2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
23 #include "pipe/p_shader_tokens.h"
24 #include "tgsi/tgsi_parse.h"
25 #include "tgsi/tgsi_scan.h"
26 #include "tgsi/tgsi_dump.h"
27 #include "util/u_format.h"
28 #include "r600_pipe.h"
31 #include "r600_opcodes.h"
36 static void r600_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
38 struct r600_pipe_state
*rstate
= &shader
->rstate
;
39 struct r600_shader
*rshader
= &shader
->shader
;
40 unsigned spi_vs_out_id
[10];
43 /* clear previous register */
46 /* so far never got proper semantic id from tgsi */
47 /* FIXME better to move this in config things so they get emited
48 * only one time per cs
50 for (i
= 0; i
< 10; i
++) {
53 for (i
= 0; i
< 32; i
++) {
54 tmp
= i
<< ((i
& 3) * 8);
55 spi_vs_out_id
[i
/ 4] |= tmp
;
57 for (i
= 0; i
< 10; i
++) {
58 r600_pipe_state_add_reg(rstate
,
59 R_028614_SPI_VS_OUT_ID_0
+ i
* 4,
60 spi_vs_out_id
[i
], 0xFFFFFFFF, NULL
);
63 r600_pipe_state_add_reg(rstate
,
64 R_0286C4_SPI_VS_OUT_CONFIG
,
65 S_0286C4_VS_EXPORT_COUNT(rshader
->noutput
- 2),
67 r600_pipe_state_add_reg(rstate
,
68 R_028868_SQ_PGM_RESOURCES_VS
,
69 S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
70 S_028868_STACK_SIZE(rshader
->bc
.nstack
),
72 r600_pipe_state_add_reg(rstate
,
73 R_0288D0_SQ_PGM_CF_OFFSET_VS
,
74 0x00000000, 0xFFFFFFFF, NULL
);
75 r600_pipe_state_add_reg(rstate
,
76 R_028858_SQ_PGM_START_VS
,
77 r600_bo_offset(shader
->bo
) >> 8, 0xFFFFFFFF, shader
->bo
);
79 r600_pipe_state_add_reg(rstate
,
80 R_03E200_SQ_LOOP_CONST_0
+ (32 * 4), 0x01000FFF,
85 int r600_find_vs_semantic_index(struct r600_shader
*vs
,
86 struct r600_shader
*ps
, int id
)
88 struct r600_shader_io
*input
= &ps
->input
[id
];
90 for (int i
= 0; i
< vs
->noutput
; i
++) {
91 if (input
->name
== vs
->output
[i
].name
&&
92 input
->sid
== vs
->output
[i
].sid
) {
99 static void r600_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
101 struct r600_pipe_state
*rstate
= &shader
->rstate
;
102 struct r600_shader
*rshader
= &shader
->shader
;
103 unsigned i
, exports_ps
, num_cout
, spi_ps_in_control_0
, spi_input_z
, spi_ps_in_control_1
;
104 int pos_index
= -1, face_index
= -1;
108 for (i
= 0; i
< rshader
->ninput
; i
++) {
109 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_POSITION
)
111 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_FACE
)
115 for (i
= 0; i
< rshader
->noutput
; i
++) {
116 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
)
117 r600_pipe_state_add_reg(rstate
,
118 R_02880C_DB_SHADER_CONTROL
,
119 S_02880C_Z_EXPORT_ENABLE(1),
120 S_02880C_Z_EXPORT_ENABLE(1), NULL
);
121 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
)
122 r600_pipe_state_add_reg(rstate
,
123 R_02880C_DB_SHADER_CONTROL
,
124 S_02880C_STENCIL_REF_EXPORT_ENABLE(1),
125 S_02880C_STENCIL_REF_EXPORT_ENABLE(1), NULL
);
130 for (i
= 0; i
< rshader
->noutput
; i
++) {
131 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
|| rshader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
)
133 else if (rshader
->output
[i
].name
== TGSI_SEMANTIC_COLOR
) {
137 exports_ps
|= S_028854_EXPORT_COLORS(num_cout
);
139 /* always at least export 1 component per pixel */
143 spi_ps_in_control_0
= S_0286CC_NUM_INTERP(rshader
->ninput
) |
144 S_0286CC_PERSP_GRADIENT_ENA(1);
146 if (pos_index
!= -1) {
147 spi_ps_in_control_0
|= (S_0286CC_POSITION_ENA(1) |
148 S_0286CC_POSITION_CENTROID(rshader
->input
[pos_index
].centroid
) |
149 S_0286CC_POSITION_ADDR(rshader
->input
[pos_index
].gpr
) |
150 S_0286CC_BARYC_SAMPLE_CNTL(1));
154 spi_ps_in_control_1
= 0;
155 if (face_index
!= -1) {
156 spi_ps_in_control_1
|= S_0286D0_FRONT_FACE_ENA(1) |
157 S_0286D0_FRONT_FACE_ADDR(rshader
->input
[face_index
].gpr
);
160 r600_pipe_state_add_reg(rstate
, R_0286CC_SPI_PS_IN_CONTROL_0
, spi_ps_in_control_0
, 0xFFFFFFFF, NULL
);
161 r600_pipe_state_add_reg(rstate
, R_0286D0_SPI_PS_IN_CONTROL_1
, spi_ps_in_control_1
, 0xFFFFFFFF, NULL
);
162 r600_pipe_state_add_reg(rstate
, R_0286D8_SPI_INPUT_Z
, spi_input_z
, 0xFFFFFFFF, NULL
);
163 r600_pipe_state_add_reg(rstate
,
164 R_028840_SQ_PGM_START_PS
,
165 r600_bo_offset(shader
->bo
) >> 8, 0xFFFFFFFF, shader
->bo
);
166 r600_pipe_state_add_reg(rstate
,
167 R_028850_SQ_PGM_RESOURCES_PS
,
168 S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
169 S_028868_STACK_SIZE(rshader
->bc
.nstack
),
171 r600_pipe_state_add_reg(rstate
,
172 R_028854_SQ_PGM_EXPORTS_PS
,
173 exports_ps
, 0xFFFFFFFF, NULL
);
174 r600_pipe_state_add_reg(rstate
,
175 R_0288CC_SQ_PGM_CF_OFFSET_PS
,
176 0x00000000, 0xFFFFFFFF, NULL
);
178 if (rshader
->fs_write_all
) {
179 r600_pipe_state_add_reg(rstate
, R_028808_CB_COLOR_CONTROL
,
180 S_028808_MULTIWRITE_ENABLE(1),
181 S_028808_MULTIWRITE_ENABLE(1),
185 if (rshader
->uses_kill
) {
186 /* only set some bits here, the other bits are set in the dsa state */
187 r600_pipe_state_add_reg(rstate
,
188 R_02880C_DB_SHADER_CONTROL
,
189 S_02880C_KILL_ENABLE(1),
190 S_02880C_KILL_ENABLE(1), NULL
);
192 r600_pipe_state_add_reg(rstate
,
193 R_03E200_SQ_LOOP_CONST_0
, 0x01000FFF,
197 static int r600_pipe_shader(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
199 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
200 struct r600_shader
*rshader
= &shader
->shader
;
203 /* copy new shader */
204 if (shader
->bo
== NULL
) {
205 shader
->bo
= r600_bo(rctx
->radeon
, rshader
->bc
.ndw
* 4, 4096, 0, 0);
206 if (shader
->bo
== NULL
) {
209 ptr
= r600_bo_map(rctx
->radeon
, shader
->bo
, 0, NULL
);
210 memcpy(ptr
, rshader
->bc
.bytecode
, rshader
->bc
.ndw
* 4);
211 r600_bo_unmap(rctx
->radeon
, shader
->bo
);
214 switch (rshader
->processor_type
) {
215 case TGSI_PROCESSOR_VERTEX
:
216 if (rshader
->family
>= CHIP_CEDAR
) {
217 evergreen_pipe_shader_vs(ctx
, shader
);
219 r600_pipe_shader_vs(ctx
, shader
);
222 case TGSI_PROCESSOR_FRAGMENT
:
223 if (rshader
->family
>= CHIP_CEDAR
) {
224 evergreen_pipe_shader_ps(ctx
, shader
);
226 r600_pipe_shader_ps(ctx
, shader
);
235 int r600_shader_from_tgsi(const struct tgsi_token
*tokens
, struct r600_shader
*shader
, u32
**literals
);
236 int r600_pipe_shader_create(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
, const struct tgsi_token
*tokens
)
238 static int dump_shaders
= -1;
239 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
243 /* Would like some magic "get_bool_option_once" routine.
