2 * Copyright 2011 Joakim Sindholt <opensource@zhasha.com>
3 * Copyright 2013 Christoph Bumiller
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * on the rights to use, copy, modify, merge, publish, distribute, sub
9 * license, and/or sell copies of the Software, and to permit persons to whom
10 * the Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
20 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
21 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
22 * USE OR OTHER DEALINGS IN THE SOFTWARE. */
24 #include "nine_shader.h"
27 #include "nine_debug.h"
28 #include "nine_state.h"
30 #include "util/u_memory.h"
31 #include "util/u_inlines.h"
32 #include "pipe/p_shader_tokens.h"
33 #include "tgsi/tgsi_ureg.h"
34 #include "tgsi/tgsi_dump.h"
36 #define DBG_CHANNEL DBG_SHADER
38 #define DUMP(args...) _nine_debug_printf(DBG_CHANNEL, NULL, args)
41 struct shader_translator
;
43 typedef HRESULT (*translate_instruction_func
)(struct shader_translator
*);
45 static INLINE
const char *d3dsio_to_string(unsigned opcode
);
48 #define NINED3D_SM1_VS 0xfffe
49 #define NINED3D_SM1_PS 0xffff
51 #define NINE_MAX_COND_DEPTH 64
52 #define NINE_MAX_LOOP_DEPTH 64
54 #define NINED3DSP_END 0x0000ffff
56 #define NINED3DSPTYPE_FLOAT4 0
57 #define NINED3DSPTYPE_INT4 1
58 #define NINED3DSPTYPE_BOOL 2
60 #define NINED3DSPR_IMMEDIATE (D3DSPR_PREDICATE + 1)
62 #define NINED3DSP_WRITEMASK_MASK D3DSP_WRITEMASK_ALL
63 #define NINED3DSP_WRITEMASK_SHIFT 16
65 #define NINED3DSHADER_INST_PREDICATED (1 << 28)
67 #define NINED3DSHADER_REL_OP_GT 1
68 #define NINED3DSHADER_REL_OP_EQ 2
69 #define NINED3DSHADER_REL_OP_GE 3
70 #define NINED3DSHADER_REL_OP_LT 4
71 #define NINED3DSHADER_REL_OP_NE 5
72 #define NINED3DSHADER_REL_OP_LE 6
74 #define NINED3DSIO_OPCODE_FLAGS_SHIFT 16
75 #define NINED3DSIO_OPCODE_FLAGS_MASK (0xff << NINED3DSIO_OPCODE_FLAGS_SHIFT)
77 #define NINED3DSI_TEXLD_PROJECT 0x1
78 #define NINED3DSI_TEXLD_BIAS 0x2
80 #define NINED3DSP_WRITEMASK_0 0x1
81 #define NINED3DSP_WRITEMASK_1 0x2
82 #define NINED3DSP_WRITEMASK_2 0x4
83 #define NINED3DSP_WRITEMASK_3 0x8
84 #define NINED3DSP_WRITEMASK_ALL 0xf
86 #define NINED3DSP_NOSWIZZLE ((0 << 0) | (1 << 2) | (2 << 4) | (3 << 6))
88 #define NINE_SWIZZLE4(x,y,z,w) \
89 TGSI_SWIZZLE_##x, TGSI_SWIZZLE_##y, TGSI_SWIZZLE_##z, TGSI_SWIZZLE_##w
91 #define NINED3DSPDM_SATURATE (D3DSPDM_SATURATE >> D3DSP_DSTMOD_SHIFT)
92 #define NINED3DSPDM_PARTIALP (D3DSPDM_PARTIALPRECISION >> D3DSP_DSTMOD_SHIFT)
93 #define NINED3DSPDM_CENTROID (D3DSPDM_MSAMPCENTROID >> D3DSP_DSTMOD_SHIFT)
96 * NEG all, not ps: m3x2, m3x3, m3x4, m4x3, m4x4
97 * BIAS <= PS 1.4 (x-0.5)
98 * BIASNEG <= PS 1.4 (-(x-0.5))
99 * SIGN <= PS 1.4 (2(x-0.5))
100 * SIGNNEG <= PS 1.4 (-2(x-0.5))
101 * COMP <= PS 1.4 (1-x)
103 * X2NEG = PS 1.4 (-2x)
104 * DZ <= PS 1.4, tex{ld,crd} (.xy/.z), z=0 => .11
105 * DW <= PS 1.4, tex{ld,crd} (.xy/.w), w=0 => .11
106 * ABS >= SM 3.0 (abs(x))
107 * ABSNEG >= SM 3.0 (-abs(x))
108 * NOT >= SM 2.0 pedication only
110 #define NINED3DSPSM_NONE (D3DSPSM_NONE >> D3DSP_SRCMOD_SHIFT)
111 #define NINED3DSPSM_NEG (D3DSPSM_NEG >> D3DSP_SRCMOD_SHIFT)
112 #define NINED3DSPSM_BIAS (D3DSPSM_BIAS >> D3DSP_SRCMOD_SHIFT)
113 #define NINED3DSPSM_BIASNEG (D3DSPSM_BIASNEG >> D3DSP_SRCMOD_SHIFT)
114 #define NINED3DSPSM_SIGN (D3DSPSM_SIGN >> D3DSP_SRCMOD_SHIFT)
115 #define NINED3DSPSM_SIGNNEG (D3DSPSM_SIGNNEG >> D3DSP_SRCMOD_SHIFT)
116 #define NINED3DSPSM_COMP (D3DSPSM_COMP >> D3DSP_SRCMOD_SHIFT)
117 #define NINED3DSPSM_X2 (D3DSPSM_X2 >> D3DSP_SRCMOD_SHIFT)
118 #define NINED3DSPSM_X2NEG (D3DSPSM_X2NEG >> D3DSP_SRCMOD_SHIFT)
119 #define NINED3DSPSM_DZ (D3DSPSM_DZ >> D3DSP_SRCMOD_SHIFT)
120 #define NINED3DSPSM_DW (D3DSPSM_DW >> D3DSP_SRCMOD_SHIFT)
121 #define NINED3DSPSM_ABS (D3DSPSM_ABS >> D3DSP_SRCMOD_SHIFT)
122 #define NINED3DSPSM_ABSNEG (D3DSPSM_ABSNEG >> D3DSP_SRCMOD_SHIFT)
123 #define NINED3DSPSM_NOT (D3DSPSM_NOT >> D3DSP_SRCMOD_SHIFT)
125 static const char *sm1_mod_str
[] =
127 [NINED3DSPSM_NONE
] = "",
128 [NINED3DSPSM_NEG
] = "-",
129 [NINED3DSPSM_BIAS
] = "bias",
130 [NINED3DSPSM_BIASNEG
] = "biasneg",
131 [NINED3DSPSM_SIGN
] = "sign",
132 [NINED3DSPSM_SIGNNEG
] = "signneg",
133 [NINED3DSPSM_COMP
] = "comp",
134 [NINED3DSPSM_X2
] = "x2",
135 [NINED3DSPSM_X2NEG
] = "x2neg",
136 [NINED3DSPSM_DZ
] = "dz",
137 [NINED3DSPSM_DW
] = "dw",
138 [NINED3DSPSM_ABS
] = "abs",
139 [NINED3DSPSM_ABSNEG
] = "-abs",
140 [NINED3DSPSM_NOT
] = "not"
144 sm1_dump_writemask(BYTE mask
)
146 if (mask
& 1) DUMP("x"); else DUMP("_");
147 if (mask
& 2) DUMP("y"); else DUMP("_");
148 if (mask
& 4) DUMP("z"); else DUMP("_");
149 if (mask
& 8) DUMP("w"); else DUMP("_");
153 sm1_dump_swizzle(BYTE s
)
155 char c
[4] = { 'x', 'y', 'z', 'w' };
157 c
[(s
>> 0) & 3], c
[(s
>> 2) & 3], c
[(s
>> 4) & 3], c
[(s
>> 6) & 3]);
160 static const char sm1_file_char
[] =
163 [D3DSPR_INPUT
] = 'v',
164 [D3DSPR_CONST
] = 'c',
166 [D3DSPR_RASTOUT
] = 'R',
167 [D3DSPR_ATTROUT
] = 'D',
168 [D3DSPR_OUTPUT
] = 'o',
169 [D3DSPR_CONSTINT
] = 'I',
170 [D3DSPR_COLOROUT
] = 'C',
171 [D3DSPR_DEPTHOUT
] = 'D',
172 [D3DSPR_SAMPLER
] = 's',
173 [D3DSPR_CONST2
] = 'c',
174 [D3DSPR_CONST3
] = 'c',
175 [D3DSPR_CONST4
] = 'c',
176 [D3DSPR_CONSTBOOL
] = 'B',
178 [D3DSPR_TEMPFLOAT16
] = 'h',
179 [D3DSPR_MISCTYPE
] = 'M',
180 [D3DSPR_LABEL
] = 'X',
181 [D3DSPR_PREDICATE
] = 'p'
185 sm1_dump_reg(BYTE file
, INT index
)
191 case D3DSPR_COLOROUT
:
194 case D3DSPR_DEPTHOUT
:
198 DUMP("oRast%i", index
);
200 case D3DSPR_CONSTINT
:
201 DUMP("iconst[%i]", index
);
203 case D3DSPR_CONSTBOOL
:
204 DUMP("bconst[%i]", index
);
207 DUMP("%c%i", sm1_file_char
[file
], index
);
215 struct sm1_src_param
*rel
;
228 sm1_parse_immediate(struct shader_translator
*, struct sm1_src_param
*);
233 struct sm1_src_param
*rel
;
237 int8_t shift
; /* sint4 */
242 assert_replicate_swizzle(const struct ureg_src
*reg
)
244 assert(reg
->SwizzleY
== reg
->SwizzleX
&&
245 reg
->SwizzleZ
== reg
->SwizzleX
&&
246 reg
->SwizzleW
== reg
->SwizzleX
);
250 sm1_dump_immediate(const struct sm1_src_param
*param
)
252 switch (param
->type
) {
253 case NINED3DSPTYPE_FLOAT4
:
254 DUMP("{ %f %f %f %f }",
255 param
->imm
.f
[0], param
->imm
.f
[1],
256 param
->imm
.f
[2], param
->imm
.f
[3]);
258 case NINED3DSPTYPE_INT4
:
259 DUMP("{ %i %i %i %i }",
260 param
->imm
.i
[0], param
->imm
.i
[1],
261 param
->imm
.i
[2], param
->imm
.i
[3]);
263 case NINED3DSPTYPE_BOOL
:
264 DUMP("%s", param
->imm
.b
? "TRUE" : "FALSE");
273 sm1_dump_src_param(const struct sm1_src_param
*param
)
275 if (param
->file
== NINED3DSPR_IMMEDIATE
) {
276 assert(!param
->mod
&&
278 param
->swizzle
== NINED3DSP_NOSWIZZLE
);
279 sm1_dump_immediate(param
);
284 DUMP("%s(", sm1_mod_str
[param
->mod
]);
286 DUMP("%c[", sm1_file_char
[param
->file
]);
287 sm1_dump_src_param(param
->rel
);
288 DUMP("+%i]", param
->idx
);
290 sm1_dump_reg(param
->file
, param
->idx
);
294 if (param
->swizzle
!= NINED3DSP_NOSWIZZLE
) {
296 sm1_dump_swizzle(param
->swizzle
);
301 sm1_dump_dst_param(const struct sm1_dst_param
*param
)
303 if (param
->mod
& NINED3DSPDM_SATURATE
)
305 if (param
->mod
& NINED3DSPDM_PARTIALP
)
307 if (param
->mod
& NINED3DSPDM_CENTROID
)
309 if (param
->shift
< 0)
310 DUMP("/%u ", 1 << -param
->shift
);
311 if (param
->shift
> 0)
312 DUMP("*%u ", 1 << param
->shift
);
315 DUMP("%c[", sm1_file_char
[param
->file
]);
316 sm1_dump_src_param(param
->rel
);
317 DUMP("+%i]", param
->idx
);
319 sm1_dump_reg(param
->file
, param
->idx
);
321 if (param
->mask
!= NINED3DSP_WRITEMASK_ALL
) {
323 sm1_dump_writemask(param
->mask
);
329 struct sm1_dst_param reg
;
337 /* NOTE: 0 is a valid TGSI opcode, but if handler is set, this parameter
338 * should be ignored completely */
340 unsigned opcode
; /* TGSI_OPCODE_x */
342 /* versions are still set even handler is set */
346 } vert_version
, frag_version
;
348 /* number of regs parsed outside of special handler */
352 /* some instructions don't map perfectly, so use a special handler */
353 translate_instruction_func handler
;
356 struct sm1_instruction
358 D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode
;
364 struct sm1_src_param src
[4];
365 struct sm1_src_param src_rel
[4];
366 struct sm1_src_param pred
;
367 struct sm1_src_param dst_rel
[1];
368 struct sm1_dst_param dst
[1];
370 struct sm1_op_info
*info
;
374 sm1_dump_instruction(struct sm1_instruction
*insn
, unsigned indent
)
378 /* no info stored for these: */
379 if (insn
->opcode
== D3DSIO_DCL
)
381 for (i
= 0; i
< indent
; ++i
)
384 if (insn
->predicated
) {
386 sm1_dump_src_param(&insn
->pred
);
389 DUMP("%s", d3dsio_to_string(insn
->opcode
));
391 switch (insn
->opcode
) {
393 DUMP(insn
->flags
== NINED3DSI_TEXLD_PROJECT
? "p" : "b");
396 DUMP("_%x", insn
->flags
);
404 for (i
= 0; i
< insn
->ndst
&& i
< Elements(insn
->dst
); ++i
) {
405 sm1_dump_dst_param(&insn
->dst
[i
]);
409 for (i
= 0; i
< insn
->nsrc
&& i
< Elements(insn
->src
); ++i
) {
410 sm1_dump_src_param(&insn
->src
[i
]);
413 if (insn
->opcode
== D3DSIO_DEF
||
414 insn
->opcode
== D3DSIO_DEFI
||
415 insn
->opcode
== D3DSIO_DEFB
)
416 sm1_dump_immediate(&insn
->src
[0]);
