D0_CHANNEL_ALL
);
123 case VARYING_SLOT_COL1
:
124 src
= i915_emit_decl(p
, REG_TYPE_T
, T_SPECULAR
, D0_CHANNEL_XYZ
);
125 src
= swizzle(src
, X
, Y
, Z
, ONE
);
127 case VARYING_SLOT_FOGC
:
128 src
= i915_emit_decl(p
, REG_TYPE_T
, T_FOG_W
, D0_CHANNEL_W
);
129 src
= swizzle(src
, W
, ZERO
, ZERO
, ONE
);
131 case VARYING_SLOT_TEX0
:
132 case VARYING_SLOT_TEX1
:
133 case VARYING_SLOT_TEX2
:
134 case VARYING_SLOT_TEX3
:
135 case VARYING_SLOT_TEX4
:
136 case VARYING_SLOT_TEX5
:
137 case VARYING_SLOT_TEX6
:
138 case VARYING_SLOT_TEX7
:
139 unit
= get_texcoord_mapping(p
, (source
->Index
-
140 VARYING_SLOT_TEX0
) | TEXCOORD_TEX
);
141 src
= i915_emit_decl(p
, REG_TYPE_T
,
146 case VARYING_SLOT_VAR0
:
147 case VARYING_SLOT_VAR0
+ 1:
148 case VARYING_SLOT_VAR0
+ 2:
149 case VARYING_SLOT_VAR0
+ 3:
150 case VARYING_SLOT_VAR0
+ 4:
151 case VARYING_SLOT_VAR0
+ 5:
152 case VARYING_SLOT_VAR0
+ 6:
153 case VARYING_SLOT_VAR0
+ 7:
154 unit
= get_texcoord_mapping(p
, (source
->Index
-
155 VARYING_SLOT_VAR0
) | TEXCOORD_VAR
);
156 src
= i915_emit_decl(p
, REG_TYPE_T
,
162 i915_program_error(p
, "Bad source->Index: %d", source
->Index
);
168 switch (source
->Index
) {
169 case FRAG_RESULT_COLOR
:
170 case FRAG_RESULT_DATA0
:
171 src
= UREG(REG_TYPE_OC
, 0);
173 case FRAG_RESULT_DEPTH
:
174 src
= UREG(REG_TYPE_OD
, 0);
177 i915_program_error(p
, "Bad source->Index: %d", source
->Index
);
182 /* Various paramters and env values. All emitted to
183 * hardware as program constants.
185 case PROGRAM_CONSTANT
:
186 case PROGRAM_STATE_VAR
:
187 case PROGRAM_UNIFORM
:
188 src
= i915_emit_param4fv(p
,
189 &program
->Base
.Parameters
->ParameterValues
[source
->Index
][0].f
);
193 i915_program_error(p
, "Bad source->File: %d", source
->File
);
198 GET_SWZ(source
->Swizzle
, 0),
199 GET_SWZ(source
->Swizzle
, 1),
200 GET_SWZ(source
->Swizzle
, 2), GET_SWZ(source
->Swizzle
, 3));
204 GET_BIT(source
->Negate
, 0),
205 GET_BIT(source
->Negate
, 1),
206 GET_BIT(source
->Negate
, 2),
207 GET_BIT(source
->Negate
, 3));
214 get_result_vector(struct i915_fragment_program
*p
,
215 const struct prog_instruction
*inst
)
217 switch (inst
->DstReg
.File
) {
219 switch (inst
->DstReg
.Index
) {
220 case FRAG_RESULT_COLOR
:
221 case FRAG_RESULT_DATA0
:
222 return UREG(REG_TYPE_OC
, 0);
223 case FRAG_RESULT_DEPTH
:
224 p
->depth_written
= 1;
225 return UREG(REG_TYPE_OD
, 0);
227 i915_program_error(p
, "Bad inst->DstReg.Index: %d",
231 case PROGRAM_TEMPORARY
:
232 return UREG(REG_TYPE_R
, inst
->DstReg
.Index
);
234 i915_program_error(p
, "Bad inst->DstReg.File: %d", inst
->DstReg
.File
);
240 get_result_flags(const struct prog_instruction
*inst
)
245 flags
|= A0_DEST_SATURATE
;
246 if (inst
->DstReg
.WriteMask
& WRITEMASK_X
)
247 flags
|= A0_DEST_CHANNEL_X
;
248 if (inst
->DstReg
.WriteMask
& WRITEMASK_Y
)
249 flags
|= A0_DEST_CHANNEL_Y
;
250 if (inst
->DstReg
.WriteMask
& WRITEMASK_Z
)
251 flags
|= A0_DEST_CHANNEL_Z
;
252 if (inst
->DstReg
.WriteMask
& WRITEMASK_W
)
253 flags
|= A0_DEST_CHANNEL_W
;
259 translate_tex_src_target(struct i915_fragment_program
*p
, GLubyte bit
)
262 case TEXTURE_1D_INDEX
:
263 return D0_SAMPLE_TYPE_2D
;
264 case TEXTURE_2D_INDEX
:
265 return D0_SAMPLE_TYPE_2D
;
266 case TEXTURE_RECT_INDEX
:
267 return D0_SAMPLE_TYPE_2D
;
268 case TEXTURE_3D_INDEX
:
269 return D0_SAMPLE_TYPE_VOLUME
;
270 case TEXTURE_CUBE_INDEX
:
271 return D0_SAMPLE_TYPE_CUBE
;
273 i915_program_error(p
, "TexSrcBit: %d", bit
);
278 #define EMIT_TEX( OP ) \
280 GLuint dim = translate_tex_src_target( p, inst->TexSrcTarget ); \
281 const struct gl_fragment_program *program = &p->FragProg; \
282 GLuint unit = program->Base.SamplerUnits[inst->TexSrcUnit]; \
283 GLuint sampler = i915_emit_decl(p, REG_TYPE_S, \
285 GLuint coord = src_vector( p, &inst->SrcReg[0], program); \
288 i915_emit_texld( p, get_live_regs(p, inst), \
289 get_result_vector( p, inst ), \
290 get_result_flags( inst ), \
296 #define EMIT_ARITH( OP, N ) \
298 i915_emit_arith( p, \
300 get_result_vector( p, inst ), \
301 get_result_flags( inst ), 0, \
302 (N<1)?0:src_vector( p, &inst->SrcReg[0], program), \
303 (N<2)?0:src_vector( p, &inst->SrcReg[1], program), \
304 (N<3)?0:src_vector( p, &inst->SrcReg[2], program)); \
307 #define EMIT_1ARG_ARITH( OP ) EMIT_ARITH( OP, 1 )
308 #define EMIT_2ARG_ARITH( OP ) EMIT_ARITH( OP, 2 )
309 #define EMIT_3ARG_ARITH( OP ) EMIT_ARITH( OP, 3 )