245 if (dump_shaders
== -1)
246 dump_shaders
= debug_get_bool_option("R600_DUMP_SHADERS", FALSE
);
249 fprintf(stderr
, "--------------------------------------------------------------\n");
250 tgsi_dump(tokens
, 0);
252 shader
->shader
.family
= r600_get_family(rctx
->radeon
);
253 r
= r600_shader_from_tgsi(tokens
, &shader
->shader
, &literals
);
255 R600_ERR("translation from TGSI failed !\n");
258 r
= r600_bc_build(&shader
->shader
.bc
);
261 R600_ERR("building bytecode failed !\n");
265 r600_bc_dump(&shader
->shader
.bc
);
266 fprintf(stderr
, "______________________________________________________________\n");
268 return r600_pipe_shader(ctx
, shader
);
271 void r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
273 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
275 r600_bo_reference(rctx
->radeon
, &shader
->bo
, NULL
);
276 r600_bc_clear(&shader
->shader
.bc
);
280 * tgsi -> r600 shader
282 struct r600_shader_tgsi_instruction
;
284 struct r600_shader_ctx
{
285 struct tgsi_shader_info info
;
286 struct tgsi_parse_context parse
;
287 const struct tgsi_token
*tokens
;
289 unsigned file_offset
[TGSI_FILE_COUNT
];
291 struct r600_shader_tgsi_instruction
*inst_info
;
293 struct r600_shader
*shader
;
296 u32 max_driver_temp_used
;
297 /* needed for evergreen interpolation */
298 boolean input_centroid
;
299 boolean input_linear
;
300 boolean input_perspective
;
304 struct r600_shader_tgsi_instruction
{
305 unsigned tgsi_opcode
;
307 unsigned r600_opcode
;
308 int (*process
)(struct r600_shader_ctx
*ctx
);
311 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[], eg_shader_tgsi_instruction
[];
312 static int tgsi_helper_tempx_replicate(struct r600_shader_ctx
*ctx
);
314 static int tgsi_is_supported(struct r600_shader_ctx
*ctx
)
316 struct tgsi_full_instruction
*i
= &ctx
->parse
.FullToken
.FullInstruction
;
319 if (i
->Instruction
.NumDstRegs
> 1) {
320 R600_ERR("too many dst (%d)\n", i
->Instruction
.NumDstRegs
);
323 if (i
->Instruction
.Predicate
) {
324 R600_ERR("predicate unsupported\n");
328 if (i
->Instruction
.Label
) {
329 R600_ERR("label unsupported\n");
333 for (j
= 0; j
< i
->Instruction
.NumSrcRegs
; j
++) {
334 if (i
->Src
[j
].Register
.Dimension
) {
335 R600_ERR("unsupported src %d (dimension %d)\n", j
,
336 i
->Src
[j
].Register
.Dimension
);
340 for (j
= 0; j
< i
->Instruction
.NumDstRegs
; j
++) {
341 if (i
->Dst
[j
].Register
.Dimension
) {
342 R600_ERR("unsupported dst (dimension)\n");
349 static int evergreen_interp_alu(struct r600_shader_ctx
*ctx
, int input
)
352 struct r600_bc_alu alu
;
353 int gpr
= 0, base_chan
= 0;
356 if (ctx
->shader
->input
[input
].interpolate
== TGSI_INTERPOLATE_PERSPECTIVE
) {
358 if (ctx
->shader
->input
[input
].centroid
)
360 } else if (ctx
->shader
->input
[input
].interpolate
== TGSI_INTERPOLATE_LINEAR
) {
362 /* if we have perspective add one */
363 if (ctx
->input_perspective
) {
365 /* if we have perspective centroid */
366 if (ctx
->input_centroid
)
369 if (ctx
->shader
->input
[input
].centroid
)
373 /* work out gpr and base_chan from index */
375 base_chan
= (2 * (ij_index
% 2)) + 1;
377 for (i
= 0; i
< 8; i
++) {
378 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
381 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_ZW
;
383 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_XY
;
385 if ((i
> 1) && (i
< 6)) {
386 alu
.dst
.sel
= ctx
->shader
->input
[input
].gpr
;
390 alu
.dst
.chan
= i
% 4;
392 alu
.src
[0].sel
= gpr
;
393 alu
.src
[0].chan
= (base_chan
- (i
% 2));
395 alu
.src
[1].sel
= V_SQ_ALU_SRC_PARAM_BASE
+ ctx
->shader
->input
[input
].lds_pos
;
397 alu
.bank_swizzle_force
= SQ_ALU_VEC_210
;
400 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
408 static int tgsi_declaration(struct r600_shader_ctx
*ctx
)
410 struct tgsi_full_declaration
*d
= &ctx
->parse
.FullToken
.FullDeclaration
;
413 switch (d
->Declaration
.File
) {
414 case TGSI_FILE_INPUT
:
415 i
= ctx
->shader
->ninput
++;
416 ctx
->shader
->input
[i
].name
= d
->Semantic
.Name
;
417 ctx
->shader
->input
[i
].sid
= d
->Semantic
.Index
;
418 ctx
->shader
->input
[i
].interpolate
= d
->Declaration
.Interpolate
;
419 ctx
->shader
->input
[i
].centroid
= d
->Declaration
.Centroid
;
420 ctx
->shader
->input
[i
].gpr
= ctx
->file_offset
[TGSI_FILE_INPUT
] + i
;
421 if (ctx
->type
== TGSI_PROCESSOR_FRAGMENT
&& ctx
->bc
->chiprev
== CHIPREV_EVERGREEN
) {
422 /* turn input into interpolate on EG */
423 if (ctx
->shader
->input
[i
].name
!= TGSI_SEMANTIC_POSITION
) {
424 if (ctx
->shader
->input
[i
].interpolate
> 0) {
425 ctx
->shader
->input
[i
].lds_pos
= ctx
->shader
->nlds
++;
426 evergreen_interp_alu(ctx
, i
);
431 case TGSI_FILE_OUTPUT
:
432 i
= ctx
->shader
->noutput
++;
433 ctx
->shader
->output
[i
].name
= d
->Semantic
.Name
;
434 ctx
->shader
->output
[i
].sid
= d
->Semantic
.Index
;
435 ctx
->shader
->output
[i
].gpr
= ctx
->file_offset
[TGSI_FILE_OUTPUT
] + i
;
436 ctx
->shader
->output
[i
].interpolate
= d
->Declaration
.Interpolate
;
438 case TGSI_FILE_CONSTANT
:
439 case TGSI_FILE_TEMPORARY
:
440 case TGSI_FILE_SAMPLER
:
441 case TGSI_FILE_ADDRESS
:
444 R600_ERR("unsupported file %d declaration\n", d
->Declaration
.File
);
450 static int r600_get_temp(struct r600_shader_ctx
*ctx
)
452 return ctx
->temp_reg
+ ctx
->max_driver_temp_used
++;
456 * for evergreen we need to scan the shader to find the number of GPRs we need to
457 * reserve for interpolation.
459 * we need to know if we are going to emit
460 * any centroid inputs
461 * if perspective and linear are required
463 static int evergreen_gpr_count(struct r600_shader_ctx
*ctx
)
468 ctx
->input_linear
= FALSE
;
469 ctx
->input_perspective
= FALSE
;
470 ctx
->input_centroid
= FALSE
;
471 ctx
->num_interp_gpr
= 1;
473 /* any centroid inputs */
474 for (i
= 0; i
< ctx
->info
.num_inputs
; i
++) {
475 /* skip position/face */
476 if (ctx
->info
.input_semantic_name
[i
] == TGSI_SEMANTIC_POSITION
||
477 ctx
->info
.input_semantic_name
[i
] == TGSI_SEMANTIC_FACE
)
479 if (ctx
->info
.input_interpolate
[i
] == TGSI_INTERPOLATE_LINEAR
)
480 ctx
->input_linear
= TRUE
;
481 if (ctx
->info
.input_interpolate
[i
] == TGSI_INTERPOLATE_PERSPECTIVE
)
482 ctx
->input_perspective
= TRUE
;
483 if (ctx
->info
.input_centroid
[i
])
484 ctx
->input_centroid
= TRUE
;
488 /* ignoring sample for now */
489 if (ctx
->input_perspective
)
491 if (ctx
->input_linear
)
493 if (ctx
->input_centroid
)
496 ctx
->num_interp_gpr
+= (num_baryc
+ 1) >> 1;
498 /* TODO PULL MODEL and LINE STIPPLE, FIXED PT POS */
499 return ctx
->num_interp_gpr
;
502 int r600_shader_from_tgsi(const struct tgsi_token
*tokens
, struct r600_shader
*shader
, u32
**literals
)
504 struct tgsi_full_immediate
*immediate
;
505 struct tgsi_full_property
*property
;
506 struct r600_shader_ctx ctx
;
507 struct r600_bc_output output
[32];
508 unsigned output_done
, noutput
;
512 ctx
.bc
= &shader
->bc
;
514 r
= r600_bc_init(ctx
.bc
, shader
->family
);
518 tgsi_scan_shader(tokens
, &ctx
.info
);
519 tgsi_parse_init(&ctx
.parse
, tokens
);
520 ctx
.type
= ctx
.parse
.FullHeader
.Processor
.Processor
;
521 shader
->processor_type
= ctx
.type
;
522 ctx
.bc
->type
= shader
->processor_type
;
524 /* register allocations */
525 /* Values [0,127] correspond to GPR[0..127].
526 * Values [128,159] correspond to constant buffer bank 0
527 * Values [160,191] correspond to constant buffer bank 1
528 * Values [256,511] correspond to cfile constants c[0..255]. (Gone on EG)
529 * Values [256,287] correspond to constant buffer bank 2 (EG)
530 * Values [288,319] correspond to constant buffer bank 3 (EG)
531 * Other special values are shown in the list below.
532 * 244 ALU_SRC_1_DBL_L: special constant 1.0 double-float, LSW. (RV670+)
533 * 245 ALU_SRC_1_DBL_M: special constant 1.0 double-float, MSW. (RV670+)
534 * 246 ALU_SRC_0_5_DBL_L: special constant 0.5 double-float, LSW. (RV670+)
535 * 247 ALU_SRC_0_5_DBL_M: special constant 0.5 double-float, MSW. (RV670+)
536 * 248 SQ_ALU_SRC_0: special constant 0.0.
537 * 249 SQ_ALU_SRC_1: special constant 1.0 float.
538 * 250 SQ_ALU_SRC_1_INT: special constant 1 integer.
539 * 251 SQ_ALU_SRC_M_1_INT: special constant -1 integer.
540 * 252 SQ_ALU_SRC_0_5: special constant 0.5 float.
541 * 253 SQ_ALU_SRC_LITERAL: literal constant.
542 * 254 SQ_ALU_SRC_PV: previous vector result.
543 * 255 SQ_ALU_SRC_PS: previous scalar result.