421 struct sm1_local_const
432 struct shader_translator
434 const DWORD
*byte_code
;
436 const DWORD
*parse_next
;
438 struct ureg_program
*ureg
;
445 unsigned processor
; /* TGSI_PROCESSOR_VERTEX/FRAMGENT */
447 boolean native_integers
;
448 boolean inline_subroutines
;
450 boolean want_texcoord
;
452 unsigned texcoord_sn
;
454 struct sm1_instruction insn
; /* current instruction */
458 struct ureg_dst oPos
;
459 struct ureg_dst oFog
;
460 struct ureg_dst oPts
;
461 struct ureg_dst oCol
[4];
462 struct ureg_dst o
[PIPE_MAX_SHADER_OUTPUTS
];
463 struct ureg_dst oDepth
;
464 struct ureg_src v
[PIPE_MAX_SHADER_INPUTS
];
465 struct ureg_src vPos
;
466 struct ureg_src vFace
;
470 struct ureg_dst tS
[8]; /* texture stage registers */
471 struct ureg_dst tdst
; /* scratch dst if we need extra modifiers */
472 struct ureg_dst t
[5]; /* scratch TEMPs */
473 struct ureg_src vC
[2]; /* PS color in */
474 struct ureg_src vT
[8]; /* PS texcoord in */
475 struct ureg_dst rL
[NINE_MAX_LOOP_DEPTH
]; /* loop ctr */
476 struct ureg_dst aL
[NINE_MAX_LOOP_DEPTH
]; /* loop ctr ADDR register */
478 unsigned num_temp
; /* Elements(regs.r) */
479 unsigned num_scratch
;
481 unsigned loop_depth_max
;
483 unsigned loop_labels
[NINE_MAX_LOOP_DEPTH
];
484 unsigned cond_labels
[NINE_MAX_COND_DEPTH
];
486 unsigned *inst_labels
; /* LABEL op */
487 unsigned num_inst_labels
;
489 unsigned sampler_targets
[NINE_MAX_SAMPLERS
]; /* TGSI_TEXTURE_x */
491 struct sm1_local_const
*lconstf
;
492 unsigned num_lconstf
;
493 struct sm1_local_const lconsti
[NINE_MAX_CONST_I
];
494 struct sm1_local_const lconstb
[NINE_MAX_CONST_B
];
496 boolean indirect_const_access
;
498 struct nine_shader_info
*info
;
500 int16_t op_info_map
[D3DSIO_BREAKP
+ 1];
503 #define IS_VS (tx->processor == TGSI_PROCESSOR_VERTEX)
504 #define IS_PS (tx->processor == TGSI_PROCESSOR_FRAGMENT)
507 sm1_read_semantic(struct shader_translator
*, struct sm1_semantic
*);
510 sm1_instruction_check(const struct sm1_instruction
*insn
)
512 if (insn
->opcode
== D3DSIO_CRS
)
514 if (insn
->dst
[0].mask
& NINED3DSP_WRITEMASK_3
)
522 tx_lconstf(struct shader_translator
*tx
, struct ureg_src
*src
, INT index
)
525 assert(index
>= 0 && index
< (NINE_MAX_CONST_F
* 2));
526 for (i
= 0; i
< tx
->num_lconstf
; ++i
) {
527 if (tx
->lconstf
[i
].idx
== index
) {
528 *src
= tx
->lconstf
[i
].reg
;
535 tx_lconsti(struct shader_translator
*tx
, struct ureg_src
*src
, INT index
)
537 assert(index
>= 0 && index
< NINE_MAX_CONST_I
);
538 if (tx
->lconsti
[index
].idx
== index
)
539 *src
= tx
->lconsti
[index
].reg
;
540 return tx
->lconsti
[index
].idx
== index
;
543 tx_lconstb(struct shader_translator
*tx
, struct ureg_src
*src
, INT index
)
545 assert(index
>= 0 && index
< NINE_MAX_CONST_B
);
546 if (tx
->lconstb
[index
].idx
== index
)
547 *src
= tx
->lconstb
[index
].reg
;
548 return tx
->lconstb
[index
].idx
== index
;
552 tx_set_lconstf(struct shader_translator
*tx
, INT index
, float f
[4])
556 /* Anno1404 sets out of range constants. */
557 assert(index
>= 0 && index
< (NINE_MAX_CONST_F
* 2));
558 if (index
>= NINE_MAX_CONST_F
)
559 WARN("lconstf index %i too high, indirect access won't work\n", index
);
561 for (n
= 0; n
< tx
->num_lconstf
; ++n
)
562 if (tx
->lconstf
[n
].idx
== index
)
564 if (n
== tx
->num_lconstf
) {
566 tx
->lconstf
= REALLOC(tx
->lconstf
,
567 (n
+ 0) * sizeof(tx
->lconstf
[0]),
568 (n
+ 8) * sizeof(tx
->lconstf
[0]));
573 tx
->lconstf
[n
].idx
= index
;
574 tx
->lconstf
[n
].reg
= ureg_imm4f(tx
->ureg
, f
[0], f
[1], f
[2], f
[3]);
576 memcpy(tx
->lconstf
[n
].imm
.f
, f
, sizeof(tx
->lconstf
[n
].imm
.f
));
579 tx_set_lconsti(struct shader_translator
*tx
, INT index
, int i
[4])
581 assert(index
>= 0 && index
< NINE_MAX_CONST_I
);
582 tx
->lconsti
[index
].idx
= index
;
583 tx
->lconsti
[index
].reg
= tx
->native_integers
?
584 ureg_imm4i(tx
->ureg
, i
[0], i
[1], i
[2], i
[3]) :
585 ureg_imm4f(tx
->ureg
, i
[0], i
[1], i
[2], i
[3]);
588 tx_set_lconstb(struct shader_translator
*tx
, INT index
, BOOL b
)
590 assert(index
>= 0 && index
< NINE_MAX_CONST_B
);
591 tx
->lconstb
[index
].idx
= index
;
592 tx
->lconstb
[index
].reg
= tx
->native_integers
?
593 ureg_imm1u(tx
->ureg
, b
? 0xffffffff : 0) :
594 ureg_imm1f(tx
->ureg
, b
? 1.0f
: 0.0f
);
597 static INLINE
struct ureg_dst
598 tx_scratch(struct shader_translator
*tx
)
600 assert(tx
->num_scratch
< Elements(tx
->regs
.t
));
601 if (ureg_dst_is_undef(tx
->regs
.t
[tx
->num_scratch
]))
602 tx
->regs
.t
[tx
->num_scratch
] = ureg_DECL_local_temporary(tx
->ureg
);
603 return tx
->regs
.t
[tx
->num_scratch
++];
606 static INLINE
struct ureg_dst
607 tx_scratch_scalar(struct shader_translator
*tx
)
609 return ureg_writemask(tx_scratch(tx
), TGSI_WRITEMASK_X
);
612 static INLINE
struct ureg_src
613 tx_src_scalar(struct ureg_dst dst
)
615 struct ureg_src src
= ureg_src(dst
);
616 int c
= ffs(dst
.WriteMask
) - 1;
617 if (dst
.WriteMask
== (1 << c
))
618 src
= ureg_scalar(src
, c
);
622 /* Need to declare all constants if indirect addressing is used,
623 * otherwise we could scan the shader to determine the maximum.
624 * TODO: It doesn't really matter for nv50 so I won't do the scan,
625 * but radeon drivers might care, if they don't infer it from TGSI.
628 tx_decl_constants(struct shader_translator
*tx
)
632 for (i
= 0; i
< NINE_MAX_CONST_F
; ++i
)
633 ureg_DECL_constant(tx
->ureg
, n
++);
634 for (i
= 0; i
< NINE_MAX_CONST_I
; ++i
)
635 ureg_DECL_constant(tx
->ureg
, n
++);
636 for (i
= 0; i
< (NINE_MAX_CONST_B
/ 4); ++i
)
637 ureg_DECL_constant(tx
->ureg
, n
++);
641 tx_temp_alloc(struct shader_translator
*tx
, INT idx
)
644 if (idx
>= tx
->num_temp
) {
645 unsigned k
= tx
->num_temp
;
646 unsigned n
= idx
+ 1;
647 tx
->regs
.r
= REALLOC(tx
->regs
.r
,
648 k
* sizeof(tx
->regs
.r
[0]),
649 n
* sizeof(tx
->regs
.r
[0]));
651 tx
->regs
.r
[k
] = ureg_dst_undef();
654 if (ureg_dst_is_undef(tx
->regs
.r
[idx
]))
655 tx
->regs
.r
[idx
] = ureg_DECL_temporary(tx
->ureg
);
659 tx_addr_alloc(struct shader_translator
*tx
, INT idx
)
662 if (ureg_dst_is_undef(tx
->regs
.a
))
663 tx
->regs
.a
= ureg_DECL_address(tx
->ureg
);
667 tx_pred_alloc(struct shader_translator
*tx
, INT idx
)
670 if (ureg_dst_is_undef(tx
->regs
.p
))
671 tx
->regs
.p
= ureg_DECL_predicate(tx
->ureg
);
675 tx_texcoord_alloc(struct shader_translator
*tx
, INT idx
)
678 assert(idx
>= 0 && idx
< Elements(tx
->regs
.vT
));
679 if (ureg_src_is_undef(tx
->regs
.vT
[idx
]))
680 tx
->regs
.vT
[idx
] = ureg_DECL_fs_input(tx
->ureg
, tx
->texcoord_sn
, idx
,
681 TGSI_INTERPOLATE_PERSPECTIVE
);
684 static INLINE
unsigned *
685 tx_bgnloop(struct shader_translator
*tx
)
688 if (tx
->loop_depth_max
< tx
->loop_depth
)
689 tx
->loop_depth_max
= tx
->loop_depth
;
690 assert(tx
->loop_depth
< NINE_MAX_LOOP_DEPTH
);
691 return &tx
->loop_labels
[tx
->loop_depth
- 1];
694 static INLINE
unsigned *
695 tx_endloop(struct shader_translator
*tx
)
697 assert(tx
->loop_depth
);
699 ureg_fixup_label(tx
->ureg
, tx
->loop_labels
[tx
->loop_depth
],
700 ureg_get_instruction_number(tx
->ureg
));
701 return &tx
->loop_labels
[tx
->loop_depth
];
704 static struct ureg_dst
705 tx_get_loopctr(struct shader_translator
*tx
)
707 const unsigned l
= tx
->loop_depth
- 1;
711 DBG("loop counter requested outside of loop\n");
712 return ureg_dst_undef();
715 if (ureg_dst_is_undef(tx
->regs
.aL
[l
]))
717 struct ureg_dst rreg
= ureg_DECL_local_temporary(tx
->ureg
);
718 struct ureg_dst areg
= ureg_DECL_address(tx
->ureg
);
722 for (c
= l
; c
< (l
+ 4) && c
< Elements(tx
->regs
.aL
); ++c
) {
723 tx
->regs
.rL
[c
] = ureg_writemask(rreg
, 1 << (c
& 3));
724 tx
->regs
.aL
[c
] = ureg_writemask(areg
, 1 << (c
& 3));
727 return tx
->regs
.rL
[l
];
729 static struct ureg_dst
730 tx_get_aL(struct shader_translator
*tx
)
732 if (!ureg_dst_is_undef(tx_get_loopctr(tx
)))
733 return tx
->regs
.aL
[tx
->loop_depth
- 1];
734 return ureg_dst_undef();
737 static INLINE
unsigned *
738 tx_cond(struct shader_translator
*tx
)
740 assert(tx
->cond_depth
<= NINE_MAX_COND_DEPTH
);
742 return &tx
->cond_labels
[tx
->cond_depth
- 1];
745 static INLINE
unsigned *
746 tx_elsecond(struct shader_translator
*tx
)
748 assert(tx
->cond_depth
);
749 return &tx
->cond_labels
[tx
->cond_depth
- 1];
753 tx_endcond(struct shader_translator
*tx
)
755 assert(tx
->cond_depth
);
757 ureg_fixup_label(tx
->ureg
, tx
->cond_labels
[tx
->cond_depth
],
758 ureg_get_instruction_number(tx
->ureg
));
761 static INLINE
struct ureg_dst
762 nine_ureg_dst_register(unsigned file
, int index
)
764 return ureg_dst(ureg_src_register(file
, index
));
767 static struct ureg_src
768 tx_src_param(struct shader_translator
*tx
, const struct sm1_src_param
*param
)
770 struct ureg_program
*ureg
= tx
->ureg
;
778 tx_temp_alloc(tx
, param
->idx
);
779 src
= ureg_src(tx
->regs
.r
[param
->idx
]);
781 /* case D3DSPR_TEXTURE: == D3DSPR_ADDR */
785 tx_addr_alloc(tx
, param
->idx
);
786 src
= ureg_src(tx
->regs
.a
);
788 if (tx
->version
.major
< 2 && tx
->version
.minor
< 4) {
789 /* no subroutines, so should be defined */
790 src
= ureg_src(tx
->regs
.tS
[param
->idx
]);