312 * TODO: consider moving this into core
314 static bool calc_live_regs( struct i915_fragment_program
*p
)
316 const struct gl_fragment_program
*program
= &p
->FragProg
;
317 GLuint regsUsed
= ~((1 << I915_MAX_TEMPORARY
) - 1);
318 uint8_t live_components
[I915_MAX_TEMPORARY
] = { 0, };
321 for (i
= program
->Base
.NumInstructions
- 1; i
>= 0; i
--) {
322 struct prog_instruction
*inst
= &program
->Base
.Instructions
[i
];
323 int opArgs
= _mesa_num_inst_src_regs(inst
->Opcode
);
326 /* Register is written to: unmark as live for this and preceeding ops */
327 if (inst
->DstReg
.File
== PROGRAM_TEMPORARY
) {
328 if (inst
->DstReg
.Index
>= I915_MAX_TEMPORARY
)
331 live_components
[inst
->DstReg
.Index
] &= ~inst
->DstReg
.WriteMask
;
332 if (live_components
[inst
->DstReg
.Index
] == 0)
333 regsUsed
&= ~(1 << inst
->DstReg
.Index
);
336 for (a
= 0; a
< opArgs
; a
++) {
337 /* Register is read from: mark as live for this and preceeding ops */
338 if (inst
->SrcReg
[a
].File
== PROGRAM_TEMPORARY
) {
341 if (inst
->SrcReg
[a
].Index
>= I915_MAX_TEMPORARY
)
344 regsUsed
|= 1 << inst
->SrcReg
[a
].Index
;
346 for (c
= 0; c
< 4; c
++) {
347 const unsigned field
= GET_SWZ(inst
->SrcReg
[a
].Swizzle
, c
);
349 if (field
<= SWIZZLE_W
)
350 live_components
[inst
->SrcReg
[a
].Index
] |= (1U << field
);
355 p
->usedRegs
[i
] = regsUsed
;
361 static GLuint
get_live_regs( struct i915_fragment_program
*p
,
362 const struct prog_instruction
*inst
)
364 const struct gl_fragment_program
*program
= &p
->FragProg
;
365 GLuint nr
= inst
- program
->Base
.Instructions
;
367 return p
->usedRegs
[nr
];
371 /* Possible concerns:
373 * SIN, COS -- could use another taylor step?
374 * LIT -- results seem a little different to sw mesa
375 * LOG -- different to mesa on negative numbers, but this is conformant.
377 * Parse failures -- Mesa doesn't currently give a good indication
378 * internally whether a particular program string parsed or not. This
379 * can lead to confusion -- hopefully we cope with it ok now.
383 upload_program(struct i915_fragment_program
*p
)
385 const struct gl_fragment_program
*program
= &p
->FragProg
;
386 const struct prog_instruction
*inst
= program
->Base
.Instructions
;
388 if (INTEL_DEBUG
& DEBUG_WM
)
389 _mesa_print_program(&program
->Base
);
391 /* Is this a parse-failed program? Ensure a valid program is
392 * loaded, as the flagging of an error isn't sufficient to stop
393 * this being uploaded to hardware.
395 if (inst
[0].Opcode
== OPCODE_END
) {
396 GLuint tmp
= i915_get_utemp(p
);
399 UREG(REG_TYPE_OC
, 0),
400 A0_DEST_CHANNEL_ALL
, 0,
401 swizzle(tmp
, ONE
, ZERO
, ONE
, ONE
), 0, 0);
405 if (program
->Base
.NumInstructions
> I915_MAX_INSN
) {
406 i915_program_error(p
, "Exceeded max instructions (%d out of %d)",
407 program
->Base
.NumInstructions
, I915_MAX_INSN
);
411 /* Not always needed:
413 if (!calc_live_regs(p
)) {
414 i915_program_error(p
, "Could not allocate registers");
419 GLuint src0
, src1
, src2
, flags
;
420 GLuint tmp
= 0, dst
, consts0
= 0, consts1
= 0;
422 switch (inst
->Opcode
) {
424 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
427 get_result_vector(p
, inst
),
428 get_result_flags(inst
), 0,
429 src0
, negate(src0
, 1, 1, 1, 1), 0);
433 EMIT_2ARG_ARITH(A0_ADD
);
437 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
438 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
439 src2
= src_vector(p
, &inst
->SrcReg
[2], program
);
440 i915_emit_arith(p
, A0_CMP
, get_result_vector(p
, inst
), get_result_flags(inst
), 0, src0
, src2
, src1
); /* NOTE: order of src2, src1 */
444 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
445 tmp
= i915_get_utemp(p
);
446 consts0
= i915_emit_const4fv(p
, sin_quad_constants
[0]);
447 consts1
= i915_emit_const4fv(p
, sin_quad_constants
[1]);
449 /* Reduce range from repeating about [-pi,pi] to [-1,1] */
452 tmp
, A0_DEST_CHANNEL_X
, 0,
454 swizzle(consts1
, Z
, ZERO
, ZERO
, ZERO
), /* 1/(2pi) */
455 swizzle(consts0
, W
, ZERO
, ZERO
, ZERO
)); /* .75 */
457 i915_emit_arith(p
, A0_FRC
, tmp
, A0_DEST_CHANNEL_X
, 0, tmp
, 0, 0);
461 tmp
, A0_DEST_CHANNEL_X
, 0,
463 swizzle(consts0
, X
, ZERO
, ZERO
, ZERO
), /* 2 */
464 swizzle(consts0
, Y
, ZERO
, ZERO
, ZERO
)); /* -1 */
466 /* Compute COS with the same calculation used for SIN, but a
467 * different source range has been mapped to [-1,1] this time.