545 for (i
= 0; i
< TGSI_FILE_COUNT
; i
++) {
546 ctx
.file_offset
[i
] = 0;
548 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
549 ctx
.file_offset
[TGSI_FILE_INPUT
] = 1;
550 if (ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
) {
551 r600_bc_add_cfinst(ctx
.bc
, EG_V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS
);
553 r600_bc_add_cfinst(ctx
.bc
, V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS
);
556 if (ctx
.type
== TGSI_PROCESSOR_FRAGMENT
&& ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
) {
557 ctx
.file_offset
[TGSI_FILE_INPUT
] = evergreen_gpr_count(&ctx
);
559 ctx
.file_offset
[TGSI_FILE_OUTPUT
] = ctx
.file_offset
[TGSI_FILE_INPUT
] +
560 ctx
.info
.file_count
[TGSI_FILE_INPUT
];
561 ctx
.file_offset
[TGSI_FILE_TEMPORARY
] = ctx
.file_offset
[TGSI_FILE_OUTPUT
] +
562 ctx
.info
.file_count
[TGSI_FILE_OUTPUT
];
564 /* Outside the GPR range. This will be translated to one of the
565 * kcache banks later. */
566 ctx
.file_offset
[TGSI_FILE_CONSTANT
] = 512;
568 ctx
.file_offset
[TGSI_FILE_IMMEDIATE
] = V_SQ_ALU_SRC_LITERAL
;
569 ctx
.temp_reg
= ctx
.file_offset
[TGSI_FILE_TEMPORARY
] +
570 ctx
.info
.file_count
[TGSI_FILE_TEMPORARY
];
574 shader
->fs_write_all
= FALSE
;
575 while (!tgsi_parse_end_of_tokens(&ctx
.parse
)) {
576 tgsi_parse_token(&ctx
.parse
);
577 switch (ctx
.parse
.FullToken
.Token
.Type
) {
578 case TGSI_TOKEN_TYPE_IMMEDIATE
:
579 immediate
= &ctx
.parse
.FullToken
.FullImmediate
;
580 ctx
.literals
= realloc(ctx
.literals
, (ctx
.nliterals
+ 1) * 16);
581 if(ctx
.literals
== NULL
) {
585 ctx
.literals
[ctx
.nliterals
* 4 + 0] = immediate
->u
[0].Uint
;
586 ctx
.literals
[ctx
.nliterals
* 4 + 1] = immediate
->u
[1].Uint
;
587 ctx
.literals
[ctx
.nliterals
* 4 + 2] = immediate
->u
[2].Uint
;
588 ctx
.literals
[ctx
.nliterals
* 4 + 3] = immediate
->u
[3].Uint
;
591 case TGSI_TOKEN_TYPE_DECLARATION
:
592 r
= tgsi_declaration(&ctx
);
596 case TGSI_TOKEN_TYPE_INSTRUCTION
:
597 r
= tgsi_is_supported(&ctx
);
600 ctx
.max_driver_temp_used
= 0;
601 /* reserve first tmp for everyone */
603 opcode
= ctx
.parse
.FullToken
.FullInstruction
.Instruction
.Opcode
;
604 if (ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
)
605 ctx
.inst_info
= &eg_shader_tgsi_instruction
[opcode
];
607 ctx
.inst_info
= &r600_shader_tgsi_instruction
[opcode
];
608 r
= ctx
.inst_info
->process(&ctx
);
612 case TGSI_TOKEN_TYPE_PROPERTY
:
613 property
= &ctx
.parse
.FullToken
.FullProperty
;
614 if (property
->Property
.PropertyName
== TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS
) {
615 if (property
->u
[0].Data
== 1)
616 shader
->fs_write_all
= TRUE
;
620 R600_ERR("unsupported token type %d\n", ctx
.parse
.FullToken
.Token
.Type
);
626 noutput
= shader
->noutput
;
627 for (i
= 0, pos0
= 0; i
< noutput
; i
++) {
628 memset(&output
[i
], 0, sizeof(struct r600_bc_output
));
629 output
[i
].gpr
= shader
->output
[i
].gpr
;
630 output
[i
].elem_size
= 3;
631 output
[i
].swizzle_x
= 0;
632 output
[i
].swizzle_y
= 1;
633 output
[i
].swizzle_z
= 2;
634 output
[i
].swizzle_w
= 3;
635 output
[i
].burst_count
= 1;
636 output
[i
].barrier
= 1;
637 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
;
638 output
[i
].array_base
= i
- pos0
;
639 output
[i
].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
);
641 case TGSI_PROCESSOR_VERTEX
:
642 if (shader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
) {
643 output
[i
].array_base
= 60;
644 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS
;
645 /* position doesn't count in array_base */
648 if (shader
->output
[i
].name
== TGSI_SEMANTIC_PSIZE
) {
649 output
[i
].array_base
= 61;
650 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS
;
651 /* position doesn't count in array_base */
655 case TGSI_PROCESSOR_FRAGMENT
:
656 if (shader
->output
[i
].name
== TGSI_SEMANTIC_COLOR
) {
657 output
[i
].array_base
= shader
->output
[i
].sid
;
658 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
659 } else if (shader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
) {
660 output
[i
].array_base
= 61;
661 output
[i
].swizzle_x
= 2;
662 output
[i
].swizzle_y
= 7;
663 output
[i
].swizzle_z
= output
[i
].swizzle_w
= 7;
664 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
665 } else if (shader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
) {
666 output
[i
].array_base
= 61;
667 output
[i
].swizzle_x
= 7;
668 output
[i
].swizzle_y
= 1;
669 output
[i
].swizzle_z
= output
[i
].swizzle_w
= 7;
670 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
672 R600_ERR("unsupported fragment output name %d\n", shader
->output
[i
].name
);
678 R600_ERR("unsupported processor type %d\n", ctx
.type
);
683 /* add fake param output for vertex shader if no param is exported */
684 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
685 for (i
= 0, pos0
= 0; i
< noutput
; i
++) {
686 if (output
[i
].type
== V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
) {
692 memset(&output
[i
], 0, sizeof(struct r600_bc_output
));
694 output
[i
].elem_size
= 3;
695 output
[i
].swizzle_x
= 0;
696 output
[i
].swizzle_y
= 1;
697 output
[i
].swizzle_z
= 2;
698 output
[i
].swizzle_w
= 3;
699 output
[i
].burst_count
= 1;
700 output
[i
].barrier
= 1;
701 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
;
702 output
[i
].array_base
= 0;
703 output
[i
].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
);
707 /* add fake pixel export */
708 if (ctx
.type
== TGSI_PROCESSOR_FRAGMENT
&& !noutput
) {
709 memset(&output
[0], 0, sizeof(struct r600_bc_output
));
711 output
[0].elem_size
= 3;
712 output
[0].swizzle_x
= 7;
713 output
[0].swizzle_y
= 7;
714 output
[0].swizzle_z
= 7;
715 output
[0].swizzle_w
= 7;
716 output
[0].burst_count
= 1;
717 output
[0].barrier
= 1;
718 output
[0].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
719 output
[0].array_base
= 0;
720 output
[0].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
);
723 /* set export done on last export of each type */
724 for (i
= noutput
- 1, output_done
= 0; i
>= 0; i
--) {
725 if (i
== (noutput
- 1)) {
726 output
[i
].end_of_program
= 1;
728 if (!(output_done
& (1 << output
[i
].type
))) {
729 output_done
|= (1 << output
[i
].type
);
730 output
[i
].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE
);
733 /* add output to bytecode */
734 for (i
= 0; i
< noutput
; i
++) {
735 r
= r600_bc_add_output(ctx
.bc
, &output
[i
]);
739 *literals
= ctx
.literals
;
740 tgsi_parse_free(&ctx
.parse
);
744 tgsi_parse_free(&ctx
.parse
);
748 static int tgsi_unsupported(struct r600_shader_ctx
*ctx
)
750 R600_ERR("%d tgsi opcode unsupported\n", ctx
->inst_info
->tgsi_opcode
);
754 static int tgsi_end(struct r600_shader_ctx
*ctx
)
759 static void tgsi_src(struct r600_shader_ctx
*ctx
,
760 const struct tgsi_full_src_register
*tgsi_src
,
761 struct r600_bc_alu_src
*r600_src
)
763 memset(r600_src
, 0, sizeof(struct r600_bc_alu_src
));
764 r600_src
->neg
= tgsi_src
->Register
.Negate
;
765 r600_src
->abs
= tgsi_src
->Register
.Absolute
;
766 if (tgsi_src
->Register
.File
== TGSI_FILE_IMMEDIATE
) {
768 if((tgsi_src
->Register
.SwizzleX
== tgsi_src
->Register
.SwizzleY
) &&
769 (tgsi_src
->Register
.SwizzleX
== tgsi_src
->Register
.SwizzleZ
) &&
770 (tgsi_src
->Register
.SwizzleX
== tgsi_src
->Register
.SwizzleW
)) {
772 index
= tgsi_src
->Register
.Index
* 4 + tgsi_src
->Register
.SwizzleX
;
773 r600_bc_special_constants(ctx
->literals
[index
], &r600_src
->sel
, &r600_src
->neg
);
774 if (r600_src
->sel
!= V_SQ_ALU_SRC_LITERAL
)
777 index
= tgsi_src
->Register
.Index
;
778 r600_src
->sel
= V_SQ_ALU_SRC_LITERAL
;
779 r600_src
->value
= ctx
->literals
+ index
* 4;
781 if (tgsi_src
->Register
.Indirect
)
782 r600_src
->rel
= V_SQ_REL_RELATIVE
;
783 r600_src
->sel
= tgsi_src
->Register
.Index
;
784 r600_src
->sel
+= ctx
->file_offset
[tgsi_src
->Register
.File
];
788 static void tgsi_dst(struct r600_shader_ctx
*ctx
,
789 const struct tgsi_full_dst_register
*tgsi_dst
,
791 struct r600_bc_alu_dst
*r600_dst
)
793 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
795 r600_dst
->sel
= tgsi_dst
->Register
.Index
;
796 r600_dst
->sel
+= ctx
->file_offset
[tgsi_dst
->Register
.File
];
797 r600_dst
->chan
= swizzle
;
799 if (tgsi_dst
->Register
.Indirect
)
800 r600_dst
->rel
= V_SQ_REL_RELATIVE
;
801 if (inst
->Instruction
.Saturate
) {
806 static unsigned tgsi_chan(const struct tgsi_full_src_register
*tgsi_src
, unsigned swizzle
)
810 return tgsi_src
->Register
.SwizzleX
;
812 return tgsi_src
->Register
.SwizzleY
;
814 return tgsi_src
->Register
.SwizzleZ
;
816 return tgsi_src
->Register
.SwizzleW
;
822 static int tgsi_split_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
824 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
825 struct r600_bc_alu alu
;
826 int i
, j
, k
, nconst
, r
;
828 for (i
= 0, nconst
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
829 if (inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
832 tgsi_src(ctx
, &inst
->Src
[i
], &r600_src
[i
]);
834 for (i
= 0, j
= nconst
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
835 if (j
> 0 && inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
836 int treg
= r600_get_temp(ctx
);
837 for (k
= 0; k
< 4; k
++) {
838 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
839 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
840 alu
.src
[0].sel
= r600_src
[i
].sel
;
842 alu
.src
[0].rel
= r600_src
[i
].rel
;
848 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
852 r600_src
[i
].sel
= treg
;
860 /* need to move any immediate into a temp - for trig functions which use literal for PI stuff */
861 static int tgsi_split_literal_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
863 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
864 struct r600_bc_alu alu
;
865 int i
, j
, k
, nliteral
, r
;
867 for (i
= 0, nliteral
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
868 if (r600_src
[i
].sel
== V_SQ_ALU_SRC_LITERAL
) {
872 for (i
= 0, j
= nliteral
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
873 if (j
> 0 && r600_src
[i
].sel
== V_SQ_ALU_SRC_LITERAL
) {
874 int treg
= r600_get_temp(ctx
);
875 for (k
= 0; k
< 4; k
++) {
876 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
877 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
878 alu
.src
[0].sel
= r600_src
[i
].sel
;
880 alu
.src
[0].value
= r600_src
[i
].value
;