792 tx_texcoord_alloc(tx
, param
->idx
);
793 src
= tx
->regs
.vT
[param
->idx
];
799 src
= ureg_src_register(TGSI_FILE_INPUT
, param
->idx
);
801 if (tx
->version
.major
< 3) {
803 src
= ureg_DECL_fs_input(tx
->ureg
, TGSI_SEMANTIC_COLOR
,
805 TGSI_INTERPOLATE_PERSPECTIVE
);
807 assert(!param
->rel
); /* TODO */
808 assert(param
->idx
< Elements(tx
->regs
.v
));
809 src
= tx
->regs
.v
[param
->idx
];
813 case D3DSPR_PREDICATE
:
815 tx_pred_alloc(tx
, param
->idx
);
816 src
= ureg_src(tx
->regs
.p
);
819 assert(param
->mod
== NINED3DSPSM_NONE
);
820 assert(param
->swizzle
== NINED3DSP_NOSWIZZLE
);
822 src
= ureg_src_register(TGSI_FILE_SAMPLER
, param
->idx
);
826 tx
->indirect_const_access
= TRUE
;
827 if (param
->rel
|| !tx_lconstf(tx
, &src
, param
->idx
)) {
829 nine_info_mark_const_f_used(tx
->info
, param
->idx
);
830 src
= ureg_src_register(TGSI_FILE_CONSTANT
, param
->idx
);
836 DBG("CONST2/3/4 should have been collapsed into D3DSPR_CONST !\n");
837 assert(!"CONST2/3/4");
838 src
= ureg_imm1f(ureg
, 0.0f
);
840 case D3DSPR_CONSTINT
:
841 if (param
->rel
|| !tx_lconsti(tx
, &src
, param
->idx
)) {
843 nine_info_mark_const_i_used(tx
->info
, param
->idx
);
844 src
= ureg_src_register(TGSI_FILE_CONSTANT
,
845 tx
->info
->const_i_base
+ param
->idx
);
848 case D3DSPR_CONSTBOOL
:
849 if (param
->rel
|| !tx_lconstb(tx
, &src
, param
->idx
)) {
850 char r
= param
->idx
/ 4;
851 char s
= param
->idx
& 3;
853 nine_info_mark_const_b_used(tx
->info
, param
->idx
);
854 src
= ureg_src_register(TGSI_FILE_CONSTANT
,
855 tx
->info
->const_b_base
+ r
);
856 src
= ureg_swizzle(src
, s
, s
, s
, s
);
860 src
= tx_src_scalar(tx_get_aL(tx
));
862 case D3DSPR_MISCTYPE
:
863 switch (param
->idx
) {
864 case D3DSMO_POSITION
:
865 if (ureg_src_is_undef(tx
->regs
.vPos
))
866 tx
->regs
.vPos
= ureg_DECL_fs_input(ureg
,
867 TGSI_SEMANTIC_POSITION
, 0,
868 TGSI_INTERPOLATE_LINEAR
);
869 if (tx
->shift_wpos
) {
870 /* TODO: do this only once */
871 struct ureg_dst wpos
= tx_scratch(tx
);
872 ureg_SUB(ureg
, wpos
, tx
->regs
.vPos
,
873 ureg_imm4f(ureg
, 0.5f
, 0.5f
, 0.0f
, 0.0f
));
874 src
= ureg_src(wpos
);
880 if (ureg_src_is_undef(tx
->regs
.vFace
)) {
881 tx
->regs
.vFace
= ureg_DECL_fs_input(ureg
,
882 TGSI_SEMANTIC_FACE
, 0,
883 TGSI_INTERPOLATE_CONSTANT
);
884 tx
->regs
.vFace
= ureg_scalar(tx
->regs
.vFace
, TGSI_SWIZZLE_X
);
886 src
= tx
->regs
.vFace
;
889 assert(!"invalid src D3DSMO");
894 case D3DSPR_TEMPFLOAT16
:
897 assert(!"invalid src D3DSPR");
900 src
= ureg_src_indirect(src
, tx_src_param(tx
, param
->rel
));
902 if (param
->swizzle
!= NINED3DSP_NOSWIZZLE
)
903 src
= ureg_swizzle(src
,
904 (param
->swizzle
>> 0) & 0x3,
905 (param
->swizzle
>> 2) & 0x3,
906 (param
->swizzle
>> 4) & 0x3,
907 (param
->swizzle
>> 6) & 0x3);
909 switch (param
->mod
) {
910 case NINED3DSPSM_ABS
:
913 case NINED3DSPSM_ABSNEG
:
914 src
= ureg_negate(ureg_abs(src
));
916 case NINED3DSPSM_NEG
:
917 src
= ureg_negate(src
);
919 case NINED3DSPSM_BIAS
:
920 tmp
= tx_scratch(tx
);
921 ureg_SUB(ureg
, tmp
, src
, ureg_imm1f(ureg
, 0.5f
));
924 case NINED3DSPSM_BIASNEG
:
925 tmp
= tx_scratch(tx
);
926 ureg_SUB(ureg
, tmp
, ureg_imm1f(ureg
, 0.5f
), src
);
929 case NINED3DSPSM_NOT
:
930 if (tx
->native_integers
) {
931 tmp
= tx_scratch(tx
);
932 ureg_NOT(ureg
, tmp
, src
);
937 case NINED3DSPSM_COMP
:
938 tmp
= tx_scratch(tx
);
939 ureg_SUB(ureg
, tmp
, ureg_imm1f(ureg
, 1.0f
), src
);
944 /* handled in instruction */
946 case NINED3DSPSM_SIGN
:
947 tmp
= tx_scratch(tx
);
948 ureg_MAD(ureg
, tmp
, src
, ureg_imm1f(ureg
, 2.0f
), ureg_imm1f(ureg
, -1.0f
));
951 case NINED3DSPSM_SIGNNEG
:
952 tmp
= tx_scratch(tx
);
953 ureg_MAD(ureg
, tmp
, src
, ureg_imm1f(ureg
, -2.0f
), ureg_imm1f(ureg
, 1.0f
));
957 tmp
= tx_scratch(tx
);
958 ureg_ADD(ureg
, tmp
, src
, src
);
961 case NINED3DSPSM_X2NEG
:
962 tmp
= tx_scratch(tx
);
963 ureg_ADD(ureg
, tmp
, src
, src
);
964 src
= ureg_negate(ureg_src(tmp
));
967 assert(param
->mod
== NINED3DSPSM_NONE
);
974 static struct ureg_dst
975 _tx_dst_param(struct shader_translator
*tx
, const struct sm1_dst_param
*param
)
983 tx_temp_alloc(tx
, param
->idx
);
984 dst
= tx
->regs
.r
[param
->idx
];
986 /* case D3DSPR_TEXTURE: == D3DSPR_ADDR */
989 if (tx
->version
.major
< 2 && !IS_VS
) {
990 if (ureg_dst_is_undef(tx
->regs
.tS
[param
->idx
]))
991 tx
->regs
.tS
[param
->idx
] = ureg_DECL_temporary(tx
->ureg
);
992 dst
= tx
->regs
.tS
[param
->idx
];
994 if (!IS_VS
&& tx
->insn
.opcode
== D3DSIO_TEXKILL
) { /* maybe others, too */
995 tx_texcoord_alloc(tx
, param
->idx
);
996 dst
= ureg_dst(tx
->regs
.vT
[param
->idx
]);
998 tx_addr_alloc(tx
, param
->idx
);
1002 case D3DSPR_RASTOUT
:
1003 assert(!param
->rel
);
1004 switch (param
->idx
) {
1006 if (ureg_dst_is_undef(tx
->regs
.oPos
))
1008 ureg_DECL_output(tx
->ureg
, TGSI_SEMANTIC_POSITION
, 0);
1009 dst
= tx
->regs
.oPos
;
1012 if (ureg_dst_is_undef(tx
->regs
.oFog
))
1014 ureg_saturate(ureg_DECL_output(tx
->ureg
, TGSI_SEMANTIC_FOG
, 0));
1015 dst
= tx
->regs
.oFog
;
1018 if (ureg_dst_is_undef(tx
->regs
.oPts
))
1020 ureg_saturate(ureg_DECL_output(tx
->ureg
, TGSI_SEMANTIC_PSIZE
, 0));
1021 dst
= tx
->regs
.oPts
;
1028 /* case D3DSPR_TEXCRDOUT: == D3DSPR_OUTPUT */
1030 if (tx
->version
.major
< 3) {
1031 assert(!param
->rel
);
1032 dst
= ureg_DECL_output(tx
->ureg
, tx
->texcoord_sn
, param
->idx
);
1034 assert(!param
->rel
); /* TODO */
1035 assert(param
->idx
< Elements(tx
->regs
.o
));
1036 dst
= tx
->regs
.o
[param
->idx
];
1039 case D3DSPR_ATTROUT
: /* VS */
1040 case D3DSPR_COLOROUT
: /* PS */
1041 assert(param
->idx
>= 0 && param
->idx
< 4);
1042 assert(!param
->rel
);
1043 tx
->info
->rt_mask
|= 1 << param
->idx
;
1044 if (ureg_dst_is_undef(tx
->regs
.oCol
[param
->idx
]))
1045 tx
->regs
.oCol
[param
->idx
] =
1046 ureg_DECL_output(tx
->ureg
, TGSI_SEMANTIC_COLOR
, param
->idx
);
1047 dst
= tx
->regs
.oCol
[param
->idx
];
1048 if (IS_VS
&& tx
->version
.major
< 3)
1049 dst
= ureg_saturate(dst
);
1051 case D3DSPR_DEPTHOUT
:
1052 assert(!param
->rel
);
1053 if (ureg_dst_is_undef(tx
->regs
.oDepth
))
1055 ureg_DECL_output_masked(tx
->ureg
, TGSI_SEMANTIC_POSITION
, 0,
1057 dst
= tx
->regs
.oDepth
; /* XXX: must write .z component */
1059 case D3DSPR_PREDICATE
:
1060 assert(!param
->rel
);
1061 tx_pred_alloc(tx
, param
->idx
);
1064 case D3DSPR_TEMPFLOAT16
:
1065 DBG("unhandled D3DSPR: %u\n", param
->file
);
1068 assert(!"invalid dst D3DSPR");
1072 dst
= ureg_dst_indirect(dst
, tx_src_param(tx
, param
->rel
));
1074 if (param
->mask
!= NINED3DSP_WRITEMASK_ALL
)
1075 dst
= ureg_writemask(dst
, param
->mask
);
1076 if (param
->mod
& NINED3DSPDM_SATURATE
)
1077 dst
= ureg_saturate(dst
);
1082 static struct ureg_dst
1083 tx_dst_param(struct shader_translator
*tx
, const struct sm1_dst_param
*param
)
1086 tx
->regs
.tdst
= ureg_writemask(tx_scratch(tx
), param
->mask
);
1087 return tx
->regs
.tdst
;
1089 return _tx_dst_param(tx
, param
);
1093 tx_apply_dst0_modifiers(struct shader_translator
*tx
)
1095 struct ureg_dst rdst
;
1098 if (!tx
->insn
.ndst
|| !tx
->insn
.dst
[0].shift
|| tx
->insn
.opcode
== D3DSIO_TEXKILL
)
1100 rdst
= _tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1102 assert(rdst
.File
!= TGSI_FILE_ADDRESS
); /* this probably isn't possible */
1104 if (tx
->insn
.dst
[0].shift
< 0)
1105 f
= 1.0f
/ (1 << -tx
->insn
.dst
[0].shift
);
1107 f
= 1 << tx
->insn
.dst
[0].shift
;
1109 ureg_MUL(tx
->ureg
, rdst
, ureg_src(tx
->regs
.tdst
), ureg_imm1f(tx
->ureg
, f
));
1112 static struct ureg_src
1113 tx_dst_param_as_src(struct shader_translator
*tx
, const struct sm1_dst_param
*param
)
1115 struct ureg_src src
;
1117 assert(!param
->shift
);
1118 assert(!(param
->mod
& NINED3DSPDM_SATURATE
));
1120 switch (param
->file
) {
1123 src
= ureg_src_register(TGSI_FILE_INPUT
, param
->idx
);
1125 assert(!param
->rel
);
1126 assert(param
->idx
< Elements(tx
->regs
.v
));
1127 src
= tx
->regs
.v
[param
->idx
];
1131 src
= ureg_src(tx_dst_param(tx
, param
));
1135 src
= ureg_src_indirect(src
, tx_src_param(tx
, param
->rel
));
1138 WARN("mask is 0, using identity swizzle\n");
1140 if (param
->mask
&& param
->mask
!= NINED3DSP_WRITEMASK_ALL
) {
1144 for (n
= 0, c
= 0; c
< 4; ++c
)
1145 if (param
->mask
& (1 << c
))
1148 for (c
= n
; c
< 4; ++c
)
1150 src
= ureg_swizzle(src
, s
[0], s
[1], s
[2], s
[3]);
1156 NineTranslateInstruction_Mkxn(struct shader_translator
*tx
, const unsigned k
, const unsigned n
)
1158 struct ureg_program
*ureg
= tx
->ureg
;
1159 struct ureg_dst dst
;
1160 struct ureg_src src
[2];
1161 struct sm1_src_param
*src_mat
= &tx
->insn
.src
[1];
1164 dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1165 src
[0] = tx_src_param(tx
, &tx
->insn
.src
[0]);
1167 for (i
= 0; i
< n
; i
++)
1169 const unsigned m
= (1 << i
);
1171 src
[1] = tx_src_param(tx
, src_mat
);
1174 if (!(dst
.WriteMask
& m
))
1177 /* XXX: src == dst case ? */
1181 ureg_DP3(ureg
, ureg_writemask(dst
, m
), src
[0], src
[1]);
1184 ureg_DP4(ureg
, ureg_writemask(dst
, m
), src
[0], src
[1]);
1187 DBG("invalid operation: M%ux%u\n", m
, n
);
1195 #define VNOTSUPPORTED 0, 0