470 /* tmp.y = abs(tmp.x); {x, abs(x), 0, 0} */
473 tmp
, A0_DEST_CHANNEL_Y
, 0,
474 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
475 negate(swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
), 0, 1, 0, 0),
478 /* tmp.y = tmp.y * tmp.x; {x, x * abs(x), 0, 0} */
481 tmp
, A0_DEST_CHANNEL_Y
, 0,
482 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
486 /* tmp.x = tmp.xy DP sin_quad_constants[2].xy */
489 tmp
, A0_DEST_CHANNEL_X
, 0,
491 swizzle(consts1
, X
, Y
, ZERO
, ZERO
),
494 /* tmp.x now contains a first approximation (y). Now, weight it
495 * against tmp.y**2 to get closer.
499 tmp
, A0_DEST_CHANNEL_Y
, 0,
500 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
501 negate(swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
), 0, 1, 0, 0),
504 /* tmp.y = tmp.x * tmp.y - tmp.x; {y, y * abs(y) - y, 0, 0} */
507 tmp
, A0_DEST_CHANNEL_Y
, 0,
508 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
509 swizzle(tmp
, ZERO
, Y
, ZERO
, ZERO
),
510 negate(swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
), 0, 1, 0, 0));
512 /* result = .2225 * tmp.y + tmp.x =.2225(y * abs(y) - y) + y= */
515 get_result_vector(p
, inst
),
516 get_result_flags(inst
), 0,
517 swizzle(consts1
, W
, W
, W
, W
),
518 swizzle(tmp
, Y
, Y
, Y
, Y
),
519 swizzle(tmp
, X
, X
, X
, X
));
523 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
524 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
527 get_result_vector(p
, inst
),
528 get_result_flags(inst
), 0,
529 swizzle(src0
, X
, Y
, ZERO
, ZERO
),
530 swizzle(src1
, X
, Y
, ZERO
, ZERO
),
535 EMIT_2ARG_ARITH(A0_DP3
);
539 EMIT_2ARG_ARITH(A0_DP4
);
543 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
544 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
548 get_result_vector(p
, inst
),
549 get_result_flags(inst
), 0,
550 swizzle(src0
, X
, Y
, Z
, ONE
), src1
, 0);
554 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
555 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
557 /* result[0] = 1 * 1;
558 * result[1] = a[1] * b[1];
559 * result[2] = a[2] * 1;
560 * result[3] = 1 * b[3];
564 get_result_vector(p
, inst
),
565 get_result_flags(inst
), 0,
566 swizzle(src0
, ONE
, Y
, Z
, ONE
),
567 swizzle(src1
, ONE
, Y
, ONE
, W
), 0);
571 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
575 get_result_vector(p
, inst
),
576 get_result_flags(inst
), 0,
577 swizzle(src0
, X
, X
, X
, X
), 0, 0);
581 EMIT_1ARG_ARITH(A0_FLR
);
585 EMIT_1ARG_ARITH(A0_TRC
);
589 EMIT_1ARG_ARITH(A0_FRC
);
593 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
594 tmp
= i915_get_utemp(p
);
596 i915_emit_texld(p
, get_live_regs(p
, inst
),
597 tmp
, A0_DEST_CHANNEL_ALL
, /* use a dummy dest reg */
598 0, src0
, T0_TEXKILL
);
602 if (inst
->DstReg
.CondMask
== COND_TR
) {
603 tmp
= i915_get_utemp(p
);
605 /* The KIL instruction discards the fragment if any component of
606 * the source is < 0. Emit an immediate operand of {-1}.xywz.