886 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
890 r600_src
[i
].sel
= treg
;
897 static int tgsi_last_instruction(unsigned writemask
)
901 for (i
= 0; i
< 4; i
++) {
902 if (writemask
& (1 << i
)) {
909 static int tgsi_op2_s(struct r600_shader_ctx
*ctx
, int swap
)
911 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
912 struct r600_bc_alu_src r600_src
[3];
913 struct r600_bc_alu alu
;
915 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
917 r
= tgsi_split_constant(ctx
, r600_src
);
920 r
= tgsi_split_literal_constant(ctx
, r600_src
);
923 for (i
= 0; i
< lasti
+ 1; i
++) {
924 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
927 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
928 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
930 alu
.inst
= ctx
->inst_info
->r600_opcode
;
932 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
933 alu
.src
[j
] = r600_src
[j
];
934 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
937 alu
.src
[0] = r600_src
[1];
938 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], i
);
940 alu
.src
[1] = r600_src
[0];
941 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
943 /* handle some special cases */
944 switch (ctx
->inst_info
->tgsi_opcode
) {
945 case TGSI_OPCODE_SUB
:
948 case TGSI_OPCODE_ABS
:
957 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
964 static int tgsi_op2(struct r600_shader_ctx
*ctx
)
966 return tgsi_op2_s(ctx
, 0);
969 static int tgsi_op2_swap(struct r600_shader_ctx
*ctx
)
971 return tgsi_op2_s(ctx
, 1);
975 * r600 - trunc to -PI..PI range
976 * r700 - normalize by dividing by 2PI
979 static int tgsi_setup_trig(struct r600_shader_ctx
*ctx
,
980 struct r600_bc_alu_src r600_src
[3])
982 static float half_inv_pi
= 1.0 /(3.1415926535 * 2);
983 static float double_pi
= 3.1415926535 * 2;
984 static float neg_pi
= -3.1415926535;
986 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
988 struct r600_bc_alu alu
;
990 r
= tgsi_split_constant(ctx
, r600_src
);
993 r
= tgsi_split_literal_constant(ctx
, r600_src
);
997 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
998 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1002 alu
.dst
.sel
= ctx
->temp_reg
;
1005 alu
.src
[0] = r600_src
[0];
1006 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1008 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
1009 alu
.src
[1].chan
= 0;
1010 alu
.src
[1].value
= (uint32_t *)&half_inv_pi
;
1011 alu
.src
[2].sel
= V_SQ_ALU_SRC_0_5
;
1012 alu
.src
[2].chan
= 0;
1014 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1018 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1019 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
);
1022 alu
.dst
.sel
= ctx
->temp_reg
;
1025 alu
.src
[0].sel
= ctx
->temp_reg
;
1026 alu
.src
[0].chan
= 0;
1028 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1032 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1033 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1037 alu
.dst
.sel
= ctx
->temp_reg
;
1040 alu
.src
[0].sel
= ctx
->temp_reg
;
1041 alu
.src
[0].chan
= 0;
1043 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
1044 alu
.src
[1].chan
= 0;
1045 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1046 alu
.src
[2].chan
= 0;
1048 if (ctx
->bc
->chiprev
== CHIPREV_R600
) {
1049 alu
.src
[1].value
= (uint32_t *)&double_pi
;
1050 alu
.src
[2].value
= (uint32_t *)&neg_pi
;
1052 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1053 alu
.src
[2].sel
= V_SQ_ALU_SRC_0_5
;
1058 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1064 static int tgsi_trig(struct r600_shader_ctx
*ctx
)
1066 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1067 struct r600_bc_alu_src r600_src
[3];
1068 struct r600_bc_alu alu
;
1070 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1072 r
= tgsi_setup_trig(ctx
, r600_src
);
1076 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1077 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1079 alu
.dst
.sel
= ctx
->temp_reg
;
1082 alu
.src
[0].sel
= ctx
->temp_reg
;
1083 alu
.src
[0].chan
= 0;
1085 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1089 /* replicate result */
1090 for (i
= 0; i
< lasti
+ 1; i
++) {
1091 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1094 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1095 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1097 alu
.src
[0].sel
= ctx
->temp_reg
;
1098 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1101 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1108 static int tgsi_scs(struct r600_shader_ctx
*ctx
)
1110 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1111 struct r600_bc_alu_src r600_src
[3];
1112 struct r600_bc_alu alu
;
1115 /* We'll only need the trig stuff if we are going to write to the
1116 * X or Y components of the destination vector.
1118 if (likely(inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_XY
)) {
1119 r
= tgsi_setup_trig(ctx
, r600_src
);
1125 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_X
) {
1126 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1127 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
);
1128 tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
1130 alu
.src
[0].sel
= ctx
->temp_reg
;
1131 alu
.src
[0].chan
= 0;
1133 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1139 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_Y
) {
1140 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1141 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
);
1142 tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
1144 alu
.src
[0].sel
= ctx
->temp_reg
;
1145 alu
.src
[0].chan
= 0;
1147 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1153 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_Z
) {
1154 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1156 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1158 tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1160 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1161 alu
.src
[0].chan
= 0;
1165 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1171 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_W
) {
1172 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1174 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1176 tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
1178 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1179 alu
.src
[0].chan
= 0;
1183 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1191 static int tgsi_kill(struct r600_shader_ctx
*ctx
)
1193 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1194 struct r600_bc_alu alu
;
1197 for (i
= 0; i
< 4; i
++) {
1198 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1199 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1203 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1205 if (ctx
->inst_info
->tgsi_opcode
== TGSI_OPCODE_KILP
) {
1206 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1209 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1210 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1215 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1220 /* kill must be last in ALU */
1221 ctx
->bc
->force_add_cf
= 1;
1222 ctx
->shader
->uses_kill
= TRUE
;
1226 static int tgsi_lit(struct r600_shader_ctx
*ctx
)
1228 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1229 struct r600_bc_alu alu
;
1230 struct r600_bc_alu_src r600_src
[3];
1233 r
= tgsi_split_constant(ctx
, r600_src
);
1236 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1241 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1242 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1243 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
; /*1.0*/
1244 alu
.src
[0].chan
= 0;
1245 tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
1246 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 0) & 1;
1247 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1251 /* dst.y = max(src.x, 0.0) */
1252 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1253 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
);
1254 alu
.src
[0] = r600_src
[0];
1255 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
; /*0.0*/
1256 alu
.src
[1].chan
= 0;
1257 tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
1258 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 1) & 1;
1259 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1264 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1265 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1266 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1267 alu
.src
[0].chan
= 0;
1268 tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
1269 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 3) & 1;
1271 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1275 if (inst
->Dst
[0].Register
.WriteMask
& (1 << 2))
1280 /* dst.z = log(src.y) */
1281 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1282 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED
);
1283 alu
.src
[0] = r600_src
[0];
1284 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
1285 tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1287 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1291 chan
= alu
.dst
.chan
;
1294 /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */
1295 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1296 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT
);
1297 alu
.src
[0] = r600_src
[0];
1298 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1299 alu
.src
[1].sel
= sel
;
1300 alu
.src
[1].chan
= chan
;
1302 alu
.src
[2] = r600_src
[0];
1303 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], 0);
1304 alu
.dst
.sel
= ctx
->temp_reg
;
1309 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1313 /* dst.z = exp(tmp.x) */
1314 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1315 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
1316 alu
.src
[0].sel
= ctx
->temp_reg
;
1317 alu
.src
[0].chan
= 0;
1318 tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1320 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1327 static int tgsi_rsq(struct r600_shader_ctx
*ctx
)
1329 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1330 struct r600_bc_alu alu
;
1333 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1336 * For state trackers other than OpenGL, we'll want to use
1337 * _RECIPSQRT_IEEE instead.