1196 #define V(maj, min) (((maj) << 8) | (min))
1198 static INLINE
const char *
1199 d3dsio_to_string( unsigned opcode
)
1201 static const char *names
[] = {
1301 if (opcode
< Elements(names
)) return names
[opcode
];
1304 case D3DSIO_PHASE
: return "PHASE";
1305 case D3DSIO_COMMENT
: return "COMMENT";
1306 case D3DSIO_END
: return "END";
1312 #define NULL_INSTRUCTION { 0, { 0, 0 }, { 0, 0 }, 0, 0, NULL }
1313 #define IS_VALID_INSTRUCTION(inst) ((inst).vert_version.min | \
1314 (inst).vert_version.max | \
1315 (inst).frag_version.min | \
1316 (inst).frag_version.max)
1318 #define SPECIAL(name) \
1319 NineTranslateInstruction_##name
1321 #define DECL_SPECIAL(name) \
1323 NineTranslateInstruction_##name( struct shader_translator *tx )
1326 NineTranslateInstruction_Generic(struct shader_translator
*);
1330 return NineTranslateInstruction_Mkxn(tx
, 4, 4);
1335 return NineTranslateInstruction_Mkxn(tx
, 4, 3);
1340 return NineTranslateInstruction_Mkxn(tx
, 3, 4);
1345 return NineTranslateInstruction_Mkxn(tx
, 3, 3);
1350 return NineTranslateInstruction_Mkxn(tx
, 3, 2);
1355 ureg_CMP(tx
->ureg
, tx_dst_param(tx
, &tx
->insn
.dst
[0]),
1356 tx_src_param(tx
, &tx
->insn
.src
[0]),
1357 tx_src_param(tx
, &tx
->insn
.src
[2]),
1358 tx_src_param(tx
, &tx
->insn
.src
[1]));
1364 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1365 struct ureg_dst cgt
;
1366 struct ureg_src cnd
;
1368 if (tx
->insn
.coissue
&& tx
->version
.major
== 1 && tx
->version
.minor
< 4) {
1370 dst
, tx_src_param(tx
, &tx
->insn
.src
[1]));
1374 cnd
= tx_src_param(tx
, &tx
->insn
.src
[0]);
1375 cgt
= tx_scratch(tx
);
1377 if (tx
->version
.major
== 1 && tx
->version
.minor
< 4) {
1378 cgt
.WriteMask
= TGSI_WRITEMASK_W
;
1379 ureg_SGT(tx
->ureg
, cgt
, cnd
, ureg_imm1f(tx
->ureg
, 0.5f
));
1380 cnd
= ureg_scalar(cnd
, TGSI_SWIZZLE_W
);
1382 ureg_SGT(tx
->ureg
, cgt
, cnd
, ureg_imm1f(tx
->ureg
, 0.5f
));
1384 ureg_CMP(tx
->ureg
, dst
,
1385 tx_src_param(tx
, &tx
->insn
.src
[1]),
1386 tx_src_param(tx
, &tx
->insn
.src
[2]), ureg_negate(cnd
));
1392 assert(tx
->insn
.src
[0].idx
< tx
->num_inst_labels
);
1393 ureg_CAL(tx
->ureg
, &tx
->inst_labels
[tx
->insn
.src
[0].idx
]);
1397 DECL_SPECIAL(CALLNZ
)
1399 struct ureg_program
*ureg
= tx
->ureg
;
1400 struct ureg_dst tmp
= tx_scratch_scalar(tx
);
1401 struct ureg_src src
= tx_src_param(tx
, &tx
->insn
.src
[1]);
1403 /* NOTE: source should be const bool, so we can use NOT/SUB instead of [U]SNE 0 */
1404 if (!tx
->insn
.flags
) {
1405 if (tx
->native_integers
)
1406 ureg_NOT(ureg
, tmp
, src
);
1408 ureg_SUB(ureg
, tmp
, ureg_imm1f(ureg
, 1.0f
), src
);
1410 ureg_IF(ureg
, tx
->insn
.flags
? src
: tx_src_scalar(tmp
), tx_cond(tx
));
1411 ureg_CAL(ureg
, &tx
->inst_labels
[tx
->insn
.src
[0].idx
]);
1417 DECL_SPECIAL(MOV_vs1x
)
1419 if (tx
->insn
.dst
[0].file
== D3DSPR_ADDR
) {
1421 tx_dst_param(tx
, &tx
->insn
.dst
[0]),
1422 tx_src_param(tx
, &tx
->insn
.src
[0]));
1425 return NineTranslateInstruction_Generic(tx
);
1430 struct ureg_program
*ureg
= tx
->ureg
;
1432 struct ureg_src src
= tx_src_param(tx
, &tx
->insn
.src
[1]);
1433 struct ureg_src iter
= ureg_scalar(src
, TGSI_SWIZZLE_X
);
1434 struct ureg_src init
= ureg_scalar(src
, TGSI_SWIZZLE_Y
);
1435 struct ureg_src step
= ureg_scalar(src
, TGSI_SWIZZLE_Z
);
1436 struct ureg_dst ctr
;
1437 struct ureg_dst tmp
= tx_scratch_scalar(tx
);
1439 label
= tx_bgnloop(tx
);
1440 ctr
= tx_get_loopctr(tx
);
1442 ureg_MOV(tx
->ureg
, ctr
, init
);
1443 ureg_BGNLOOP(tx
->ureg
, label
);
1444 if (tx
->native_integers
) {
1445 /* we'll let the backend pull up that MAD ... */
1446 ureg_UMAD(ureg
, tmp
, iter
, step
, init
);
1447 ureg_USEQ(ureg
, tmp
, ureg_src(ctr
), tx_src_scalar(tmp
));
1448 #ifdef NINE_TGSI_LAZY_DEVS
1449 ureg_UIF(ureg
, tx_src_scalar(tmp
), tx_cond(tx
));
1452 /* can't simply use SGE for precision because step might be negative */
1453 ureg_MAD(ureg
, tmp
, iter
, step
, init
);
1454 ureg_SEQ(ureg
, tmp
, ureg_src(ctr
), tx_src_scalar(tmp
));
1455 #ifdef NINE_TGSI_LAZY_DEVS
1456 ureg_IF(ureg
, tx_src_scalar(tmp
), tx_cond(tx
));
1459 #ifdef NINE_TGSI_LAZY_DEVS
1464 ureg_BREAKC(ureg
, tx_src_scalar(tmp
));
1466 if (tx
->native_integers
) {
1467 ureg_UARL(ureg
, tx_get_aL(tx
), tx_src_scalar(ctr
));
1468 ureg_UADD(ureg
, ctr
, tx_src_scalar(ctr
), step
);
1470 ureg_ARL(ureg
, tx_get_aL(tx
), tx_src_scalar(ctr
));
1471 ureg_ADD(ureg
, ctr
, tx_src_scalar(ctr
), step
);
1482 DECL_SPECIAL(ENDLOOP
)
1484 ureg_ENDLOOP(tx
->ureg
, tx_endloop(tx
));
1490 unsigned k
= tx
->num_inst_labels
;
1491 unsigned n
= tx
->insn
.src
[0].idx
;
1494 tx
->inst_labels
= REALLOC(tx
->inst_labels
,
1495 k
* sizeof(tx
->inst_labels
[0]),
1496 n
* sizeof(tx
->inst_labels
[0]));
1498 tx
->inst_labels
[n
] = ureg_get_instruction_number(tx
->ureg
);
1502 DECL_SPECIAL(SINCOS
)
1504 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1505 struct ureg_src src
= tx_src_param(tx
, &tx
->insn
.src
[0]);
1507 assert(!(dst
.WriteMask
& 0xc));
1509 dst
.WriteMask
&= TGSI_WRITEMASK_XY
; /* z undefined, w untouched */
1510 ureg_SCS(tx
->ureg
, dst
, src
);
1517 tx_dst_param(tx
, &tx
->insn
.dst
[0]),
1518 tx_src_param(tx
, &tx
->insn
.src
[0]));
1524 struct ureg_program
*ureg
= tx
->ureg
;
1526 struct ureg_src rep
= tx_src_param(tx
, &tx
->insn
.src
[0]);
1527 struct ureg_dst ctr
;
1528 struct ureg_dst tmp
= tx_scratch_scalar(tx
);
1529 struct ureg_src imm
=
1530 tx
->native_integers
? ureg_imm1u(ureg
, 0) : ureg_imm1f(ureg
, 0.0f
);
1532 label
= tx_bgnloop(tx
);
1533 ctr
= tx_get_loopctr(tx
);
1535 /* NOTE: rep must be constant, so we don't have to save the count */
1536 assert(rep
.File
== TGSI_FILE_CONSTANT
|| rep
.File
== TGSI_FILE_IMMEDIATE
);
1538 ureg_MOV(ureg
, ctr
, imm
);
1539 ureg_BGNLOOP(ureg
, label
);
1540 if (tx
->native_integers
)
1542 ureg_USGE(ureg
, tmp
, tx_src_scalar(ctr
), rep
);
1543 ureg_UIF(ureg
, tx_src_scalar(tmp
), tx_cond(tx
));
1547 ureg_SGE(ureg
, tmp
, tx_src_scalar(ctr
), rep
);
1548 ureg_IF(ureg
, tx_src_scalar(tmp
), tx_cond(tx
));
1554 if (tx
->native_integers
) {
1555 ureg_UADD(ureg
, ctr
, tx_src_scalar(ctr
), ureg_imm1u(ureg
, 1));
1557 ureg_ADD(ureg
, ctr
, tx_src_scalar(ctr
), ureg_imm1f(ureg
, 1.0f
));
1563 DECL_SPECIAL(ENDREP
)
1565 ureg_ENDLOOP(tx
->ureg
, tx_endloop(tx
));
1572 ureg_ENDIF(tx
->ureg
);
1578 struct ureg_src src
= tx_src_param(tx
, &tx
->insn
.src
[0]);
1580 if (tx
->native_integers
&& tx
->insn
.src
[0].file
== D3DSPR_CONSTBOOL
)
1581 ureg_UIF(tx
->ureg
, src
, tx_cond(tx
));
1583 ureg_IF(tx
->ureg
, src
, tx_cond(tx
));
1588 static INLINE
unsigned
1589 sm1_insn_flags_to_tgsi_setop(BYTE flags
)
1592 case NINED3DSHADER_REL_OP_GT
: return TGSI_OPCODE_SGT
;
1593 case NINED3DSHADER_REL_OP_EQ
: return TGSI_OPCODE_SEQ
;
1594 case NINED3DSHADER_REL_OP_GE
: return TGSI_OPCODE_SGE
;
1595 case NINED3DSHADER_REL_OP_LT
: return TGSI_OPCODE_SLT
;
1596 case NINED3DSHADER_REL_OP_NE
: return TGSI_OPCODE_SNE
;
1597 case NINED3DSHADER_REL_OP_LE
: return TGSI_OPCODE_SLE
;
1599 assert(!"invalid comparison flags");
1600 return TGSI_OPCODE_SGT
;
1606 const unsigned cmp_op
= sm1_insn_flags_to_tgsi_setop(tx
->insn
.flags
);
1607 struct ureg_src src
[2];
1608 struct ureg_dst tmp
= ureg_writemask(tx_scratch(tx
), TGSI_WRITEMASK_X
);
1609 src
[0] = tx_src_param(tx
, &tx
->insn
.src
[0]);
1610 src
[1] = tx_src_param(tx
, &tx
->insn
.src
[1]);
1611 ureg_insn(tx
->ureg
, cmp_op
, &tmp
, 1, src
, 2);
1612 ureg_IF(tx
->ureg
, ureg_scalar(ureg_src(tmp
), TGSI_SWIZZLE_X
), tx_cond(tx
));
1618 ureg_ELSE(tx
->ureg
, tx_elsecond(tx
));
1622 DECL_SPECIAL(BREAKC
)
1624 const unsigned cmp_op
= sm1_insn_flags_to_tgsi_setop(tx
->insn
.flags
);
1625 struct ureg_src src
[2];
1626 struct ureg_dst tmp
= ureg_writemask(tx_scratch(tx
), TGSI_WRITEMASK_X
);
1627 src
[0] = tx_src_param(tx
, &tx
->insn
.src
[0]);
1628 src
[1] = tx_src_param(tx
, &tx
->insn
.src
[1]);
1629 ureg_insn(tx
->ureg
, cmp_op
, &tmp
, 1, src
, 2);
1630 ureg_IF(tx
->ureg
, ureg_scalar(ureg_src(tmp
), TGSI_SWIZZLE_X
), tx_cond(tx
));
1633 ureg_ENDIF(tx
->ureg
);
1637 static const char *sm1_declusage_names
[] =
1639 [D3DDECLUSAGE_POSITION
] = "POSITION",
1640 [D3DDECLUSAGE_BLENDWEIGHT
] = "BLENDWEIGHT",
1641 [D3DDECLUSAGE_BLENDINDICES
] = "BLENDINDICES",
1642 [D3DDECLUSAGE_NORMAL
] = "NORMAL",
1643 [D3DDECLUSAGE_PSIZE
] = "PSIZE",
1644 [D3DDECLUSAGE_TEXCOORD
] = "TEXCOORD",
1645 [D3DDECLUSAGE_TANGENT
] = "TANGENT",
1646 [D3DDECLUSAGE_BINORMAL
] = "BINORMAL",
1647 [D3DDECLUSAGE_TESSFACTOR
] = "TESSFACTOR",
1648 [D3DDECLUSAGE_POSITIONT
] = "POSITIONT",
1649 [D3DDECLUSAGE_COLOR
] = "COLOR",
1650 [D3DDECLUSAGE_FOG
] = "FOG",
1651 [D3DDECLUSAGE_DEPTH
] = "DEPTH",
1652 [D3DDECLUSAGE_SAMPLE
] = "SAMPLE"
1655 static INLINE
unsigned
1656 sm1_to_nine_declusage(struct sm1_semantic
*dcl
)
1658 return nine_d3d9_to_nine_declusage(dcl
->usage
, dcl
->usage_idx
);
1662 sm1_declusage_to_tgsi(struct tgsi_declaration_semantic
*sem
,
1664 struct sm1_semantic
*dcl
)
1666 BYTE index
= dcl
->usage_idx
;
1668 /* For everything that is not matching to a TGSI_SEMANTIC_****,
1669 * we match to a TGSI_SEMANTIC_GENERIC with index.