608 i915_emit_texld(p
, get_live_regs(p
, inst
),
609 tmp
, A0_DEST_CHANNEL_ALL
,
610 0, /* use a dummy dest reg */
611 negate(swizzle(tmp
, ONE
, ONE
, ONE
, ONE
),
616 i915_program_error(p
, "Unsupported KIL_NV condition code: %d",
617 inst
->DstReg
.CondMask
);
622 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
626 get_result_vector(p
, inst
),
627 get_result_flags(inst
), 0,
628 swizzle(src0
, X
, X
, X
, X
), 0, 0);
632 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
633 tmp
= i915_get_utemp(p
);
635 /* tmp = max( a.xyzw, a.00zw )
636 * XXX: Clamp tmp.w to -128..128
638 * tmp.y = tmp.w * tmp.y
640 * result = cmp (a.11-x1, a.1x01, a.1xy1 )
642 i915_emit_arith(p
, A0_MAX
, tmp
, A0_DEST_CHANNEL_ALL
, 0,
643 src0
, swizzle(src0
, ZERO
, ZERO
, Z
, W
), 0);
645 i915_emit_arith(p
, A0_LOG
, tmp
, A0_DEST_CHANNEL_Y
, 0,
646 swizzle(tmp
, Y
, Y
, Y
, Y
), 0, 0);
648 i915_emit_arith(p
, A0_MUL
, tmp
, A0_DEST_CHANNEL_Y
, 0,
649 swizzle(tmp
, ZERO
, Y
, ZERO
, ZERO
),
650 swizzle(tmp
, ZERO
, W
, ZERO
, ZERO
), 0);
652 i915_emit_arith(p
, A0_EXP
, tmp
, A0_DEST_CHANNEL_Y
, 0,
653 swizzle(tmp
, Y
, Y
, Y
, Y
), 0, 0);
655 i915_emit_arith(p
, A0_CMP
,
656 get_result_vector(p
, inst
),
657 get_result_flags(inst
), 0,
658 negate(swizzle(tmp
, ONE
, ONE
, X
, ONE
), 0, 0, 1, 0),
659 swizzle(tmp
, ONE
, X
, ZERO
, ONE
),
660 swizzle(tmp
, ONE
, X
, Y
, ONE
));
665 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
666 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
667 src2
= src_vector(p
, &inst
->SrcReg
[2], program
);
668 flags
= get_result_flags(inst
);
669 tmp
= i915_get_utemp(p
);
676 * result = (-c)*a + tmp
678 i915_emit_arith(p
, A0_MAD
, tmp
,
679 flags
& A0_DEST_CHANNEL_ALL
, 0, src1
, src0
, src2
);
681 i915_emit_arith(p
, A0_MAD
,
682 get_result_vector(p
, inst
),
683 flags
, 0, negate(src2
, 1, 1, 1, 1), src0
, tmp
);
687 EMIT_3ARG_ARITH(A0_MAD
);
691 EMIT_2ARG_ARITH(A0_MAX
);
695 EMIT_2ARG_ARITH(A0_MIN
);
699 EMIT_1ARG_ARITH(A0_MOV
);
703 EMIT_2ARG_ARITH(A0_MUL
);
707 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
708 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
709 tmp
= i915_get_utemp(p
);
710 flags
= get_result_flags(inst
);
712 /* XXX: masking on intermediate values, here and elsewhere.
716 tmp
, A0_DEST_CHANNEL_X
, 0,
717 swizzle(src0
, X
, X
, X
, X
), 0, 0);
719 i915_emit_arith(p
, A0_MUL
, tmp
, A0_DEST_CHANNEL_X
, 0, tmp
, src1
, 0);
724 get_result_vector(p
, inst
),
725 flags
, 0, swizzle(tmp
, X
, X
, X
, X
), 0, 0);
730 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
734 get_result_vector(p
, inst
),
735 get_result_flags(inst
), 0,
736 swizzle(src0
, X
, X
, X
, X
), 0, 0);
741 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
745 get_result_vector(p
, inst
),
746 get_result_flags(inst
), 0,
747 swizzle(src0
, X
, X
, X
, X
), 0, 0);
751 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
752 tmp
= i915_get_utemp(p
);
755 * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1
756 * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x
757 * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x
758 * scs.x = DP4 t1, sin_constants
759 * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1
760 * scs.y = DP4 t1, cos_constants
764 tmp
, A0_DEST_CHANNEL_XY
, 0,
765 swizzle(src0
, X
, X
, ONE
, ONE
),
766 swizzle(src0
, X
, ONE
, ONE
, ONE
), 0);
770 tmp
, A0_DEST_CHANNEL_ALL
, 0,
771 swizzle(tmp
, X
, Y
, X
, Y
),
772 swizzle(tmp
, X
, X
, ONE
, ONE
), 0);
774 if (inst
->DstReg
.WriteMask
& WRITEMASK_Y
) {
777 if (inst
->DstReg
.WriteMask
& WRITEMASK_X
)
778 tmp1
= i915_get_utemp(p
);
784 tmp1
, A0_DEST_CHANNEL_ALL
, 0,
785 swizzle(tmp
, X
, Y
, Y
, W
),
786 swizzle(tmp
, X
, Z
, ONE
, ONE
), 0);
790 get_result_vector(p
, inst
),
791 A0_DEST_CHANNEL_Y
, 0,
792 swizzle(tmp1
, W
, Z
, Y
, X
),
793 i915_emit_const4fv(p
, sin_constants
), 0);
796 if (inst
->DstReg
.WriteMask
& WRITEMASK_X
) {
799 tmp
, A0_DEST_CHANNEL_XYZ
, 0,
800 swizzle(tmp
, X
, X
, Z
, ONE
),
801 swizzle(tmp
, Z
, ONE
, ONE
, ONE
), 0);
805 get_result_vector(p
, inst
),
806 A0_DEST_CHANNEL_X
, 0,
807 swizzle(tmp
, ONE
, Z
, Y
, X
),
808 i915_emit_const4fv(p
, cos_constants
), 0);
813 tmp
= i915_get_utemp(p
);
814 flags
= get_result_flags(inst
);
815 dst
= get_result_vector(p
, inst
);
817 /* If both operands are uniforms or constants, we get 5 instructions
820 * U[1] = MOV CONST[1]
821 * U[0].xyz = SGE CONST[0].xxxx, U[1]
822 * U[1] = MOV CONST[1].-x-y-z-w
823 * R[0].xyz = SGE CONST[0].-x-x-x-x, U[1]
824 * R[0].xyz = MUL R[0], U[0]
826 * This code is stupid. Instead of having the individual calls to
827 * i915_emit_arith generate the moves to utemps, do it in the caller.