1339 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_CLAMPED
);
1341 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1342 tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1343 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1346 alu
.dst
.sel
= ctx
->temp_reg
;
1349 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1352 /* replicate result */
1353 return tgsi_helper_tempx_replicate(ctx
);
1356 static int tgsi_helper_tempx_replicate(struct r600_shader_ctx
*ctx
)
1358 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1359 struct r600_bc_alu alu
;
1362 for (i
= 0; i
< 4; i
++) {
1363 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1364 alu
.src
[0].sel
= ctx
->temp_reg
;
1365 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1367 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1368 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
1371 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1378 static int tgsi_trans_srcx_replicate(struct r600_shader_ctx
*ctx
)
1380 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1381 struct r600_bc_alu alu
;
1384 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1385 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1386 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1387 tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1388 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1390 alu
.dst
.sel
= ctx
->temp_reg
;
1393 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1396 /* replicate result */
1397 return tgsi_helper_tempx_replicate(ctx
);
1400 static int tgsi_pow(struct r600_shader_ctx
*ctx
)
1402 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1403 struct r600_bc_alu alu
;
1407 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1408 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
1409 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1410 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1411 alu
.dst
.sel
= ctx
->temp_reg
;
1414 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1418 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1419 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1420 tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[0]);
1421 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], 0);
1422 alu
.src
[1].sel
= ctx
->temp_reg
;
1423 alu
.dst
.sel
= ctx
->temp_reg
;
1426 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1429 /* POW(a,b) = EXP2(b * LOG2(a))*/
1430 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1431 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
1432 alu
.src
[0].sel
= ctx
->temp_reg
;
1433 alu
.dst
.sel
= ctx
->temp_reg
;
1436 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1439 return tgsi_helper_tempx_replicate(ctx
);
1442 static int tgsi_ssg(struct r600_shader_ctx
*ctx
)
1444 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1445 struct r600_bc_alu alu
;
1446 struct r600_bc_alu_src r600_src
[3];
1449 r
= tgsi_split_constant(ctx
, r600_src
);
1452 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1456 /* tmp = (src > 0 ? 1 : src) */
1457 for (i
= 0; i
< 4; i
++) {
1458 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1459 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
);
1462 alu
.dst
.sel
= ctx
->temp_reg
;
1465 alu
.src
[0] = r600_src
[0];
1466 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1468 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1470 alu
.src
[2] = r600_src
[0];
1471 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], i
);
1474 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1479 /* dst = (-tmp > 0 ? -1 : tmp) */
1480 for (i
= 0; i
< 4; i
++) {
1481 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1482 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
);
1484 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1486 alu
.src
[0].sel
= ctx
->temp_reg
;
1487 alu
.src
[0].chan
= i
;
1490 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1493 alu
.src
[2].sel
= ctx
->temp_reg
;
1494 alu
.src
[2].chan
= i
;
1498 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1505 static int tgsi_helper_copy(struct r600_shader_ctx
*ctx
, struct tgsi_full_instruction
*inst
)
1507 struct r600_bc_alu alu
;
1510 for (i
= 0; i
< 4; i
++) {
1511 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1512 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
))) {
1513 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
);
1516 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1517 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1518 alu
.src
[0].sel
= ctx
->temp_reg
;
1519 alu
.src
[0].chan
= i
;
1524 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1531 static int tgsi_op3(struct r600_shader_ctx
*ctx
)
1533 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1534 struct r600_bc_alu_src r600_src
[3];
1535 struct r600_bc_alu alu
;
1537 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1539 r
= tgsi_split_constant(ctx
, r600_src
);
1542 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1545 for (i
= 0; i
< lasti
+ 1; i
++) {
1546 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1549 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1550 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1551 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1552 alu
.src
[j
] = r600_src
[j
];
1553 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1556 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1563 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1570 static int tgsi_dp(struct r600_shader_ctx
*ctx
)
1572 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1573 struct r600_bc_alu_src r600_src
[3];
1574 struct r600_bc_alu alu
;
1577 r
= tgsi_split_constant(ctx
, r600_src
);
1580 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1583 for (i
= 0; i
< 4; i
++) {
1584 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1585 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1586 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1587 alu
.src
[j
] = r600_src
[j
];
1588 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1591 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1593 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
1594 /* handle some special cases */
1595 switch (ctx
->inst_info
->tgsi_opcode
) {
1596 case TGSI_OPCODE_DP2
:
1598 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1599 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1602 case TGSI_OPCODE_DP3
:
1604 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1605 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1608 case TGSI_OPCODE_DPH
:
1610 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1611 alu
.src
[0].chan
= 0;
1621 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1628 static int tgsi_tex(struct r600_shader_ctx
*ctx
)
1630 static float one_point_five
= 1.5f
;
1631 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1632 struct r600_bc_tex tex
;
1633 struct r600_bc_alu alu
;
1637 boolean src_not_temp
=
1638 inst
->Src
[0].Register
.File
!= TGSI_FILE_TEMPORARY
&&
1639 inst
->Src
[0].Register
.File
!= TGSI_FILE_INPUT
;
1641 src_gpr
= ctx
->file_offset
[inst
->Src
[0].Register
.File
] + inst
->Src
[0].Register
.Index
;
1643 if (inst
->Instruction
.Opcode
== TGSI_OPCODE_TXP
) {
1644 /* Add perspective divide */
1645 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1646 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
1647 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1649 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1650 alu
.dst
.sel
= ctx
->temp_reg
;
1654 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1658 for (i
= 0; i
< 3; i
++) {
1659 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1660 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1661 alu
.src
[0].sel
= ctx
->temp_reg
;
1662 alu
.src
[0].chan
= 3;
1663 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1664 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1665 alu
.dst
.sel
= ctx
->temp_reg
;
1668 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1672 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1673 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1674 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1675 alu
.src
[0].chan
= 0;
1676 alu
.dst
.sel
= ctx
->temp_reg
;
1680 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1683 src_not_temp
= FALSE
;
1684 src_gpr
= ctx
->temp_reg
;
1687 if (inst
->Texture
.Texture
== TGSI_TEXTURE_CUBE
) {
1688 int src_chan
, src2_chan
;
1690 /* tmp1.xyzw = CUBE(R0.zzxy, R0.yxzz) */
1691 for (i
= 0; i
< 4; i
++) {
1692 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1693 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE
);
1717 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1718 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], src_chan
);
1719 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1720 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], src2_chan
);
1721 alu
.dst
.sel
= ctx
->temp_reg
;
1726 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1731 /* tmp1.z = RCP_e(|tmp1.z|) */
1732 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1733 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
1734 alu
.src
[0].sel
= ctx
->temp_reg
;
1735 alu
.src
[0].chan
= 2;
1737 alu
.dst
.sel
= ctx
->temp_reg
;
1741 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1745 /* MULADD R0.x, R0.x, PS1, (0x3FC00000, 1.5f).x
1746 * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x
1747 * muladd has no writemask, have to use another temp
1749 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1750 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1753 alu
.src
[0].sel
= ctx
->temp_reg
;
1754 alu
.src
[0].chan
= 0;
1755 alu
.src
[1].sel
= ctx
->temp_reg
;
1756 alu
.src
[1].chan
= 2;
1758 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1759 alu
.src
[2].chan
= 0;
1760 alu
.src
[2].value
= (u32
*)&one_point_five
;
1762 alu
.dst
.sel
= ctx
->temp_reg
;
1766 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1770 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1771 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1774 alu
.src
[0].sel
= ctx
->temp_reg
;
1775 alu
.src
[0].chan
= 1;
1776 alu
.src
[1].sel
= ctx
->temp_reg
;
1777 alu
.src
[1].chan
= 2;
1779 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1780 alu
.src
[2].chan
= 0;
1781 alu
.src
[2].value
= (u32
*)&one_point_five
;
1783 alu
.dst
.sel
= ctx
->temp_reg
;
1788 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1792 src_not_temp
= FALSE
;
1793 src_gpr
= ctx
->temp_reg
;
1797 for (i
= 0; i
< 4; i
++) {
1798 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1799 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1800 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1801 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1802 alu
.dst
.sel
= ctx
->temp_reg
;
1807 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1811 src_gpr
= ctx
->temp_reg
;
1814 opcode
= ctx
->inst_info
->r600_opcode
;
1815 if (opcode
== SQ_TEX_INST_SAMPLE
&&
1816 (inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW1D
|| inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW2D
))
1817 opcode
= SQ_TEX_INST_SAMPLE_C
;
1819 memset(&tex
, 0, sizeof(struct r600_bc_tex
));
1821 tex
.sampler_id
= ctx
->file_offset
[inst
->Src
[1].Register
.File
] + inst
->Src
[1].Register
.Index
;
1822 tex
.resource_id
= tex
.sampler_id
;
1823 tex
.src_gpr
= src_gpr
;
1824 tex
.dst_gpr
= ctx
->file_offset
[inst
->Dst
[0].Register
.File
] + inst
->Dst
[0].Register
.Index
;
1825 tex
.dst_sel_x
= (inst
->Dst
[0].Register
.WriteMask
& 1) ? 0 : 7;
1826 tex
.dst_sel_y
= (inst
->Dst
[0].Register
.WriteMask
& 2) ? 1 : 7;
1827 tex
.dst_sel_z
= (inst
->Dst
[0].Register
.WriteMask
& 4) ? 2 : 7;
1828 tex
.dst_sel_w
= (inst
->Dst
[0].Register
.WriteMask
& 8) ? 3 : 7;
1834 if (inst
->Texture
.Texture
== TGSI_TEXTURE_CUBE
) {
1841 if (inst
->Texture
.Texture
!= TGSI_TEXTURE_RECT
) {
1842 tex
.coord_type_x
= 1;
1843 tex
.coord_type_y
= 1;
1844 tex
.coord_type_z
= 1;
1845 tex
.coord_type_w
= 1;
1848 if (inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW1D
|| inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW2D
)
1851 r
= r600_bc_add_tex(ctx
->bc
, &tex
);
1855 /* add shadow ambient support - gallium doesn't do it yet */
1859 static int tgsi_lrp(struct r600_shader_ctx
*ctx
)
1861 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1862 struct r600_bc_alu_src r600_src
[3];
1863 struct r600_bc_alu alu
;
1864 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1868 r
= tgsi_split_constant(ctx
, r600_src
);
1871 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1875 /* optimize if it's just an equal balance */
1876 if(r600_src
[0].sel
== V_SQ_ALU_SRC_0_5
) {
1877 for (i
= 0; i
< lasti
+ 1; i
++) {
1878 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1881 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1882 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
);
1883 alu
.src
[0] = r600_src
[1];
1884 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], i
);
1885 alu
.src
[1] = r600_src
[2];
1886 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1888 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1893 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1901 for (i
= 0; i
< lasti
+ 1; i
++) {
1902 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1905 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1906 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
);
1907 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1908 alu
.src
[0].chan
= 0;
1909 alu
.src
[1] = r600_src
[0];
1910 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1912 alu
.dst
.sel
= ctx
->temp_reg
;