1671 * The index can be anything UINT16 and usage_idx is BYTE,
1672 * so we can fit everything. It doesn't matter if indices
1673 * are close together or low.
1676 * POSITION >= 1: 10 * index + 6
1677 * COLOR >= 2: 10 * (index-1) + 7
1678 * TEXCOORD[0..15]: index
1679 * BLENDWEIGHT: 10 * index + 18
1680 * BLENDINDICES: 10 * index + 19
1681 * NORMAL: 10 * index + 20
1682 * TANGENT: 10 * index + 21
1683 * BINORMAL: 10 * index + 22
1684 * TESSFACTOR: 10 * index + 23
1687 switch (dcl
->usage
) {
1688 case D3DDECLUSAGE_POSITION
:
1689 case D3DDECLUSAGE_POSITIONT
:
1690 case D3DDECLUSAGE_DEPTH
:
1692 sem
->Name
= TGSI_SEMANTIC_POSITION
;
1695 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1696 sem
->Index
= 10 * index
+ 6;
1699 case D3DDECLUSAGE_COLOR
:
1701 sem
->Name
= TGSI_SEMANTIC_COLOR
;
1704 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1705 sem
->Index
= 10 * (index
-1) + 7;
1708 case D3DDECLUSAGE_FOG
:
1710 sem
->Name
= TGSI_SEMANTIC_FOG
;
1713 case D3DDECLUSAGE_PSIZE
:
1715 sem
->Name
= TGSI_SEMANTIC_PSIZE
;
1718 case D3DDECLUSAGE_TEXCOORD
:
1720 if (index
< 8 && tc
)
1721 sem
->Name
= TGSI_SEMANTIC_TEXCOORD
;
1723 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1726 case D3DDECLUSAGE_BLENDWEIGHT
:
1727 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1728 sem
->Index
= 10 * index
+ 18;
1730 case D3DDECLUSAGE_BLENDINDICES
:
1731 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1732 sem
->Index
= 10 * index
+ 19;
1734 case D3DDECLUSAGE_NORMAL
:
1735 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1736 sem
->Index
= 10 * index
+ 20;
1738 case D3DDECLUSAGE_TANGENT
:
1739 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1740 sem
->Index
= 10 * index
+ 21;
1742 case D3DDECLUSAGE_BINORMAL
:
1743 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1744 sem
->Index
= 10 * index
+ 22;
1746 case D3DDECLUSAGE_TESSFACTOR
:
1747 sem
->Name
= TGSI_SEMANTIC_GENERIC
;
1748 sem
->Index
= 10 * index
+ 23;
1750 case D3DDECLUSAGE_SAMPLE
:
1751 sem
->Name
= TGSI_SEMANTIC_COUNT
;
1755 assert(!"Invalid DECLUSAGE.");
1760 #define NINED3DSTT_1D (D3DSTT_1D >> D3DSP_TEXTURETYPE_SHIFT)
1761 #define NINED3DSTT_2D (D3DSTT_2D >> D3DSP_TEXTURETYPE_SHIFT)
1762 #define NINED3DSTT_VOLUME (D3DSTT_VOLUME >> D3DSP_TEXTURETYPE_SHIFT)
1763 #define NINED3DSTT_CUBE (D3DSTT_CUBE >> D3DSP_TEXTURETYPE_SHIFT)
1764 static INLINE
unsigned
1765 d3dstt_to_tgsi_tex(BYTE sampler_type
)
1767 switch (sampler_type
) {
1768 case NINED3DSTT_1D
: return TGSI_TEXTURE_1D
;
1769 case NINED3DSTT_2D
: return TGSI_TEXTURE_2D
;
1770 case NINED3DSTT_VOLUME
: return TGSI_TEXTURE_3D
;
1771 case NINED3DSTT_CUBE
: return TGSI_TEXTURE_CUBE
;
1774 return TGSI_TEXTURE_UNKNOWN
;
1777 static INLINE
unsigned
1778 d3dstt_to_tgsi_tex_shadow(BYTE sampler_type
)
1780 switch (sampler_type
) {
1781 case NINED3DSTT_1D
: return TGSI_TEXTURE_SHADOW1D
;
1782 case NINED3DSTT_2D
: return TGSI_TEXTURE_SHADOW2D
;
1783 case NINED3DSTT_VOLUME
:
1784 case NINED3DSTT_CUBE
:
1787 return TGSI_TEXTURE_UNKNOWN
;
1790 static INLINE
unsigned
1791 ps1x_sampler_type(const struct nine_shader_info
*info
, unsigned stage
)
1793 switch ((info
->sampler_ps1xtypes
>> (stage
* 2)) & 0x3) {
1794 case 1: return TGSI_TEXTURE_1D
;
1795 case 0: return TGSI_TEXTURE_2D
;
1796 case 3: return TGSI_TEXTURE_3D
;
1798 return TGSI_TEXTURE_CUBE
;
1803 sm1_sampler_type_name(BYTE sampler_type
)
1805 switch (sampler_type
) {
1806 case NINED3DSTT_1D
: return "1D";
1807 case NINED3DSTT_2D
: return "2D";
1808 case NINED3DSTT_VOLUME
: return "VOLUME";
1809 case NINED3DSTT_CUBE
: return "CUBE";
1811 return "(D3DSTT_?)";
1815 static INLINE
unsigned
1816 nine_tgsi_to_interp_mode(struct tgsi_declaration_semantic
*sem
)
1818 switch (sem
->Name
) {
1819 case TGSI_SEMANTIC_POSITION
:
1820 case TGSI_SEMANTIC_NORMAL
:
1821 return TGSI_INTERPOLATE_LINEAR
;
1822 case TGSI_SEMANTIC_BCOLOR
:
1823 case TGSI_SEMANTIC_COLOR
:
1824 case TGSI_SEMANTIC_FOG
:
1825 case TGSI_SEMANTIC_GENERIC
:
1826 case TGSI_SEMANTIC_TEXCOORD
:
1827 case TGSI_SEMANTIC_CLIPDIST
:
1828 case TGSI_SEMANTIC_CLIPVERTEX
:
1829 return TGSI_INTERPOLATE_PERSPECTIVE
;
1830 case TGSI_SEMANTIC_EDGEFLAG
:
1831 case TGSI_SEMANTIC_FACE
:
1832 case TGSI_SEMANTIC_INSTANCEID
:
1833 case TGSI_SEMANTIC_PCOORD
:
1834 case TGSI_SEMANTIC_PRIMID
:
1835 case TGSI_SEMANTIC_PSIZE
:
1836 case TGSI_SEMANTIC_VERTEXID
:
1837 return TGSI_INTERPOLATE_CONSTANT
;
1840 return TGSI_INTERPOLATE_CONSTANT
;
1846 struct ureg_program
*ureg
= tx
->ureg
;
1849 struct tgsi_declaration_semantic tgsi
;
1850 struct sm1_semantic sem
;
1851 sm1_read_semantic(tx
, &sem
);
1853 is_input
= sem
.reg
.file
== D3DSPR_INPUT
;
1855 sem
.usage
== D3DDECLUSAGE_SAMPLE
|| sem
.reg
.file
== D3DSPR_SAMPLER
;
1858 sm1_dump_dst_param(&sem
.reg
);
1860 DUMP(" %s\n", sm1_sampler_type_name(sem
.sampler_type
));
1862 if (tx
->version
.major
>= 3)
1863 DUMP(" %s%i\n", sm1_declusage_names
[sem
.usage
], sem
.usage_idx
);
1865 if (sem
.usage
| sem
.usage_idx
)
1866 DUMP(" %u[%u]\n", sem
.usage
, sem
.usage_idx
);
1871 const unsigned m
= 1 << sem
.reg
.idx
;
1872 ureg_DECL_sampler(ureg
, sem
.reg
.idx
);
1873 tx
->info
->sampler_mask
|= m
;
1874 tx
->sampler_targets
[sem
.reg
.idx
] = (tx
->info
->sampler_mask_shadow
& m
) ?
1875 d3dstt_to_tgsi_tex_shadow(sem
.sampler_type
) :
1876 d3dstt_to_tgsi_tex(sem
.sampler_type
);
1880 sm1_declusage_to_tgsi(&tgsi
, tx
->want_texcoord
, &sem
);
1883 /* linkage outside of shader with vertex declaration */
1884 ureg_DECL_vs_input(ureg
, sem
.reg
.idx
);
1885 assert(sem
.reg
.idx
< Elements(tx
->info
->input_map
));
1886 tx
->info
->input_map
[sem
.reg
.idx
] = sm1_to_nine_declusage(&sem
);
1887 tx
->info
->num_inputs
= sem
.reg
.idx
+ 1;
1888 /* NOTE: preserving order in case of indirect access */
1890 if (tx
->version
.major
>= 3) {
1891 /* SM2 output semantic determined by file */
1892 assert(sem
.reg
.mask
!= 0);
1893 if (sem
.usage
== D3DDECLUSAGE_POSITIONT
)
1894 tx
->info
->position_t
= TRUE
;
1895 assert(sem
.reg
.idx
< Elements(tx
->regs
.o
));
1896 tx
->regs
.o
[sem
.reg
.idx
] = ureg_DECL_output_masked(
1897 ureg
, tgsi
.Name
, tgsi
.Index
, sem
.reg
.mask
);
1899 if (tgsi
.Name
== TGSI_SEMANTIC_PSIZE
)
1900 tx
->regs
.oPts
= tx
->regs
.o
[sem
.reg
.idx
];
1903 if (is_input
&& tx
->version
.major
>= 3) {
1904 /* SM3 only, SM2 input semantic determined by file */
1905 assert(sem
.reg
.idx
< Elements(tx
->regs
.v
));
1906 tx
->regs
.v
[sem
.reg
.idx
] = ureg_DECL_fs_input_cyl_centroid(
1907 ureg
, tgsi
.Name
, tgsi
.Index
,
1908 nine_tgsi_to_interp_mode(&tgsi
),
1910 sem
.reg
.mod
& NINED3DSPDM_CENTROID
);
1912 if (!is_input
&& 0) { /* declare in COLOROUT/DEPTHOUT case */
1913 /* FragColor or FragDepth */
1914 assert(sem
.reg
.mask
!= 0);
1915 ureg_DECL_output_masked(ureg
, tgsi
.Name
, tgsi
.Index
, sem
.reg
.mask
);
1923 tx_set_lconstf(tx
, tx
->insn
.dst
[0].idx
, tx
->insn
.src
[0].imm
.f
);
1929 tx_set_lconstb(tx
, tx
->insn
.dst
[0].idx
, tx
->insn
.src
[0].imm
.b
);
1935 tx_set_lconsti(tx
, tx
->insn
.dst
[0].idx
, tx
->insn
.src
[0].imm
.i
);
1941 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1942 struct ureg_src src
[2] = {
1943 tx_src_param(tx
, &tx
->insn
.src
[0]),
1944 tx_src_param(tx
, &tx
->insn
.src
[1])
1946 ureg_POW(tx
->ureg
, dst
, ureg_abs(src
[0]), src
[1]);
1952 struct ureg_program
*ureg
= tx
->ureg
;
1953 struct ureg_dst tmp
= tx_scratch_scalar(tx
);
1954 struct ureg_src nrm
= tx_src_scalar(tmp
);
1955 struct ureg_src src
= tx_src_param(tx
, &tx
->insn
.src
[0]);
1956 ureg_DP3(ureg
, tmp
, src
, src
);
1957 ureg_RSQ(ureg
, tmp
, nrm
);
1958 ureg_MUL(ureg
, tx_dst_param(tx
, &tx
->insn
.dst
[0]), src
, nrm
);
1962 DECL_SPECIAL(DP2ADD
)
1964 struct ureg_dst tmp
= tx_scratch_scalar(tx
);
1965 struct ureg_src dp2
= tx_src_scalar(tmp
);
1966 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1967 struct ureg_src src
[3];
1969 for (i
= 0; i
< 3; ++i
)
1970 src
[i
] = tx_src_param(tx
, &tx
->insn
.src
[i
]);
1971 assert_replicate_swizzle(&src
[2]);
1973 ureg_DP2(tx
->ureg
, tmp
, src
[0], src
[1]);
1974 ureg_ADD(tx
->ureg
, dst
, src
[2], dp2
);
1979 DECL_SPECIAL(TEXCOORD
)
1981 struct ureg_program
*ureg
= tx
->ureg
;
1982 const unsigned s
= tx
->insn
.dst
[0].idx
;
1983 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1985 if (ureg_src_is_undef(tx
->regs
.vT
[s
]))
1986 tx
->regs
.vT
[s
] = ureg_DECL_fs_input(ureg
, tx
->texcoord_sn
, s
, TGSI_INTERPOLATE_PERSPECTIVE
);
1987 ureg_MOV(ureg
, dst
, tx
->regs
.vT
[s
]); /* XXX is this sufficient ? */
1992 DECL_SPECIAL(TEXCOORD_ps14
)
1994 struct ureg_program
*ureg
= tx
->ureg
;
1995 const unsigned s
= tx
->insn
.src
[0].idx
;
1996 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
1998 if (ureg_src_is_undef(tx
->regs
.vT
[s
]))
1999 tx
->regs
.vT
[s
] = ureg_DECL_fs_input(ureg
, tx
->texcoord_sn
, s
, TGSI_INTERPOLATE_PERSPECTIVE
);
2000 ureg_MOV(ureg
, dst
, tx
->regs
.vT
[s
]); /* XXX is this sufficient ? */
2005 DECL_SPECIAL(TEXKILL
)
2007 struct ureg_src reg
;
2009 if (tx
->version
.major
> 1 || tx
->version
.minor
> 3) {
2010 reg
= tx_dst_param_as_src(tx
, &tx
->insn
.dst
[0]);
2012 tx_texcoord_alloc(tx
, tx
->insn
.dst
[0].idx
);
2013 reg
= tx
->regs
.vT
[tx