828 * This results in code like:
830 * U[1] = MOV CONST[1]
831 * U[0].xyz = SGE CONST[0].xxxx, U[1]
832 * R[0].xyz = SGE CONST[0].-x-x-x-x, U[1].-x-y-z-w
833 * R[0].xyz = MUL R[0], U[0]
835 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
836 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
838 if (GET_UREG_TYPE(src0
) == REG_TYPE_CONST
839 && GET_UREG_TYPE(src1
) == REG_TYPE_CONST
) {
840 unsigned tmp
= i915_get_utemp(p
);
842 i915_emit_arith(p
, A0_MOV
, tmp
, A0_DEST_CHANNEL_ALL
, 0,
848 /* tmp = src1 >= src2 */
856 /* dst = src1 <= src2 */
861 negate(src0
, 1, 1, 1, 1),
862 negate(src1
, 1, 1, 1, 1),
864 /* dst = tmp && dst */
875 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
876 tmp
= i915_get_utemp(p
);
877 consts0
= i915_emit_const4fv(p
, sin_quad_constants
[0]);
878 consts1
= i915_emit_const4fv(p
, sin_quad_constants
[1]);
880 /* Reduce range from repeating about [-pi,pi] to [-1,1] */
883 tmp
, A0_DEST_CHANNEL_X
, 0,
885 swizzle(consts1
, Z
, ZERO
, ZERO
, ZERO
), /* 1/(2pi) */
886 swizzle(consts0
, Z
, ZERO
, ZERO
, ZERO
)); /* .5 */
888 i915_emit_arith(p
, A0_FRC
, tmp
, A0_DEST_CHANNEL_X
, 0, tmp
, 0, 0);
892 tmp
, A0_DEST_CHANNEL_X
, 0,
894 swizzle(consts0
, X
, ZERO
, ZERO
, ZERO
), /* 2 */
895 swizzle(consts0
, Y
, ZERO
, ZERO
, ZERO
)); /* -1 */
897 /* Compute sin using a quadratic and quartic. It gives continuity
898 * that repeating the Taylor series lacks every 2*pi, and has
901 * The idea was described at:
902 * http://www.devmaster.net/forums/showthread.php?t=5784
905 /* tmp.y = abs(tmp.x); {x, abs(x), 0, 0} */
908 tmp
, A0_DEST_CHANNEL_Y
, 0,
909 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
910 negate(swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
), 0, 1, 0, 0),
913 /* tmp.y = tmp.y * tmp.x; {x, x * abs(x), 0, 0} */
916 tmp
, A0_DEST_CHANNEL_Y
, 0,
917 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
921 /* tmp.x = tmp.xy DP sin_quad_constants[2].xy */
924 tmp
, A0_DEST_CHANNEL_X
, 0,
926 swizzle(consts1
, X
, Y
, ZERO
, ZERO
),
929 /* tmp.x now contains a first approximation (y). Now, weight it
930 * against tmp.y**2 to get closer.
934 tmp
, A0_DEST_CHANNEL_Y
, 0,
935 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
936 negate(swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
), 0, 1, 0, 0),
939 /* tmp.y = tmp.x * tmp.y - tmp.x; {y, y * abs(y) - y, 0, 0} */
942 tmp
, A0_DEST_CHANNEL_Y
, 0,
943 swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
),
944 swizzle(tmp
, ZERO
, Y
, ZERO
, ZERO
),
945 negate(swizzle(tmp
, ZERO
, X
, ZERO
, ZERO
), 0, 1, 0, 0));
947 /* result = .2225 * tmp.y + tmp.x =.2225(y * abs(y) - y) + y= */
950 get_result_vector(p
, inst
),
951 get_result_flags(inst
), 0,
952 swizzle(consts1
, W
, W
, W
, W
),
953 swizzle(tmp
, Y
, Y
, Y
, Y
),
954 swizzle(tmp
, X
, X
, X
, X
));
959 EMIT_2ARG_ARITH(A0_SGE
);
965 get_result_vector( p
, inst
),
966 get_result_flags( inst
), 0,
967 negate(src_vector( p
, &inst
->SrcReg
[0], program
),
969 negate(src_vector( p
, &inst
->SrcReg
[1], program
),
977 get_result_vector( p
, inst
),
978 get_result_flags( inst
), 0,
979 negate(src_vector( p
, &inst
->SrcReg
[0], program
),
981 negate(src_vector( p
, &inst
->SrcReg
[1], program
),
987 EMIT_2ARG_ARITH(A0_SLT
);
991 tmp
= i915_get_utemp(p
);
992 flags
= get_result_flags(inst
);
993 dst
= get_result_vector(p
, inst
);
995 /* If both operands are uniforms or constants, we get 5 instructions
998 * U[1] = MOV CONST[1]
999 * U[0].xyz = SLT CONST[0].xxxx, U[1]
1000 * U[1] = MOV CONST[1].-x-y-z-w
1001 * R[0].xyz = SLT CONST[0].-x-x-x-x, U[1]
1002 * R[0].xyz = MUL R[0], U[0]
1004 * This code is stupid. Instead of having the individual calls to
1005 * i915_emit_arith generate the moves to utemps, do it in the caller.