1918 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1923 /* (1 - src0) * src2 */
1924 for (i
= 0; i
< lasti
+ 1; i
++) {
1925 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1928 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1929 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1930 alu
.src
[0].sel
= ctx
->temp_reg
;
1931 alu
.src
[0].chan
= i
;
1932 alu
.src
[1] = r600_src
[2];
1933 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1934 alu
.dst
.sel
= ctx
->temp_reg
;
1940 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1945 /* src0 * src1 + (1 - src0) * src2 */
1946 for (i
= 0; i
< lasti
+ 1; i
++) {
1947 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1950 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1951 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1953 alu
.src
[0] = r600_src
[0];
1954 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1955 alu
.src
[1] = r600_src
[1];
1956 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
1957 alu
.src
[2].sel
= ctx
->temp_reg
;
1958 alu
.src
[2].chan
= i
;
1960 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1965 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1972 static int tgsi_cmp(struct r600_shader_ctx
*ctx
)
1974 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1975 struct r600_bc_alu_src r600_src
[3];
1976 struct r600_bc_alu alu
;
1978 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1980 r
= tgsi_split_constant(ctx
, r600_src
);
1983 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1987 for (i
= 0; i
< lasti
+ 1; i
++) {
1988 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1991 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1992 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE
);
1993 alu
.src
[0] = r600_src
[0];
1994 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1996 alu
.src
[1] = r600_src
[2];
1997 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1999 alu
.src
[2] = r600_src
[1];
2000 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[1], i
);
2002 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2008 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2015 static int tgsi_xpd(struct r600_shader_ctx
*ctx
)
2017 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2018 struct r600_bc_alu_src r600_src
[3];
2019 struct r600_bc_alu alu
;
2020 uint32_t use_temp
= 0;
2023 if (inst
->Dst
[0].Register
.WriteMask
!= 0xf)
2026 r
= tgsi_split_constant(ctx
, r600_src
);
2029 r
= tgsi_split_literal_constant(ctx
, r600_src
);
2033 for (i
= 0; i
< 4; i
++) {
2034 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2035 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2037 alu
.src
[0] = r600_src
[0];
2040 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
2043 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2046 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
2049 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
2050 alu
.src
[0].chan
= i
;
2053 alu
.src
[1] = r600_src
[1];
2056 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
2059 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
2062 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
2065 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2066 alu
.src
[1].chan
= i
;
2069 alu
.dst
.sel
= ctx
->temp_reg
;
2075 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2080 for (i
= 0; i
< 4; i
++) {
2081 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2082 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
2084 alu
.src
[0] = r600_src
[0];
2087 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
2090 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
2093 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2096 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
2097 alu
.src
[0].chan
= i
;
2100 alu
.src
[1] = r600_src
[1];
2103 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
2106 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
2109 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
2112 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2113 alu
.src
[1].chan
= i
;
2116 alu
.src
[2].sel
= ctx
->temp_reg
;
2118 alu
.src
[2].chan
= i
;
2121 alu
.dst
.sel
= ctx
->temp_reg
;
2123 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2129 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2134 return tgsi_helper_copy(ctx
, inst
);
2138 static int tgsi_exp(struct r600_shader_ctx
*ctx
)
2140 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2141 struct r600_bc_alu_src r600_src
[3] = { { 0 } };
2142 struct r600_bc_alu alu
;
2145 /* result.x = 2^floor(src); */
2146 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
2147 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2149 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2150 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2152 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2154 alu
.dst
.sel
= ctx
->temp_reg
;
2158 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2162 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2163 alu
.src
[0].sel
= ctx
->temp_reg
;
2164 alu
.src
[0].chan
= 0;
2166 alu
.dst
.sel
= ctx
->temp_reg
;
2170 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2175 /* result.y = tmp - floor(tmp); */
2176 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
2177 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2179 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
);
2180 alu
.src
[0] = r600_src
[0];
2181 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2182 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2184 alu
.dst
.sel
= ctx
->temp_reg
;
2185 // r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
2193 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2198 /* result.z = RoughApprox2ToX(tmp);*/
2199 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 0x1) {
2200 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2201 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2202 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2203 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2205 alu
.dst
.sel
= ctx
->temp_reg
;
2211 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2216 /* result.w = 1.0;*/
2217 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 0x1) {
2218 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2220 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2221 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2222 alu
.src
[0].chan
= 0;
2224 alu
.dst
.sel
= ctx
->temp_reg
;
2228 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2232 return tgsi_helper_copy(ctx
, inst
);
2235 static int tgsi_log(struct r600_shader_ctx
*ctx
)
2237 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2238 struct r600_bc_alu alu
;
2241 /* result.x = floor(log2(src)); */
2242 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
2243 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2245 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2246 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2248 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2250 alu
.dst
.sel
= ctx
->temp_reg
;
2254 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2258 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2259 alu
.src
[0].sel
= ctx
->temp_reg
;
2260 alu
.src
[0].chan
= 0;
2262 alu
.dst
.sel
= ctx
->temp_reg
;
2267 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2272 /* result.y = src.x / (2 ^ floor(log2(src.x))); */
2273 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
2274 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2276 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2277 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2279 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2281 alu
.dst
.sel
= ctx
->temp_reg
;
2286 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2290 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2292 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2293 alu
.src
[0].sel
= ctx
->temp_reg
;
2294 alu
.src
[0].chan
= 1;
2296 alu
.dst
.sel
= ctx
->temp_reg
;
2301 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2305 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2307 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2308 alu
.src
[0].sel
= ctx
->temp_reg
;
2309 alu
.src
[0].chan
= 1;
2311 alu
.dst
.sel
= ctx
->temp_reg
;
2316 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2320 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2322 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
2323 alu
.src
[0].sel
= ctx
->temp_reg
;
2324 alu
.src
[0].chan
= 1;
2326 alu
.dst
.sel
= ctx
->temp_reg
;
2331 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2335 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2337 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2339 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2340 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2342 alu
.src
[1].sel
= ctx
->temp_reg
;
2343 alu
.src
[1].chan
= 1;
2345 alu
.dst
.sel
= ctx
->temp_reg
;
2350 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2355 /* result.z = log2(src);*/
2356 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 1) {
2357 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2359 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2360 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2361 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2363 alu
.dst
.sel
= ctx
->temp_reg
;
2368 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2373 /* result.w = 1.0; */
2374 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 1) {
2375 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2377 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2378 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2379 alu
.src
[0].chan
= 0;
2381 alu
.dst
.sel
= ctx
->temp_reg
;
2386 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2391 return tgsi_helper_copy(ctx
, inst
);
2394 static int tgsi_eg_arl(struct r600_shader_ctx
*ctx
)
2396 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2397 struct r600_bc_alu alu
;
2399 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2401 switch (inst
->Instruction
.Opcode
) {
2402 case TGSI_OPCODE_ARL
:
2403 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT_FLOOR
;
2405 case TGSI_OPCODE_ARR
:
2406 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT
;
2413 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2414 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2417 alu
.dst
.sel
= ctx
->temp_reg
;
2419 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2422 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2423 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT
;
2424 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2425 alu
.src
[0].sel
= ctx
->temp_reg
;
2426 alu
.src
[0].chan
= 0;
2428 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2433 static int tgsi_r600_arl(struct r600_shader_ctx
*ctx
)
2435 /* TODO from r600c, ar values don't persist between clauses */
2436 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2437 struct r600_bc_alu alu
;
2439 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2441 switch (inst
->Instruction
.Opcode
) {
2442 case TGSI_OPCODE_ARL
:
2443 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_FLOOR
;
2445 case TGSI_OPCODE_ARR
:
2446 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA
;
2454 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2455 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2459 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2462 ctx
->bc
->cf_last
->r6xx_uses_waterfall
= 1;
2466 static int tgsi_opdst(struct r600_shader_ctx
*ctx
)
2468 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2469 struct r600_bc_alu alu
;
2472 for (i
= 0; i
< 4; i
++) {
2473 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2475 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2476 tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2478 if (i
== 0 || i
== 3) {
2479 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2481 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2482 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2485 if (i
== 0 || i
== 2) {
2486 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
2488 tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[1]);
2489 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
2493 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2500 static int emit_logic_pred(struct r600_shader_ctx
*ctx
, int opcode
)
2502 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2503 struct r600_bc_alu alu
;
2506 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2510 alu
.dst
.sel
= ctx
->temp_reg
;
2514 tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2515 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2516 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2517 alu
.src
[1].chan
= 0;
2521 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE
));
2527 static int pops(struct r600_shader_ctx
*ctx
, int pops
)
2530 if (ctx
->bc
->cf_last
) {
2531 if (ctx
->bc
->cf_last
->inst
== CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU
) << 3)
2533 else if (ctx
->bc
->cf_last
->inst
== CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER
) << 3)
2538 ctx
->bc
->cf_last
->inst
= CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER
) << 3;
2539 ctx
->bc
->force_add_cf
= 1;
2540 } else if (alu_pop
== 2) {
2541 ctx
->bc
->cf_last
->inst
= CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP2_AFTER
) << 3;
2542 ctx
->bc
->force_add_cf
= 1;
2544 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_POP
));
2545 ctx
->bc
->cf_last
->pop_count
= pops
;
2546 ctx
->bc
->cf_last