->insn
.dst
[0].idx
];
2015 if (tx
->version
.major
< 2)
2016 reg
= ureg_swizzle(reg
, NINE_SWIZZLE4(X
,Y
,Z
,Z
));
2017 ureg_KILL_IF(tx
->ureg
, reg
);
2022 DECL_SPECIAL(TEXBEM
)
2024 STUB(D3DERR_INVALIDCALL
);
2027 DECL_SPECIAL(TEXBEML
)
2029 STUB(D3DERR_INVALIDCALL
);
2032 DECL_SPECIAL(TEXREG2AR
)
2034 STUB(D3DERR_INVALIDCALL
);
2037 DECL_SPECIAL(TEXREG2GB
)
2039 STUB(D3DERR_INVALIDCALL
);
2042 DECL_SPECIAL(TEXM3x2PAD
)
2044 STUB(D3DERR_INVALIDCALL
);
2047 DECL_SPECIAL(TEXM3x2TEX
)
2049 STUB(D3DERR_INVALIDCALL
);
2052 DECL_SPECIAL(TEXM3x3PAD
)
2054 return D3D_OK
; /* this is just padding */
2057 DECL_SPECIAL(TEXM3x3SPEC
)
2059 STUB(D3DERR_INVALIDCALL
);
2062 DECL_SPECIAL(TEXM3x3VSPEC
)
2064 STUB(D3DERR_INVALIDCALL
);
2067 DECL_SPECIAL(TEXREG2RGB
)
2069 STUB(D3DERR_INVALIDCALL
);
2072 DECL_SPECIAL(TEXDP3TEX
)
2074 STUB(D3DERR_INVALIDCALL
);
2077 DECL_SPECIAL(TEXM3x2DEPTH
)
2079 STUB(D3DERR_INVALIDCALL
);
2082 DECL_SPECIAL(TEXDP3
)
2084 STUB(D3DERR_INVALIDCALL
);
2087 DECL_SPECIAL(TEXM3x3
)
2089 struct ureg_program
*ureg
= tx
->ureg
;
2090 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
2091 struct ureg_src src
[4];
2093 const int m
= tx
->insn
.dst
[0].idx
- 2;
2094 const int n
= tx
->insn
.src
[0].idx
;
2095 assert(m
>= 0 && m
> n
);
2097 for (s
= m
; s
<= (m
+ 2); ++s
) {
2098 if (ureg_src_is_undef(tx
->regs
.vT
[s
]))
2099 tx
->regs
.vT
[s
] = ureg_DECL_fs_input(ureg
, tx
->texcoord_sn
, s
, TGSI_INTERPOLATE_PERSPECTIVE
);
2100 src
[s
] = tx
->regs
.vT
[s
];
2102 ureg_DP3(ureg
, ureg_writemask(dst
, TGSI_WRITEMASK_X
), src
[0], ureg_src(tx
->regs
.tS
[n
]));
2103 ureg_DP3(ureg
, ureg_writemask(dst
, TGSI_WRITEMASK_Y
), src
[1], ureg_src(tx
->regs
.tS
[n
]));
2104 ureg_DP3(ureg
, ureg_writemask(dst
, TGSI_WRITEMASK_Z
), src
[2], ureg_src(tx
->regs
.tS
[n
]));
2106 switch (tx
->insn
.opcode
) {
2107 case D3DSIO_TEXM3x3
:
2108 ureg_MOV(ureg
, ureg_writemask(dst
, TGSI_WRITEMASK_W
), ureg_imm1f(ureg
, 1.0f
));
2110 case D3DSIO_TEXM3x3TEX
:
2111 src
[3] = ureg_DECL_sampler(ureg
, m
+ 2);
2112 tx
->info
->sampler_mask
|= 1 << (m
+ 2);
2113 ureg_TEX(ureg
, dst
, ps1x_sampler_type(tx
->info
, m
+ 2), ureg_src(dst
), src
[3]);
2116 return D3DERR_INVALIDCALL
;
2121 DECL_SPECIAL(TEXDEPTH
)
2123 STUB(D3DERR_INVALIDCALL
);
2128 STUB(D3DERR_INVALIDCALL
);
2133 struct ureg_program
*ureg
= tx
->ureg
;
2135 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
2136 struct ureg_src src
[2] = {
2137 tx_src_param(tx
, &tx
->insn
.src
[0]),
2138 tx_src_param(tx
, &tx
->insn
.src
[1])
2140 assert(tx
->insn
.src
[1].idx
>= 0 &&
2141 tx
->insn
.src
[1].idx
< Elements(tx
->sampler_targets
));
2142 target
= tx
->sampler_targets
[tx
->insn
.src
[1].idx
];
2144 switch (tx
->insn
.flags
) {
2146 ureg_TEX(ureg
, dst
, target
, src
[0], src
[1]);
2148 case NINED3DSI_TEXLD_PROJECT
:
2149 ureg_TXP(ureg
, dst
, target
, src
[0], src
[1]);
2151 case NINED3DSI_TEXLD_BIAS
:
2152 ureg_TXB(ureg
, dst
, target
, src
[0], src
[1]);
2156 return D3DERR_INVALIDCALL
;
2161 DECL_SPECIAL(TEXLD_14
)
2163 struct ureg_program
*ureg
= tx
->ureg
;
2164 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
2165 struct ureg_src src
= tx_src_param(tx
, &tx
->insn
.src
[0]);
2166 const unsigned s
= tx
->insn
.dst
[0].idx
;
2167 const unsigned t
= ps1x_sampler_type(tx
->info
, s
);
2169 tx
->info
->sampler_mask
|= 1 << s
;
2170 ureg_TEX(ureg
, dst
, t
, src
, ureg_DECL_sampler(ureg
, s
));
2177 struct ureg_program
*ureg
= tx
->ureg
;
2178 const unsigned s
= tx
->insn
.dst
[0].idx
;
2179 const unsigned t
= ps1x_sampler_type(tx
->info
, s
);
2180 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
2181 struct ureg_src src
[2];
2183 if (ureg_src_is_undef(tx
->regs
.vT
[s
]))
2184 tx
->regs
.vT
[s
] = ureg_DECL_fs_input(ureg
, tx
->texcoord_sn
, s
, TGSI_INTERPOLATE_PERSPECTIVE
);
2186 src
[0] = tx
->regs
.vT
[s
];
2187 src
[1] = ureg_DECL_sampler(ureg
, s
);
2188 tx
->info
->sampler_mask
|= 1 << s
;
2190 ureg_TEX(ureg
, dst
, t
, src
[0], src
[1]);
2195 DECL_SPECIAL(TEXLDD
)
2198 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
2199 struct ureg_src src
[4] = {
2200 tx_src_param(tx
, &tx
->insn
.src
[0]),
2201 tx_src_param(tx
, &tx
->insn
.src
[1]),
2202 tx_src_param(tx
, &tx
->insn
.src
[2]),
2203 tx_src_param(tx
, &tx
->insn
.src
[3])
2205 assert(tx
->insn
.src
[3].idx
>= 0 &&
2206 tx
->insn
.src
[3].idx
< Elements(tx
->sampler_targets
));
2207 target
= tx
->sampler_targets
[tx
->insn
.src
[1].idx
];
2209 ureg_TXD(tx
->ureg
, dst
, target
, src
[0], src
[2], src
[3], src
[1]);
2213 DECL_SPECIAL(TEXLDL
)
2216 struct ureg_dst dst
= tx_dst_param(tx
, &tx
->insn
.dst
[0]);
2217 struct ureg_src src
[2] = {
2218 tx_src_param(tx
, &tx
->insn
.src
[0]),
2219 tx_src_param(tx
, &tx
->insn
.src
[1])
2221 assert(tx
->insn
.src
[3].idx
>= 0 &&
2222 tx
->insn
.src
[3].idx
< Elements(tx
->sampler_targets
));
2223 target
= tx
->sampler_targets
[tx
->insn
.src
[1].idx
];
2225 ureg_TXL(tx
->ureg
, dst
, target
, src
[0], src
[1]);
2231 STUB(D3DERR_INVALIDCALL
);
2234 DECL_SPECIAL(BREAKP
)
2236 STUB(D3DERR_INVALIDCALL
);
2241 return D3D_OK
; /* we don't care about phase */
2244 DECL_SPECIAL(COMMENT
)
2246 return D3D_OK
; /* nothing to do */
2250 #define _OPI(o,t,vv1,vv2,pv1,pv2,d,s,h) \
2251 { D3DSIO_##o, TGSI_OPCODE_##t, { vv1, vv2 }, { pv1, pv2, }, d, s, h }
2253 struct sm1_op_info inst_table
[] =
2255 _OPI(NOP
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 0, 0, NULL
), /* 0 */
2256 _OPI(MOV
, MOV
, V(0,0), V(1,1), V(0,0), V(0,0), 1, 1, SPECIAL(MOV_vs1x
)),
2257 _OPI(MOV
, MOV
, V(2,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
),
2258 _OPI(ADD
, ADD
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 2 */
2259 _OPI(SUB
, SUB
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 3 */
2260 _OPI(MAD
, MAD
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 3, NULL
), /* 4 */
2261 _OPI(MUL
, MUL
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 5 */
2262 _OPI(RCP
, RCP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
), /* 6 */
2263 _OPI(RSQ
, RSQ
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
), /* 7 */
2264 _OPI(DP3
, DP3
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 8 */
2265 _OPI(DP4
, DP4
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 9 */
2266 _OPI(MIN
, MIN
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 10 */
2267 _OPI(MAX
, MAX
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 11 */
2268 _OPI(SLT
, SLT
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 12 */
2269 _OPI(SGE
, SGE
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 13 */
2270 _OPI(EXP
, EX2
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
), /* 14 */
2271 _OPI(LOG
, LG2
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
), /* 15 */
2272 _OPI(LIT
, LIT
, V(0,0), V(3,0), V(0,0), V(0,0), 1, 1, NULL
), /* 16 */
2273 _OPI(DST
, DST
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* 17 */
2274 _OPI(LRP
, LRP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 3, NULL
), /* 18 */
2275 _OPI(FRC
, FRC
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
), /* 19 */
2277 _OPI(M4x4
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, SPECIAL(M4x4
)),
2278 _OPI(M4x3
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, SPECIAL(M4x3
)),
2279 _OPI(M3x4
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, SPECIAL(M3x4
)),
2280 _OPI(M3x3
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, SPECIAL(M3x3
)),
2281 _OPI(M3x2
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, SPECIAL(M3x2
)),
2283 _OPI(CALL
, CAL
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(CALL
)),
2284 _OPI(CALLNZ
, CAL
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(CALLNZ
)),
2285 _OPI(LOOP
, BGNLOOP
, V(2,0), V(3,0), V(3,0), V(3,0), 0, 2, SPECIAL(LOOP
)),
2286 _OPI(RET
, RET
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(RET
)),
2287 _OPI(ENDLOOP
, ENDLOOP
, V(2,0), V(3,0), V(3,0), V(3,0), 0, 0, SPECIAL(ENDLOOP
)),
2288 _OPI(LABEL
, NOP
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(LABEL
)),
2290 _OPI(DCL
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 0, 0, SPECIAL(DCL
)),
2292 _OPI(POW
, POW
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, SPECIAL(POW
)),
2293 _OPI(CRS
, XPD
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 2, NULL
), /* XXX: .w */
2294 _OPI(SGN
, SSG
, V(2,0), V(3,0), V(0,0), V(0,0), 1, 3, SPECIAL(SGN
)), /* ignore src1,2 */
2295 _OPI(ABS
, ABS
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, NULL
),
2296 _OPI(NRM
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 1, SPECIAL(NRM
)), /* NRM doesn't fit */
2298 _OPI(SINCOS
, SCS
, V(2,0), V(2,1), V(2,0), V(2,1), 1, 3, SPECIAL(SINCOS
)),
2299 _OPI(SINCOS
, SCS
, V(3,0), V(3,0), V(3,0), V(3,0), 1, 1, SPECIAL(SINCOS
)),
2301 /* More flow control */
2302 _OPI(REP
, NOP
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 1, SPECIAL(REP
)),
2303 _OPI(ENDREP
, NOP
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(ENDREP
)),
2304 _OPI(IF
, IF
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 1, SPECIAL(IF
)),
2305 _OPI(IFC
, IF
, V(2,1), V(3,0), V(2,1), V(3,0), 0, 2, SPECIAL(IFC
)),
2306 _OPI(ELSE
, ELSE