1006 * This results in code like:
1008 * U[1] = MOV CONST[1]
1009 * U[0].xyz = SLT CONST[0].xxxx, U[1]
1010 * R[0].xyz = SLT CONST[0].-x-x-x-x, U[1].-x-y-z-w
1011 * R[0].xyz = MUL R[0], U[0]
1013 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
1014 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
1016 if (GET_UREG_TYPE(src0
) == REG_TYPE_CONST
1017 && GET_UREG_TYPE(src1
) == REG_TYPE_CONST
) {
1018 unsigned tmp
= i915_get_utemp(p
);
1020 i915_emit_arith(p
, A0_MOV
, tmp
, A0_DEST_CHANNEL_ALL
, 0,
1026 /* tmp = src1 < src2 */
1034 /* dst = src1 > src2 */
1039 negate(src0
, 1, 1, 1, 1),
1040 negate(src1
, 1, 1, 1, 1),
1042 /* dst = tmp || dst */
1046 flags
| A0_DEST_SATURATE
, 0,
1053 dst
= get_result_vector(p
, inst
);
1054 flags
= get_result_flags(inst
);
1055 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
1056 tmp
= i915_get_utemp(p
);
1058 /* tmp = (src < 0.0) */
1064 swizzle(src0
, ZERO
, ZERO
, ZERO
, ZERO
),
1067 /* dst = (0.0 < src) */
1072 swizzle(src0
, ZERO
, ZERO
, ZERO
, ZERO
),
1076 /* dst = (src > 0.0) - (src < 0.0) */
1082 negate(tmp
, 1, 1, 1, 1),
1088 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
1089 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
1093 get_result_vector(p
, inst
),
1094 get_result_flags(inst
), 0,
1095 src0
, negate(src1
, 1, 1, 1, 1), 0);
1099 EMIT_1ARG_ARITH(A0_MOV
); /* extended swizzle handled natively */
1107 EMIT_TEX(T0_TEXLDB
);
1111 EMIT_TEX(T0_TEXLDP
);
1116 * result.x = src0.y * src1.z - src0.z * src1.y;
1117 * result.y = src0.z * src1.x - src0.x * src1.z;
1118 * result.z = src0.x * src1.y - src0.y * src1.x;
1121 src0
= src_vector(p
, &inst
->SrcReg
[0], program
);
1122 src1
= src_vector(p
, &inst
->SrcReg
[1], program
);
1123 tmp
= i915_get_utemp(p
);
1127 tmp
, A0_DEST_CHANNEL_ALL
, 0,
1128 swizzle(src0
, Z
, X
, Y
, ONE
),
1129 swizzle(src1
, Y
, Z
, X
, ONE
), 0);
1133 get_result_vector(p
, inst
),
1134 get_result_flags(inst
), 0,
1135 swizzle(src0
, Y
, Z
, X
, ONE
),
1136 swizzle(src1
, Z
, X
, Y
, ONE
),
1137 negate(tmp
, 1, 1, 1, 0));
1143 case OPCODE_BGNLOOP
:
1152 case OPCODE_ENDLOOP
:
1157 i915_program_error(p
, "Unsupported opcode: %s",
1158 _mesa_opcode_string(inst
->Opcode
));
1163 /* These opcodes are claimed as GLSL, NV_vp, and ARB_vp in
1164 * prog_instruction.h, but apparently GLSL doesn't ever emit them.
1165 * Instead, it translates to EX2 or LG2.
1169 /* These opcodes are claimed by GLSL in prog_instruction.h, but
1170 * only NV_vp/fp appears to emit them.
1173 i915_program_error(p
, "bad opcode: %s",
1174 _mesa_opcode_string(inst
->Opcode
));
1179 i915_release_utemps(p
);
1183 /* Rather than trying to intercept and jiggle depth writes during
1184 * emit, just move the value into its correct position at the end of
1188 fixup_depth_write(struct i915_fragment_program
*p
)
1190 if (p
->depth_written
) {
1191 GLuint depth
= UREG(REG_TYPE_OD
, 0);
1195 depth
, A0_DEST_CHANNEL_W
, 0,
1196 swizzle(depth
, X
, Y
, Z
, Z
), 0, 0);
1201 check_texcoord_mapping(struct i915_fragment_program
*p
)
1203 GLbitfield64 inputs
= p
->FragProg
.Base
.InputsRead
;
1206 for (unsigned i
= 0; i
< p
->ctx
->Const
.MaxTextureCoordUnits
; i
++) {
1207 if (inputs
& VARYING_BIT_TEX(i
)) {
1208 if (unit
>= p
->ctx
->Const
.MaxTextureCoordUnits
) {
1212 p
->texcoord_mapping
[unit
++] = i
| TEXCOORD_TEX
;
1214 if (inputs
& VARYING_BIT_VAR(i
)) {
1215 if (unit
>= p
->ctx
->Const
.MaxTextureCoordUnits
) {
1219 p
->texcoord_mapping
[unit
++] = i
| TEXCOORD_VAR
;
1223 if (unit
> p
->ctx
->Const
.MaxTextureCoordUnits
)
1224 i915_program_error(p
, "Too many texcoord units");
1228 check_wpos(struct i915_fragment_program
*p
)
1230 GLbitfield64 inputs
= p
->FragProg
.Base
.InputsRead
;
1236 if ((inputs
& VARYING_BIT_POS
) == 0)
1239 for (i
= 0; i
< p
->ctx
->Const
.MaxTextureCoordUnits
; i
++) {
1240 unit
+= !!(inputs
& VARYING_BIT_TEX(i
));
1241 unit
+= !!(inputs
& VARYING_BIT_VAR(i