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2551 static inline void callstack_decrease_current(struct r600_shader_ctx
*ctx
, unsigned reason
)
2555 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
--;
2559 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
-= 4;
2562 /* TOODO : for 16 vp asic should -= 2; */
2563 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
--;
2568 static inline void callstack_check_depth(struct r600_shader_ctx
*ctx
, unsigned reason
, unsigned check_max_only
)
2570 if (check_max_only
) {
2583 if ((ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+ diff
) >
2584 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
) {
2585 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
=
2586 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+ diff
;
2592 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
++;
2596 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+= 4;
2599 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
++;
2603 if ((ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
) >
2604 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
) {
2605 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
=
2606 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
;
2610 static void fc_set_mid(struct r600_shader_ctx
*ctx
, int fc_sp
)
2612 struct r600_cf_stack_entry
*sp
= &ctx
->bc
->fc_stack
[fc_sp
];
2614 sp
->mid
= (struct r600_bc_cf
**)realloc((void *)sp
->mid
,
2615 sizeof(struct r600_bc_cf
*) * (sp
->num_mid
+ 1));
2616 sp
->mid
[sp
->num_mid
] = ctx
->bc
->cf_last
;
2620 static void fc_pushlevel(struct r600_shader_ctx
*ctx
, int type
)
2623 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
= type
;
2624 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
= ctx
->bc
->cf_last
;
2627 static void fc_poplevel(struct r600_shader_ctx
*ctx
)
2629 struct r600_cf_stack_entry
*sp
= &ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
];
2641 static int emit_return(struct r600_shader_ctx
*ctx
)
2643 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_RETURN
);
2647 static int emit_jump_to_offset(struct r600_shader_ctx
*ctx
, int pops
, int offset
)
2650 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_JUMP
);
2651 ctx
->bc
->cf_last
->pop_count
= pops
;
2652 /* TODO work out offset */
2656 static int emit_setret_in_loop_flag(struct r600_shader_ctx
*ctx
, unsigned flag_value
)
2661 static void emit_testflag(struct r600_shader_ctx
*ctx
)
2666 static void emit_return_on_flag(struct r600_shader_ctx
*ctx
, unsigned ifidx
)
2669 emit_jump_to_offset(ctx
, 1, 4);
2670 emit_setret_in_loop_flag(ctx
, V_SQ_ALU_SRC_0
);
2671 pops(ctx
, ifidx
+ 1);
2675 static void break_loop_on_flag(struct r600_shader_ctx
*ctx
, unsigned fc_sp
)
2679 r600_bc_add_cfinst(ctx
->bc
, ctx
->inst_info
->r600_opcode
);
2680 ctx
->bc
->cf_last
->pop_count
= 1;
2682 fc_set_mid(ctx
, fc_sp
);
2688 static int tgsi_if(struct r600_shader_ctx
*ctx
)
2690 emit_logic_pred(ctx
, CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE
));
2692 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_JUMP
));
2694 fc_pushlevel(ctx
, FC_IF
);
2696 callstack_check_depth(ctx
, FC_PUSH_VPM
, 0);
2700 static int tgsi_else(struct r600_shader_ctx
*ctx
)
2702 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_ELSE
));
2703 ctx
->bc
->cf_last
->pop_count
= 1;
2705 fc_set_mid(ctx
, ctx
->bc
->fc_sp
);
2706 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
;
2710 static int tgsi_endif(struct r600_shader_ctx
*ctx
)
2713 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_IF
) {
2714 R600_ERR("if/endif unbalanced in shader\n");
2718 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
== NULL
) {
2719 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2720 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->pop_count
= 1;
2722 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
[0]->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2726 callstack_decrease_current(ctx
, FC_PUSH_VPM
);
2730 static int tgsi_bgnloop(struct r600_shader_ctx
*ctx
)
2732 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL
));
2734 fc_pushlevel(ctx
, FC_LOOP
);
2736 /* check stack depth */
2737 callstack_check_depth(ctx
, FC_LOOP
, 0);
2741 static int tgsi_endloop(struct r600_shader_ctx
*ctx
)
2745 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END
));
2747 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_LOOP
) {
2748 R600_ERR("loop/endloop in shader code are not paired.\n");
2752 /* fixup loop pointers - from r600isa
2753 LOOP END points to CF after LOOP START,
2754 LOOP START point to CF after LOOP END
2755 BRK/CONT point to LOOP END CF
2757 ctx
->bc
->cf_last
->cf_addr
= ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->id
+ 2;
2759 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2761 for (i
= 0; i
< ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].num_mid
; i
++) {
2762 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
[i
]->cf_addr
= ctx
->bc
->cf_last
->id
;
2764 /* TODO add LOOPRET support */
2766 callstack_decrease_current(ctx
, FC_LOOP
);
2770 static int tgsi_loop_brk_cont(struct r600_shader_ctx
*ctx
)
2774 for (fscp
= ctx
->bc
->fc_sp
; fscp
> 0; fscp
--)
2776 if (FC_LOOP
== ctx
->bc
->fc_stack
[fscp
].type
)
2781 R600_ERR("Break not inside loop/endloop pair\n");
2785 r600_bc_add_cfinst(ctx
->bc
, ctx
->inst_info
->r600_opcode
);
2786 ctx
->bc
->cf_last
->pop_count
= 1;
2788 fc_set_mid(ctx
, fscp
);
2791 callstack_check_depth(ctx
, FC_PUSH_VPM
, 1);
2795 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[] = {
2796 {TGSI_OPCODE_ARL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_r600_arl
},
2797 {TGSI_OPCODE_MOV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
2798 {TGSI_OPCODE_LIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
2801 * For state trackers other than OpenGL, we'll want to use
2802 * _RECIP_IEEE instead.
2804 {TGSI_OPCODE_RCP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_CLAMPED
, tgsi_trans_srcx_replicate
},
2806 {TGSI_OPCODE_RSQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_rsq
},
2807 {TGSI_OPCODE_EXP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
2808 {TGSI_OPCODE_LOG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_log
},
2809 {TGSI_OPCODE_MUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
2810 {TGSI_OPCODE_ADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
2811 {TGSI_OPCODE_DP3
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2812 {TGSI_OPCODE_DP4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2813 {TGSI_OPCODE_DST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_opdst
},
2814 {TGSI_OPCODE_MIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
2815 {TGSI_OPCODE_MAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
2816 {TGSI_OPCODE_SLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
2817 {TGSI_OPCODE_SGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
2818 {TGSI_OPCODE_MAD
, 1, V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
2819 {TGSI_OPCODE_SUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
2820 {TGSI_OPCODE_LRP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
2821 {TGSI_OPCODE_CND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2823 {20, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2824 {TGSI_OPCODE_DP2A
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2826 {22, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2827 {23, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2828 {TGSI_OPCODE_FRC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
2829 {TGSI_OPCODE_CLAMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2830 {TGSI_OPCODE_FLR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
2831 {TGSI_OPCODE_ROUND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2832 {TGSI_OPCODE_EX2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
2833 {TGSI_OPCODE_LG2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
2834 {TGSI_OPCODE_POW
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
2835 {TGSI_OPCODE_XPD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
2837 {32, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2838 {TGSI_OPCODE_ABS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
2839 {TGSI_OPCODE_RCC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2840 {TGSI_OPCODE_DPH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2841 {TGSI_OPCODE_COS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
, tgsi_trig
},
2842 {TGSI_OPCODE_DDX
, 0, SQ_TEX_INST_GET_GRADIENTS_H
, tgsi_tex
},
2843 {TGSI_OPCODE_DDY
, 0, SQ_TEX_INST_GET_GRADIENTS_V
, tgsi_tex
},
2844 {TGSI_OPCODE_KILP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* predicated kill */
2845 {TGSI_OPCODE_PK2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2846 {TGSI_OPCODE_PK2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2847 {TGSI_OPCODE_PK4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2848 {TGSI_OPCODE_PK4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2849 {TGSI_OPCODE_RFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2850 {TGSI_OPCODE_SEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
2851 {TGSI_OPCODE_SFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2852 {TGSI_OPCODE_SGT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
2853 {TGSI_OPCODE_SIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
2854 {TGSI_OPCODE_SLE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
2855 {TGSI_OPCODE_SNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
2856 {TGSI_OPCODE_STR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2857 {TGSI_OPCODE_TEX
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
2858 {TGSI_OPCODE_TXD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2859 {TGSI_OPCODE_TXP
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
2860 {TGSI_OPCODE_UP2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2861 {TGSI_OPCODE_UP2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2862 {TGSI_OPCODE_UP4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2863 {TGSI_OPCODE_UP4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2864 {TGSI_OPCODE_X2D
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2865 {TGSI_OPCODE_ARA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2866 {TGSI_OPCODE_ARR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_r600_arl
},
2867 {TGSI_OPCODE_BRA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2868 {TGSI_OPCODE_CAL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2869 {TGSI_OPCODE_RET
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2870 {TGSI_OPCODE_SSG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
2871 {TGSI_OPCODE_CMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
2872 {TGSI_OPCODE_SCS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_scs
},
2873 {TGSI_OPCODE_TXB
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
2874 {TGSI_OPCODE_NRM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2875 {TGSI_OPCODE_DIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2876 {TGSI_OPCODE_DP2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2877 {TGSI_OPCODE_TXL
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
2878 {TGSI_OPCODE_BRK
, 0, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK
, tgsi_loop_brk_cont
},
2879 {TGSI_OPCODE_IF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
2881 {75, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2882 {76, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2883 {TGSI_OPCODE_ELSE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
2884 {TGSI_OPCODE_ENDIF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
2886 {79, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2887 {80, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2888 {TGSI_OPCODE_PUSHA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2889 {TGSI_OPCODE_POPA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2890 {TGSI_OPCODE_CEIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2891 {TGSI_OPCODE_I2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2892 {TGSI_OPCODE_NOT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2893 {TGSI_OPCODE_TRUNC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
2894 {TGSI_OPCODE_SHL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2896 {88, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2897 {TGSI_OPCODE_AND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2898 {TGSI_OPCODE_OR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2899 {TGSI_OPCODE_MOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2900 {TGSI_OPCODE_XOR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2901 {TGSI_OPCODE_SAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2902 {TGSI_OPCODE_TXF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2903 {TGSI_OPCODE_TXQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2904 {TGSI_OPCODE_CONT
, 0, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE
, tgsi_loop_brk_cont
},
2905 {TGSI_OPCODE_EMIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2906 {TGSI_OPCODE_ENDPRIM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2907 {TGSI_OPCODE_BGNLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_bgnloop
},
2908 {TGSI_OPCODE_BGNSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2909 {TGSI_OPCODE_ENDLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endloop
},