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(ELSE
)),
2307 _OPI(ENDIF
, ENDIF
, V(2,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(ENDIF
)),
2308 _OPI(BREAK
, BRK
, V(2,1), V(3,0), V(2,1), V(3,0), 0, 0, NULL
),
2309 _OPI(BREAKC
, BREAKC
, V(2,1), V(3,0), V(2,1), V(3,0), 0, 2, SPECIAL(BREAKC
)),
2311 _OPI(MOVA
, ARR
, V(2,0), V(3,0), V(0,0), V(0,0), 1, 1, NULL
),
2313 _OPI(DEFB
, NOP
, V(0,0), V(3,0) , V(0,0), V(3,0) , 1, 0, SPECIAL(DEFB
)),
2314 _OPI(DEFI
, NOP
, V(0,0), V(3,0) , V(0,0), V(3,0) , 1, 0, SPECIAL(DEFI
)),
2316 _OPI(TEXCOORD
, NOP
, V(0,0), V(0,0), V(0,0), V(1,3), 1, 0, SPECIAL(TEXCOORD
)),
2317 _OPI(TEXCOORD
, MOV
, V(0,0), V(0,0), V(1,4), V(1,4), 1, 1, SPECIAL(TEXCOORD_ps14
)),
2318 _OPI(TEXKILL
, KILL_IF
, V(0,0), V(0,0), V(0,0), V(3,0), 1, 0, SPECIAL(TEXKILL
)),
2319 _OPI(TEX
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 1, 0, SPECIAL(TEX
)),
2320 _OPI(TEX
, TEX
, V(0,0), V(0,0), V(1,4), V(1,4), 1, 1, SPECIAL(TEXLD_14
)),
2321 _OPI(TEX
, TEX
, V(0,0), V(0,0), V(2,0), V(3,0), 1, 2, SPECIAL(TEXLD
)),
2322 _OPI(TEXBEM
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXBEM
)),
2323 _OPI(TEXBEML
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXBEML
)),
2324 _OPI(TEXREG2AR
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXREG2AR
)),
2325 _OPI(TEXREG2GB
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXREG2GB
)),
2326 _OPI(TEXM3x2PAD
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXM3x2PAD
)),
2327 _OPI(TEXM3x2TEX
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXM3x2TEX
)),
2328 _OPI(TEXM3x3PAD
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXM3x3PAD
)),
2329 _OPI(TEXM3x3TEX
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXM3x3
)),
2330 _OPI(TEXM3x3SPEC
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXM3x3SPEC
)),
2331 _OPI(TEXM3x3VSPEC
, TEX
, V(0,0), V(0,0), V(0,0), V(1,3), 0, 0, SPECIAL(TEXM3x3VSPEC
)),
2333 _OPI(EXPP
, EXP
, V(0,0), V(1,1), V(0,0), V(0,0), 1, 1, NULL
),
2334 _OPI(EXPP
, EX2
, V(2,0), V(3,0), V(0,0), V(0,0), 1, 1, NULL
),
2335 _OPI(LOGP
, LG2
, V(0,0), V(3,0), V(0,0), V(0,0), 1, 1, NULL
),
2336 _OPI(CND
, NOP
, V(0,0), V(0,0), V(0,0), V(1,4), 1, 3, SPECIAL(CND
)),
2338 _OPI(DEF
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 1, 0, SPECIAL(DEF
)),
2340 /* More tex stuff */
2341 _OPI(TEXREG2RGB
, TEX
, V(0,0), V(0,0), V(1,2), V(1,3), 0, 0, SPECIAL(TEXREG2RGB
)),
2342 _OPI(TEXDP3TEX
, TEX
, V(0,0), V(0,0), V(1,2), V(1,3), 0, 0, SPECIAL(TEXDP3TEX
)),
2343 _OPI(TEXM3x2DEPTH
, TEX
, V(0,0), V(0,0), V(1,3), V(1,3), 0, 0, SPECIAL(TEXM3x2DEPTH
)),
2344 _OPI(TEXDP3
, TEX
, V(0,0), V(0,0), V(1,2), V(1,3), 0, 0, SPECIAL(TEXDP3
)),
2345 _OPI(TEXM3x3
, TEX
, V(0,0), V(0,0), V(1,2), V(1,3), 0, 0, SPECIAL(TEXM3x3
)),
2346 _OPI(TEXDEPTH
, TEX
, V(0,0), V(0,0), V(1,4), V(1,4), 0, 0, SPECIAL(TEXDEPTH
)),
2349 _OPI(CMP
, CMP
, V(0,0), V(0,0), V(1,2), V(3,0), 1, 3, SPECIAL(CMP
)), /* reversed */
2350 _OPI(BEM
, NOP
, V(0,0), V(0,0), V(1,4), V(1,4), 0, 0, SPECIAL(BEM
)),
2351 _OPI(DP2ADD
, NOP
, V(0,0), V(0,0), V(2,0), V(3,0), 1, 3, SPECIAL(DP2ADD
)),
2352 _OPI(DSX
, DDX
, V(0,0), V(0,0), V(2,1), V(3,0), 1, 1, NULL
),
2353 _OPI(DSY
, DDY
, V(0,0), V(0,0), V(2,1), V(3,0), 1, 1, NULL
),
2354 _OPI(TEXLDD
, TXD
, V(0,0), V(0,0), V(2,1), V(3,0), 1, 4, SPECIAL(TEXLDD
)),
2355 _OPI(SETP
, NOP
, V(0,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(SETP
)),
2356 _OPI(TEXLDL
, TXL
, V(3,0), V(3,0), V(3,0), V(3,0), 1, 2, SPECIAL(TEXLDL
)),
2357 _OPI(BREAKP
, BRK
, V(0,0), V(3,0), V(2,1), V(3,0), 0, 0, SPECIAL(BREAKP
))
2360 struct sm1_op_info inst_phase
=
2361 _OPI(PHASE
, NOP
, V(0,0), V(0,0), V(1,4), V(1,4), 0, 0, SPECIAL(PHASE
));
2363 struct sm1_op_info inst_comment
=
2364 _OPI(COMMENT
, NOP
, V(0,0), V(3,0), V(0,0), V(3,0), 0, 0, SPECIAL(COMMENT
));
2367 create_op_info_map(struct shader_translator
*tx
)
2369 const unsigned version
= (tx
->version
.major
<< 8) | tx
->version
.minor
;
2372 for (i
= 0; i
< Elements(tx
->op_info_map
); ++i
)
2373 tx
->op_info_map
[i
] = -1;
2375 if (tx
->processor
== TGSI_PROCESSOR_VERTEX
) {
2376 for (i
= 0; i
< Elements(inst_table
); ++i
) {
2377 assert(inst_table
[i
].sio
< Elements(tx
->op_info_map
));
2378 if (inst_table
[i
].vert_version
.min
<= version
&&
2379 inst_table
[i
].vert_version
.max
>= version
)
2380 tx
->op_info_map
[inst_table
[i
].sio
] = i
;
2383 for (i
= 0; i
< Elements(inst_table
); ++i
) {
2384 assert(inst_table
[i
].sio
< Elements(tx
->op_info_map
));
2385 if (inst_table
[i
].frag_version
.min
<= version
&&
2386 inst_table
[i
].frag_version
.max
>= version
)
2387 tx
->op_info_map
[inst_table
[i
].sio
] = i
;
2392 static INLINE HRESULT
2393 NineTranslateInstruction_Generic(struct shader_translator
*tx
)
2395 struct ureg_dst dst
[1];
2396 struct ureg_src src
[4];
2399 for (i
= 0; i
< tx
->insn
.ndst
&& i
< Elements(dst
); ++i
)
2400 dst
[i
] = tx_dst_param(tx
, &tx
->insn
.dst
[i
]);
2401 for (i
= 0; i
< tx
->insn
.nsrc
&& i
< Elements(src
); ++i
)
2402 src
[i
] = tx_src_param(tx
, &tx
->insn
.src
[i
]);
2404 ureg_insn(tx
->ureg
, tx
->insn
.info
->opcode
,
2406 src
, tx
->insn
.nsrc
);
2411 TOKEN_PEEK(struct shader_translator
*tx
)
2413 return *(tx
->parse
);
2417 TOKEN_NEXT(struct shader_translator
*tx
)
2419 return *(tx
->parse
)++;
2423 TOKEN_JUMP(struct shader_translator
*tx
)
2425 if (tx
->parse_next
&& tx
->parse
!= tx
->parse_next
) {
2426 WARN("parse(%p) != parse_next(%p) !\n", tx
->parse
, tx
->parse_next
);
2427 tx
->parse
= tx
->parse_next
;
2431 static INLINE boolean
2432 sm1_parse_eof(struct shader_translator
*tx
)
2434 return TOKEN_PEEK(tx
) == NINED3DSP_END
;
2438 sm1_read_version(struct shader_translator
*tx
)
2440 const DWORD tok
= TOKEN_NEXT(tx
);
2442 tx
->version
.major
= D3DSHADER_VERSION_MAJOR(tok
);
2443 tx
->version
.minor
= D3DSHADER_VERSION_MINOR(tok
);
2445 switch (tok
>> 16) {
2446 case NINED3D_SM1_VS
: tx
->processor
= TGSI_PROCESSOR_VERTEX
; break;
2447 case NINED3D_SM1_PS
: tx
->processor
= TGSI_PROCESSOR_FRAGMENT
; break;
2449 DBG("Invalid shader type: %x\n", tok
);
2455 /* This is just to check if we parsed the instruction properly. */
2457 sm1_parse_get_skip(struct shader_translator
*tx
)
2459 const DWORD tok
= TOKEN_PEEK(tx
);
2461 if (tx
->version
.major
>= 2) {
2462 tx
->parse_next
= tx
->parse
+ 1 /* this */ +
2463 ((tok
& D3DSI_INSTLENGTH_MASK
) >> D3DSI_INSTLENGTH_SHIFT
);
2465 tx
->parse_next
= NULL
; /* TODO: determine from param count */
2470 sm1_print_comment(const char *comment
, UINT size
)
2478 sm1_parse_comments(struct shader_translator
*tx
, BOOL print
)
2480 DWORD tok
= TOKEN_PEEK(tx
);
2482 while ((tok
& D3DSI_OPCODE_MASK
) == D3DSIO_COMMENT
)
2484 const char *comment
= "";
2485 UINT size
= (tok
& D3DSI_COMMENTSIZE_MASK
) >> D3DSI_COMMENTSIZE_SHIFT
;
2486 tx
->parse
+= size
+ 1;
2489 sm1_print_comment(comment
, size
);
2491 tok
= TOKEN_PEEK(tx
);
2496 sm1_parse_get_param(struct shader_translator
*tx
, DWORD
*reg
, DWORD
*rel
)
2498 *reg
= TOKEN_NEXT(tx
);
2500 if (*reg
& D3DSHADER_ADDRMODE_RELATIVE
)
2502 if (tx
->version
.major
< 2)
2504 ((D3DSPR_ADDR
<< D3DSP_REGTYPE_SHIFT2
) & D3DSP_REGTYPE_MASK2
) |
2505 ((D3DSPR_ADDR
<< D3DSP_REGTYPE_SHIFT
) & D3DSP_REGTYPE_MASK
) |
2506 (D3DSP_NOSWIZZLE
<< D3DSP_SWIZZLE_SHIFT
);
2508 *rel
= TOKEN_NEXT(tx
);
2513 sm1_parse_dst_param(struct sm1_dst_param
*dst
, DWORD tok
)
2517 (tok
& D3DSP_REGTYPE_MASK
) >> D3DSP_REGTYPE_SHIFT
|
2518 (tok
& D3DSP_REGTYPE_MASK2
) >> D3DSP_REGTYPE_SHIFT2
;
2519 dst
->type
= TGSI_RETURN_TYPE_FLOAT
;
2520 dst
->idx
= tok
& D3DSP_REGNUM_MASK
;
2522 dst
->mask
= (tok
& NINED3DSP_WRITEMASK_MASK
) >> NINED3DSP_WRITEMASK_SHIFT
;
2523 dst
->mod
= (tok
& D3DSP_DSTMOD_MASK
) >> D3DSP_DSTMOD_SHIFT
;
2524 shift
= (tok
& D3DSP_DSTSHIFT_MASK
) >> D3DSP_DSTSHIFT_SHIFT
;
2525 dst
->shift
= (shift
& 0x8) ? -(shift
& 0x7) : shift
& 0x7;
2529 sm1_parse_src_param(struct sm1_src_param
*src
, DWORD tok
)
2532 ((tok
& D3DSP_REGTYPE_MASK
) >> D3DSP_REGTYPE_SHIFT
) |
2533 ((tok
& D3DSP_REGTYPE_MASK2
) >> D3DSP_REGTYPE_SHIFT2
);
2534 src
->type
= TGSI_RETURN_TYPE_FLOAT
;
2535 src
->idx
= tok
& D3DSP_REGNUM_MASK
;
2537 src
->swizzle
= (tok
& D3DSP_SWIZZLE_MASK
) >> D3DSP_SWIZZLE_SHIFT
;
2538 src
->mod
= (tok
& D3DSP_SRCMOD_MASK
) >> D3DSP_SRCMOD_SHIFT
;
2540 switch (src
->file
) {
2541 case D3DSPR_CONST2
: src
->file
= D3DSPR_CONST
; src
->idx
+= 2048; break;
2542 case D3DSPR_CONST3
: src
->file
= D3DSPR_CONST
; src
->idx
+= 4096; break;
2543 case D3DSPR_CONST4
: src
->file
= D3DSPR_CONST
; src
->idx
+= 6144; break;
2550 sm1_parse_immediate(struct shader_translator
*tx
,
2551 struct sm1_src_param
*imm
)
2553 imm
->file
= NINED3DSPR_IMMEDIATE
;
2556 imm
->swizzle
= NINED3DSP_NOSWIZZLE
;
2558 switch (tx
->insn
.opcode
) {
2560 imm
->type
= NINED3DSPTYPE_FLOAT4
;
2561 memcpy(&imm
->imm
.d
[0], tx
->parse
, 4 * sizeof(DWORD
));
2565 imm
->type
= NINED3DSPTYPE_INT4
;
2566 memcpy(&imm
->imm
.d
[0], tx
->parse
, 4 * sizeof(DWORD
));
2570 imm
->type
= NINED3DSPTYPE_BOOL
;
2571 memcpy(&imm
->imm
.d
[0], tx
->parse
, 1 * sizeof(DWORD
));
2581 sm1_read_dst_param(struct shader_translator
*tx
,
2582 struct sm1_dst_param
*dst
,
2583 struct sm1_src_param
*rel
)
2585 DWORD tok_dst
, tok_rel
= 0;