));
1244 if (unit
< p
->ctx
->Const
.MaxTextureCoordUnits
)
1247 i915_program_error(p
, "No free texcoord for wpos value");
1252 translate_program(struct i915_fragment_program
*p
)
1254 struct i915_context
*i915
= I915_CONTEXT(p
->ctx
);
1256 if (INTEL_DEBUG
& DEBUG_WM
) {
1258 _mesa_print_program(&p
->FragProg
.Base
);
1262 i915_init_program(i915
, p
);
1263 check_texcoord_mapping(p
);
1266 fixup_depth_write(p
);
1267 i915_fini_program(p
);
1274 track_params(struct i915_fragment_program
*p
)
1279 _mesa_load_state_parameters(p
->ctx
, p
->FragProg
.Base
.Parameters
);
1281 for (i
= 0; i
< p
->nr_params
; i
++) {
1282 GLint reg
= p
->param
[i
].reg
;
1283 COPY_4V(p
->constant
[reg
], p
->param
[i
].values
);
1286 p
->params_uptodate
= 1;
1287 p
->on_hardware
= 0; /* overkill */
1292 i915BindProgram(struct gl_context
* ctx
, GLenum target
, struct gl_program
*prog
)
1294 if (target
== GL_FRAGMENT_PROGRAM_ARB
) {
1295 struct i915_context
*i915
= I915_CONTEXT(ctx
);
1296 struct i915_fragment_program
*p
= (struct i915_fragment_program
*) prog
;
1298 if (i915
->current_program
== p
)
1301 if (i915
->current_program
) {
1302 i915
->current_program
->on_hardware
= 0;
1303 i915
->current_program
->params_uptodate
= 0;
1306 i915
->current_program
= p
;
1308 assert(p
->on_hardware
== 0);
1309 assert(p
->params_uptodate
== 0);
1314 static struct gl_program
*
1315 i915NewProgram(struct gl_context
* ctx
, GLenum target
, GLuint id
)
1318 case GL_VERTEX_PROGRAM_ARB
:
1319 return _mesa_init_vertex_program(ctx
, CALLOC_STRUCT(gl_vertex_program
),
1322 case GL_FRAGMENT_PROGRAM_ARB
:{
1323 struct i915_fragment_program
*prog
=
1324 CALLOC_STRUCT(i915_fragment_program
);
1326 i915_init_program(I915_CONTEXT(ctx
), prog
);
1328 return _mesa_init_fragment_program(ctx
, &prog
->FragProg
,
1338 return _mesa_new_program(ctx
, target
, id
);
1343 i915DeleteProgram(struct gl_context
* ctx
, struct gl_program
*prog
)
1345 if (prog
->Target
== GL_FRAGMENT_PROGRAM_ARB
) {
1346 struct i915_context
*i915
= I915_CONTEXT(ctx
);
1347 struct i915_fragment_program
*p
= (struct i915_fragment_program
*) prog
;
1349 if (i915
->current_program
== p
)
1350 i915
->current_program
= 0;
1353 _mesa_delete_program(ctx
, prog
);
1358 i915IsProgramNative(struct gl_context
* ctx
, GLenum target
, struct gl_program
*prog
)
1360 if (target
== GL_FRAGMENT_PROGRAM_ARB
) {
1361 struct i915_fragment_program
*p
= (struct i915_fragment_program
*) prog
;
1364 translate_program(p
);
1373 i915ProgramStringNotify(struct gl_context
* ctx
,
1374 GLenum target
, struct gl_program
*prog
)
1376 if (target
== GL_FRAGMENT_PROGRAM_ARB
) {
1377 struct i915_fragment_program
*p
= (struct i915_fragment_program
*) prog
;
1381 (void) _tnl_program_string(ctx
, target
, prog
);
1383 /* XXX check if program is legal, within limits */
1388 i915SamplerUniformChange(struct gl_context
*ctx
,
1389 GLenum target
, struct gl_program
*prog
)
1391 i915ProgramStringNotify(ctx
, target
, prog
);
1395 i915_update_program(struct gl_context
*ctx
)
1397 struct intel_context
*intel
= intel_context(ctx
);
1398 struct i915_context
*i915
= i915_context(&intel
->ctx
);
1399 struct i915_fragment_program
*fp
=
1400 (struct i915_fragment_program
*) ctx
->FragmentProgram
._Current
;
1402 if (i915
->current_program
!= fp
) {
1403 if (i915
->current_program
) {
1404 i915
->current_program
->on_hardware
= 0;
1405 i915
->current_program
->params_uptodate
= 0;
1408 i915
->current_program
= fp
;
1411 if (!fp
->translated
)
1412 translate_program(fp
);
1414 FALLBACK(&i915
->intel
, I915_FALLBACK_PROGRAM
, fp
->error
);
1418 i915ValidateFragmentProgram(struct i915_context
*i915
)
1420 struct gl_context
*ctx
= &i915
->intel
.ctx
;
1421 struct intel_context
*intel
= intel_context(ctx
);
1422 TNLcontext
*tnl
= TNL_CONTEXT(ctx
);
1423 struct vertex_buffer
*VB
= &tnl
->vb
;
1425 struct i915_fragment_program
*p
=
1426 (struct i915_fragment_program