2910 {TGSI_OPCODE_ENDSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2912 {103, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2913 {104, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2914 {105, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2915 {106, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2916 {TGSI_OPCODE_NOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2918 {108, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2919 {109, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2920 {110, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2921 {111, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2922 {TGSI_OPCODE_NRM4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2923 {TGSI_OPCODE_CALLNZ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2924 {TGSI_OPCODE_IFC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2925 {TGSI_OPCODE_BREAKC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2926 {TGSI_OPCODE_KIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
2927 {TGSI_OPCODE_END
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
2929 {118, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2930 {TGSI_OPCODE_F2I
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2931 {TGSI_OPCODE_IDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2932 {TGSI_OPCODE_IMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2933 {TGSI_OPCODE_IMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2934 {TGSI_OPCODE_INEG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2935 {TGSI_OPCODE_ISGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2936 {TGSI_OPCODE_ISHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2937 {TGSI_OPCODE_ISLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2938 {TGSI_OPCODE_F2U
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2939 {TGSI_OPCODE_U2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2940 {TGSI_OPCODE_UADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2941 {TGSI_OPCODE_UDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2942 {TGSI_OPCODE_UMAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2943 {TGSI_OPCODE_UMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2944 {TGSI_OPCODE_UMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2945 {TGSI_OPCODE_UMOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2946 {TGSI_OPCODE_UMUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2947 {TGSI_OPCODE_USEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2948 {TGSI_OPCODE_USGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2949 {TGSI_OPCODE_USHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2950 {TGSI_OPCODE_USLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2951 {TGSI_OPCODE_USNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2952 {TGSI_OPCODE_SWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2953 {TGSI_OPCODE_CASE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2954 {TGSI_OPCODE_DEFAULT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2955 {TGSI_OPCODE_ENDSWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2956 {TGSI_OPCODE_LAST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2959 static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction
[] = {
2960 {TGSI_OPCODE_ARL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_eg_arl
},
2961 {TGSI_OPCODE_MOV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
2962 {TGSI_OPCODE_LIT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
2963 {TGSI_OPCODE_RCP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
, tgsi_trans_srcx_replicate
},
2964 {TGSI_OPCODE_RSQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE
, tgsi_trans_srcx_replicate
},
2965 {TGSI_OPCODE_EXP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
2966 {TGSI_OPCODE_LOG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2967 {TGSI_OPCODE_MUL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
2968 {TGSI_OPCODE_ADD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
2969 {TGSI_OPCODE_DP3
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2970 {TGSI_OPCODE_DP4
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2971 {TGSI_OPCODE_DST
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_opdst
},
2972 {TGSI_OPCODE_MIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
2973 {TGSI_OPCODE_MAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
2974 {TGSI_OPCODE_SLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
2975 {TGSI_OPCODE_SGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
2976 {TGSI_OPCODE_MAD
, 1, EG_V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
2977 {TGSI_OPCODE_SUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
2978 {TGSI_OPCODE_LRP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
2979 {TGSI_OPCODE_CND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2981 {20, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2982 {TGSI_OPCODE_DP2A
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2984 {22, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2985 {23, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2986 {TGSI_OPCODE_FRC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
2987 {TGSI_OPCODE_CLAMP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2988 {TGSI_OPCODE_FLR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
2989 {TGSI_OPCODE_ROUND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2990 {TGSI_OPCODE_EX2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
2991 {TGSI_OPCODE_LG2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
2992 {TGSI_OPCODE_POW
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
2993 {TGSI_OPCODE_XPD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
2995 {32, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2996 {TGSI_OPCODE_ABS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
2997 {TGSI_OPCODE_RCC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2998 {TGSI_OPCODE_DPH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2999 {TGSI_OPCODE_COS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
, tgsi_trig
},
3000 {TGSI_OPCODE_DDX
, 0, SQ_TEX_INST_GET_GRADIENTS_H
, tgsi_tex
},
3001 {TGSI_OPCODE_DDY
, 0, SQ_TEX_INST_GET_GRADIENTS_V
, tgsi_tex
},
3002 {TGSI_OPCODE_KILP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* predicated kill */
3003 {TGSI_OPCODE_PK2H
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3004 {TGSI_OPCODE_PK2US
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3005 {TGSI_OPCODE_PK4B
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3006 {TGSI_OPCODE_PK4UB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3007 {TGSI_OPCODE_RFL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3008 {TGSI_OPCODE_SEQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
3009 {TGSI_OPCODE_SFL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3010 {TGSI_OPCODE_SGT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
3011 {TGSI_OPCODE_SIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
3012 {TGSI_OPCODE_SLE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
3013 {TGSI_OPCODE_SNE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
3014 {TGSI_OPCODE_STR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3015 {TGSI_OPCODE_TEX
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
3016 {TGSI_OPCODE_TXD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3017 {TGSI_OPCODE_TXP
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
3018 {TGSI_OPCODE_UP2H
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3019 {TGSI_OPCODE_UP2US
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3020 {TGSI_OPCODE_UP4B
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3021 {TGSI_OPCODE_UP4UB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3022 {TGSI_OPCODE_X2D
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3023 {TGSI_OPCODE_ARA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3024 {TGSI_OPCODE_ARR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_eg_arl
},
3025 {TGSI_OPCODE_BRA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3026 {TGSI_OPCODE_CAL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3027 {TGSI_OPCODE_RET
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3028 {TGSI_OPCODE_SSG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
3029 {TGSI_OPCODE_CMP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
3030 {TGSI_OPCODE_SCS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_scs
},
3031 {TGSI_OPCODE_TXB
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
3032 {TGSI_OPCODE_NRM
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3033 {TGSI_OPCODE_DIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3034 {TGSI_OPCODE_DP2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3035 {TGSI_OPCODE_TXL
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
3036 {TGSI_OPCODE_BRK
, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK
, tgsi_loop_brk_cont
},
3037 {TGSI_OPCODE_IF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
3039 {75, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3040 {76, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3041 {TGSI_OPCODE_ELSE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
3042 {TGSI_OPCODE_ENDIF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
3044 {79, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3045 {80, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3046 {TGSI_OPCODE_PUSHA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3047 {TGSI_OPCODE_POPA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3048 {TGSI_OPCODE_CEIL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3049 {TGSI_OPCODE_I2F
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3050 {TGSI_OPCODE_NOT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3051 {TGSI_OPCODE_TRUNC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
3052 {TGSI_OPCODE_SHL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3054 {88, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3055 {TGSI_OPCODE_AND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3056 {TGSI_OPCODE_OR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3057 {TGSI_OPCODE_MOD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3058 {TGSI_OPCODE_XOR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3059 {TGSI_OPCODE_SAD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3060 {TGSI_OPCODE_TXF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3061 {TGSI_OPCODE_TXQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3062 {TGSI_OPCODE_CONT
, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE
, tgsi_loop_brk_cont
},
3063 {TGSI_OPCODE_EMIT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3064 {TGSI_OPCODE_ENDPRIM
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3065 {TGSI_OPCODE_BGNLOOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_bgnloop
},
3066 {TGSI_OPCODE_BGNSUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3067 {TGSI_OPCODE_ENDLOOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endloop
},
3068 {TGSI_OPCODE_ENDSUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3070 {103, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3071 {104, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3072 {105, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3073 {106, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3074 {TGSI_OPCODE_NOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3076 {108, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3077 {109, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3078 {110, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3079 {111, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3080 {TGSI_OPCODE_NRM4
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3081 {TGSI_OPCODE_CALLNZ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3082 {TGSI_OPCODE_IFC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3083 {TGSI_OPCODE_BREAKC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3084 {TGSI_OPCODE_KIL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
3085 {TGSI_OPCODE_END
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
3087 {118, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3088 {TGSI_OPCODE_F2I
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3089 {TGSI_OPCODE_IDIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3090 {TGSI_OPCODE_IMAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3091 {TGSI_OPCODE_IMIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3092 {TGSI_OPCODE_INEG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3093 {TGSI_OPCODE_ISGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3094 {TGSI_OPCODE_ISHR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3095 {TGSI_OPCODE_ISLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3096 {TGSI_OPCODE_F2U
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3097 {TGSI_OPCODE_U2F
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3098 {TGSI_OPCODE_UADD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3099 {TGSI_OPCODE_UDIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3100 {TGSI_OPCODE_UMAD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3101 {TGSI_OPCODE_UMAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3102 {TGSI_OPCODE_UMIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3103 {TGSI_OPCODE_UMOD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3104 {TGSI_OPCODE_UMUL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3105 {TGSI_OPCODE_USEQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3106 {TGSI_OPCODE_USGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3107 {TGSI_OPCODE_USHR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3108 {TGSI_OPCODE_USLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3109 {TGSI_OPCODE_USNE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3110 {TGSI_OPCODE_SWITCH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3111 {TGSI_OPCODE_CASE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3112 {TGSI_OPCODE_DEFAULT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3113 {TGSI_OPCODE_ENDSWITCH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3114 {TGSI_OPCODE_LAST
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},