2587 sm1_parse_get_param(tx
, &tok_dst
, &tok_rel
);
2588 sm1_parse_dst_param(dst
, tok_dst
);
2589 if (tok_dst
& D3DSHADER_ADDRMODE_RELATIVE
) {
2590 sm1_parse_src_param(rel
, tok_rel
);
2596 sm1_read_src_param(struct shader_translator
*tx
,
2597 struct sm1_src_param
*src
,
2598 struct sm1_src_param
*rel
)
2600 DWORD tok_src
, tok_rel
= 0;
2602 sm1_parse_get_param(tx
, &tok_src
, &tok_rel
);
2603 sm1_parse_src_param(src
, tok_src
);
2604 if (tok_src
& D3DSHADER_ADDRMODE_RELATIVE
) {
2606 sm1_parse_src_param(rel
, tok_rel
);
2612 sm1_read_semantic(struct shader_translator
*tx
,
2613 struct sm1_semantic
*sem
)
2615 const DWORD tok_usg
= TOKEN_NEXT(tx
);
2616 const DWORD tok_dst
= TOKEN_NEXT(tx
);
2618 sem
->sampler_type
= (tok_usg
& D3DSP_TEXTURETYPE_MASK
) >> D3DSP_TEXTURETYPE_SHIFT
;
2619 sem
->usage
= (tok_usg
& D3DSP_DCL_USAGE_MASK
) >> D3DSP_DCL_USAGE_SHIFT
;
2620 sem
->usage_idx
= (tok_usg
& D3DSP_DCL_USAGEINDEX_MASK
) >> D3DSP_DCL_USAGEINDEX_SHIFT
;
2622 sm1_parse_dst_param(&sem
->reg
, tok_dst
);
2626 sm1_parse_instruction(struct shader_translator
*tx
)
2628 struct sm1_instruction
*insn
= &tx
->insn
;
2630 struct sm1_op_info
*info
= NULL
;
2633 sm1_parse_comments(tx
, TRUE
);
2634 sm1_parse_get_skip(tx
);
2636 tok
= TOKEN_NEXT(tx
);
2638 insn
->opcode
= tok
& D3DSI_OPCODE_MASK
;
2639 insn
->flags
= (tok
& NINED3DSIO_OPCODE_FLAGS_MASK
) >> NINED3DSIO_OPCODE_FLAGS_SHIFT
;
2640 insn
->coissue
= !!(tok
& D3DSI_COISSUE
);
2641 insn
->predicated
= !!(tok
& NINED3DSHADER_INST_PREDICATED
);
2643 if (insn
->opcode
< Elements(tx
->op_info_map
)) {
2644 int k
= tx
->op_info_map
[insn
->opcode
];
2646 assert(k
< Elements(inst_table
));
2647 info
= &inst_table
[k
];
2650 if (insn
->opcode
== D3DSIO_PHASE
) info
= &inst_phase
;
2651 if (insn
->opcode
== D3DSIO_COMMENT
) info
= &inst_comment
;
2654 DBG("illegal or unhandled opcode: %08x\n", insn
->opcode
);
2659 insn
->ndst
= info
->ndst
;
2660 insn
->nsrc
= info
->nsrc
;
2662 assert(!insn
->predicated
&& "TODO: predicated instructions");
2666 unsigned min
= IS_VS
? info
->vert_version
.min
: info
->frag_version
.min
;
2667 unsigned max
= IS_VS
? info
->vert_version
.max
: info
->frag_version
.max
;
2668 unsigned ver
= (tx
->version
.major
<< 8) | tx
->version
.minor
;
2669 if (ver
< min
|| ver
> max
) {
2670 DBG("opcode not supported in this shader version: %x <= %x <= %x\n",
2676 for (i
= 0; i
< insn
->ndst
; ++i
)
2677 sm1_read_dst_param(tx
, &insn
->dst
[i
], &insn
->dst_rel
[i
]);
2678 if (insn
->predicated
)
2679 sm1_read_src_param(tx
, &insn
->pred
, NULL
);
2680 for (i
= 0; i
< insn
->nsrc
; ++i
)
2681 sm1_read_src_param(tx
, &insn
->src
[i
], &insn
->src_rel
[i
]);
2683 /* parse here so we can dump them before processing */
2684 if (insn
->opcode
== D3DSIO_DEF
||
2685 insn
->opcode
== D3DSIO_DEFI
||
2686 insn
->opcode
== D3DSIO_DEFB
)
2687 sm1_parse_immediate(tx
, &tx
->insn
.src
[0]);
2689 sm1_dump_instruction(insn
, tx
->cond_depth
+ tx
->loop_depth
);
2690 sm1_instruction_check(insn
);
2695 NineTranslateInstruction_Generic(tx
);
2696 tx_apply_dst0_modifiers(tx
);
2698 tx
->num_scratch
= 0; /* reset */
2704 tx_ctor(struct shader_translator
*tx
, struct nine_shader_info
*info
)
2710 tx
->byte_code
= info
->byte_code
;
2711 tx
->parse
= info
->byte_code
;
2713 for (i
= 0; i
< Elements(info
->input_map
); ++i
)
2714 info
->input_map
[i
] = NINE_DECLUSAGE_NONE
;
2715 info
->num_inputs
= 0;
2717 info
->position_t
= FALSE
;
2718 info
->point_size
= FALSE
;
2720 tx
->info
->const_used_size
= 0;
2722 info
->sampler_mask
= 0x0;
2723 info
->rt_mask
= 0x0;
2725 info
->lconstf
.data
= NULL
;
2726 info
->lconstf
.ranges
= NULL
;
2728 for (i
= 0; i
< Elements(tx
->regs
.aL
); ++i
) {
2729 tx
->regs
.aL
[i
] = ureg_dst_undef();
2730 tx
->regs
.rL
[i
] = ureg_dst_undef();
2732 tx
->regs
.a
= ureg_dst_undef();
2733 tx
->regs
.p
= ureg_dst_undef();
2734 tx
->regs
.oDepth
= ureg_dst_undef();
2735 tx
->regs
.vPos
= ureg_src_undef();
2736 tx
->regs
.vFace
= ureg_src_undef();
2737 for (i
= 0; i
< Elements(tx
->regs
.o
); ++i
)
2738 tx
->regs
.o
[i
] = ureg_dst_undef();
2739 for (i
= 0; i
< Elements(tx
->regs
.oCol
); ++i
)
2740 tx
->regs
.oCol
[i
] = ureg_dst_undef();
2741 for (i
= 0; i
< Elements(tx
->regs
.vC
); ++i
)
2742 tx
->regs
.vC
[i
] = ureg_src_undef();
2743 for (i
= 0; i
< Elements(tx
->regs
.vT
); ++i
)
2744 tx
->regs
.vT
[i
] = ureg_src_undef();
2746 for (i
= 0; i
< Elements(tx
->lconsti
); ++i
)
2747 tx
->lconsti
[i
].idx
= -1;
2748 for (i
= 0; i
< Elements(tx
->lconstb
); ++i
)
2749 tx
->lconstb
[i
].idx
= -1;
2751 sm1_read_version(tx
);
2753 info
->version
= (tx
->version
.major
<< 4) | tx
->version
.minor
;
2755 create_op_info_map(tx
);
2759 tx_dtor(struct shader_translator
*tx
)
2761 if (tx
->num_inst_labels
)
2762 FREE(tx
->inst_labels
);
2768 static INLINE
unsigned
2769 tgsi_processor_from_type(unsigned shader_type
)
2771 switch (shader_type
) {
2772 case PIPE_SHADER_VERTEX
: return TGSI_PROCESSOR_VERTEX
;
2773 case PIPE_SHADER_FRAGMENT
: return TGSI_PROCESSOR_FRAGMENT
;
2779 #define GET_CAP(n) device->screen->get_param( \
2780 device->screen, PIPE_CAP_##n)
2781 #define GET_SHADER_CAP(n) device->screen->get_shader_param( \
2782 device->screen, info->type, PIPE_SHADER_CAP_##n)
2785 nine_translate_shader(struct NineDevice9
*device
, struct nine_shader_info
*info
)
2787 struct shader_translator
*tx
;
2788 HRESULT hr
= D3D_OK
;
2789 const unsigned processor
= tgsi_processor_from_type(info
->type
);
2791 user_assert(processor
!= ~0, D3DERR_INVALIDCALL
);
2793 tx
= CALLOC_STRUCT(shader_translator
);
2795 return E_OUTOFMEMORY
;
2798 if (((tx
->version
.major
<< 16) | tx
->version
.minor
) > 0x00030000) {
2799 hr
= D3DERR_INVALIDCALL
;
2800 DBG("Unsupported shader version: %u.%u !\n",
2801 tx
->version
.major
, tx
->version
.minor
);
2804 if (tx
->processor
!= processor
) {
2805 hr
= D3DERR_INVALIDCALL
;
2806 DBG("Shader type mismatch: %u / %u !\n", tx
->processor
, processor
);
2809 DUMP("%s%u.%u\n", processor
== TGSI_PROCESSOR_VERTEX
? "VS" : "PS",
2810 tx
->version
.major
, tx
->version
.minor
);
2812 tx
->ureg
= ureg_create(processor
);
2817 tx_decl_constants(tx
);
2819 tx
->native_integers
= GET_SHADER_CAP(INTEGERS
);
2820 tx
->inline_subroutines
= !GET_SHADER_CAP(SUBROUTINES
);
2821 tx
->lower_preds
= !GET_SHADER_CAP(MAX_PREDS
);
2822 tx
->want_texcoord
= GET_CAP(TGSI_TEXCOORD
);
2823 tx
->shift_wpos
= !GET_CAP(TGSI_FS_COORD_PIXEL_CENTER_INTEGER
);
2824 tx
->texcoord_sn
= tx
->want_texcoord
?
2825 TGSI_SEMANTIC_TEXCOORD
: TGSI_SEMANTIC_GENERIC
;
2827 /* VS must always write position. Declare it here to make it the 1st output.
2828 * (Some drivers like nv50 are buggy and rely on that.)
2831 tx
->regs
.oPos
= ureg_DECL_output(tx
->ureg
, TGSI_SEMANTIC_POSITION
, 0);
2833 ureg_property(tx
->ureg
, TGSI_PROPERTY_FS_COORD_ORIGIN
, TGSI_FS_COORD_ORIGIN_UPPER_LEFT
);
2834 if (!tx
->shift_wpos
)
2835 ureg_property(tx
->ureg
, TGSI_PROPERTY_FS_COORD_PIXEL_CENTER
, TGSI_FS_COORD_PIXEL_CENTER_INTEGER
);
2838 while (!sm1_parse_eof(tx
))
2839 sm1_parse_instruction(tx
);
2840 tx
->parse
++; /* for byte_size */
2842 if (IS_PS
&& (tx
->version
.major
< 2) && tx
->num_temp
) {
2843 ureg_MOV(tx
->ureg
, ureg_DECL_output(tx
->ureg
, TGSI_SEMANTIC_COLOR
, 0),
2844 ureg_src(tx
->regs
.r
[0]));
2845 info
->rt_mask
|= 0x1;
2848 if (info
->position_t
)
2849 ureg_property(tx
->ureg
, TGSI_PROPERTY_VS_WINDOW_SPACE_POSITION
, TRUE
);
2853 if (IS_VS
&& !ureg_dst_is_undef(tx
->regs
.oPts
))
2854 info
->point_size
= TRUE
;
2856 if (debug_get_bool_option("NINE_TGSI_DUMP", FALSE
)) {
2858 const struct tgsi_token
*toks
= ureg_get_tokens(tx
->ureg
, &count
);
2860 ureg_free_tokens(toks
);
2863 /* record local constants */
2864 if (tx
->num_lconstf
&& tx
->indirect_const_access
) {
2865 struct nine_range
*ranges
;
2872 data
= MALLOC(tx
->num_lconstf
* 4 * sizeof(float));
2875 info
->lconstf
.data
= data
;
2877 indices
= MALLOC(tx
->num_lconstf
* sizeof(indices
[0]));
2881 /* lazy sort, num_lconstf should be small */
2882 for (n
= 0; n
< tx
->num_lconstf
; ++n
) {
2883 for (k
= 0, i
= 0; i
< tx
->num_lconstf
; ++i
) {
2884 if (tx
->lconstf
[i
].idx
< tx
->lconstf
[k
].idx
)
2887 indices
[n
] = tx
->lconstf
[k
].idx
;
2888 memcpy(&data
[n
* 4], &tx
->lconstf
[k
].imm
.f
[0], 4 * sizeof(float));
2889 tx
->lconstf
[k
].idx
= INT_MAX
;
2893 for (n
= 1, i
= 1; i
< tx
->num_lconstf
; ++i
)
2894 if (indices
[i
] != indices
[i
- 1] + 1)
2896 ranges
= MALLOC(n
* sizeof(ranges
[0]));
2901 info
->lconstf
.ranges
= ranges
;
2904 ranges
[k
].bgn
= indices
[0];
2905 for (i
= 1; i
< tx
->num_lconstf
; ++i
) {
2906 if (indices
[i
] != indices
[i
- 1] + 1) {
2907 ranges
[k
].next
= &ranges
[k
+ 1];
2908 ranges
[k
].end
= indices
[i
- 1] + 1;
2910 ranges
[k
].bgn
= indices
[i
];
2913 ranges
[k
].end
= indices
[i
- 1] + 1;
2914 ranges
[k
].next
= NULL
;
2915 assert(n
== (k
+ 1));
2921 if (tx
->indirect_const_access
)
2922 info
->const_used_size
= ~0;
2924 info
->cso
= ureg_create_shader_and_destroy(tx
->ureg
, device
->pipe
);
2926 hr
= D3DERR_DRIVERINTERNALERROR
;
2927 FREE(info
->lconstf
.data
);
2928 FREE(info
->lconstf
.ranges
);
2932 info
->byte_size
= (tx
->parse
- tx
->byte_code
) * sizeof(DWORD
);