*) ctx
->FragmentProgram
._Current
;
1428 const GLbitfield64 inputsRead
= p
->FragProg
.Base
.InputsRead
;
1429 GLuint s4
= i915
->state
.Ctx
[I915_CTXREG_LIS4
] & ~S4_VFMT_MASK
;
1430 GLuint s2
= S2_TEXCOORD_NONE
;
1435 VB
->AttribPtr
[VERT_ATTRIB_POS
] = VB
->NdcPtr
;
1438 translate_program(p
);
1440 intel
->vertex_attr_count
= 0;
1441 intel
->wpos_offset
= 0;
1442 intel
->coloroffset
= 0;
1443 intel
->specoffset
= 0;
1445 if (inputsRead
& VARYING_BITS_TEX_ANY
|| p
->wpos_tex
!= -1) {
1446 EMIT_ATTR(_TNL_ATTRIB_POS
, EMIT_4F_VIEWPORT
, S4_VFMT_XYZW
, 16);
1449 EMIT_ATTR(_TNL_ATTRIB_POS
, EMIT_3F_VIEWPORT
, S4_VFMT_XYZ
, 12);
1452 /* Handle gl_PointSize builtin var here */
1453 if (ctx
->Point
._Attenuated
|| ctx
->VertexProgram
.PointSizeEnabled
)
1454 EMIT_ATTR(_TNL_ATTRIB_POINTSIZE
, EMIT_1F
, S4_VFMT_POINT_WIDTH
, 4);
1456 if (inputsRead
& VARYING_BIT_COL0
) {
1457 intel
->coloroffset
= offset
/ 4;
1458 EMIT_ATTR(_TNL_ATTRIB_COLOR0
, EMIT_4UB_4F_BGRA
, S4_VFMT_COLOR
, 4);
1461 if (inputsRead
& VARYING_BIT_COL1
) {
1462 intel
->specoffset
= offset
/ 4;
1463 EMIT_ATTR(_TNL_ATTRIB_COLOR1
, EMIT_4UB_4F_BGRA
, S4_VFMT_SPEC_FOG
, 4);
1466 if ((inputsRead
& VARYING_BIT_FOGC
)) {
1467 EMIT_ATTR(_TNL_ATTRIB_FOG
, EMIT_1F
, S4_VFMT_FOG_PARAM
, 4);
1470 for (i
= 0; i
< p
->ctx
->Const
.MaxTextureCoordUnits
; i
++) {
1471 if (inputsRead
& VARYING_BIT_TEX(i
)) {
1472 int unit
= get_texcoord_mapping(p
, i
| TEXCOORD_TEX
);
1473 int sz
= VB
->AttribPtr
[_TNL_ATTRIB_TEX0
+ i
]->size
;
1475 s2
&= ~S2_TEXCOORD_FMT(unit
, S2_TEXCOORD_FMT0_MASK
);
1476 s2
|= S2_TEXCOORD_FMT(unit
, SZ_TO_HW(sz
));
1478 EMIT_ATTR(_TNL_ATTRIB_TEX0
+ i
, EMIT_SZ(sz
), 0, sz
* 4);
1480 if (inputsRead
& VARYING_BIT_VAR(i
)) {
1481 int unit
= get_texcoord_mapping(p
, i
| TEXCOORD_VAR
);
1482 int sz
= VB
->AttribPtr
[_TNL_ATTRIB_GENERIC0
+ i
]->size
;
1484 s2
&= ~S2_TEXCOORD_FMT(unit
, S2_TEXCOORD_FMT0_MASK
);
1485 s2
|= S2_TEXCOORD_FMT(unit
, SZ_TO_HW(sz
));
1487 EMIT_ATTR(_TNL_ATTRIB_GENERIC0
+ i
, EMIT_SZ(sz
), 0, sz
* 4);
1489 if (i
== p
->wpos_tex
) {
1490 int wpos_size
= 4 * sizeof(float);
1491 /* If WPOS is required, duplicate the XYZ position data in an
1492 * unused texture coordinate:
1494 s2
&= ~S2_TEXCOORD_FMT(i
, S2_TEXCOORD_FMT0_MASK
);
1495 s2
|= S2_TEXCOORD_FMT(i
, SZ_TO_HW(wpos_size
));
1497 intel
->wpos_offset
= offset
;
1498 EMIT_PAD(wpos_size
);
1502 if (s2
!= i915
->state
.Ctx
[I915_CTXREG_LIS2
] ||
1503 s4
!= i915
->state
.Ctx
[I915_CTXREG_LIS4
]) {
1504 I915_STATECHANGE(i915
, I915_UPLOAD_CTX
);
1506 /* Must do this *after* statechange, so as not to affect
1507 * buffered vertices reliant on the old state:
1509 intel
->vertex_size
= _tnl_install_attrs(&intel
->ctx
,
1510 intel
->vertex_attrs
,
1511 intel
->vertex_attr_count
,
1512 intel
->ViewportMatrix
.m
, 0);
1514 assert(intel
->prim
.current_offset
== intel
->prim
.start_offset
);
1515 intel
->prim
.start_offset
= (intel
->prim
.current_offset
+ intel
->vertex_size
-1) / intel
->vertex_size
* intel
->vertex_size
;
1516 intel
->prim
.current_offset
= intel
->prim
.start_offset
;
1518 intel
->vertex_size
>>= 2;
1520 i915
->state
.Ctx
[I915_CTXREG_LIS2
] = s2
;
1521 i915
->state
.Ctx
[I915_CTXREG_LIS4
] = s4
;
1523 assert(intel
->vtbl
.check_vertex_size(intel
, intel
->vertex_size
));
1526 if (!p
->params_uptodate
)
1529 if (!p
->on_hardware
)
1530 i915_upload_program(i915
, p
);
1532 if (INTEL_DEBUG
& DEBUG_WM
) {
1534 i915_disassemble_program(i915
->state
.Program
, i915
->state
.ProgramSize
);
1539 i915InitFragProgFuncs(struct dd_function_table
*functions
)
1541 functions
->BindProgram
= i915BindProgram
;
1542 functions
->NewProgram
= i915NewProgram
;
1543 functions
->DeleteProgram
= i915DeleteProgram
;
1544 functions
->IsProgramNative
= i915IsProgramNative
;
1545 functions
->ProgramStringNotify
= i915ProgramStringNotify
;
1546 functions
->SamplerUniformChange
= i915SamplerUniformChange
;