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 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 int 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
];
789 static int tgsi_dst(struct r600_shader_ctx
*ctx
,
790 const struct tgsi_full_dst_register
*tgsi_dst
,
792 struct r600_bc_alu_dst
*r600_dst
)
794 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
796 r600_dst
->sel
= tgsi_dst
->Register
.Index
;
797 r600_dst
->sel
+= ctx
->file_offset
[tgsi_dst
->Register
.File
];
798 r600_dst
->chan
= swizzle
;
800 if (tgsi_dst
->Register
.Indirect
)
801 r600_dst
->rel
= V_SQ_REL_RELATIVE
;
802 if (inst
->Instruction
.Saturate
) {
808 static unsigned tgsi_chan(const struct tgsi_full_src_register
*tgsi_src
, unsigned swizzle
)
812 return tgsi_src
->Register
.SwizzleX
;
814 return tgsi_src
->Register
.SwizzleY
;
816 return tgsi_src
->Register
.SwizzleZ
;
818 return tgsi_src
->Register
.SwizzleW
;
824 static int tgsi_split_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
826 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
827 struct r600_bc_alu alu
;
828 int i
, j
, k
, nconst
, r
;
830 for (i
= 0, nconst
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
831 if (inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
834 r
= tgsi_src(ctx
, &inst
->Src
[i
], &r600_src
[i
]);
839 for (i
= 0, j
= nconst
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
840 if (j
> 0 && inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
841 int treg
= r600_get_temp(ctx
);
842 for (k
= 0; k
< 4; k
++) {
843 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
844 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
845 alu
.src
[0].sel
= r600_src
[i
].sel
;
847 alu
.src
[0].rel
= r600_src
[i
].rel
;
853 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
857 r600_src
[i
].sel
= treg
;
865 /* need to move any immediate into a temp - for trig functions which use literal for PI stuff */
866 static int tgsi_split_literal_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
868 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
869 struct r600_bc_alu alu
;
870 int i
, j
, k
, nliteral
, r
;
872 for (i
= 0, nliteral
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
873 if (r600_src
[i
].sel
== V_SQ_ALU_SRC_LITERAL
) {
877 for (i
= 0, j
= nliteral
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
878 if (j
> 0 && r600_src
[i
].sel
== V_SQ_ALU_SRC_LITERAL
) {
879 int treg
= r600_get_temp(ctx
);
880 for (k
= 0; k
< 4; k
++) {
881 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
882 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
883 alu
.src
[0].sel
= r600_src
[i
].sel
;
885 alu
.src
[0].value
= r600_src
[i
].value
;
891 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
895 r600_src
[i
].sel
= treg
;
902 static int tgsi_last_instruction(unsigned writemask
)
906 for (i
= 0; i
< 4; i
++) {
907 if (writemask
& (1 << i
)) {
914 static int tgsi_op2_s(struct r600_shader_ctx
*ctx
, int swap
)
916 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
917 struct r600_bc_alu_src r600_src
[3];
918 struct r600_bc_alu alu
;
920 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
922 r
= tgsi_split_constant(ctx
, r600_src
);
925 r
= tgsi_split_literal_constant(ctx
, r600_src
);
928 for (i
= 0; i
< lasti
+ 1; i
++) {
929 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
932 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
933 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
937 alu
.inst
= ctx
->inst_info
->r600_opcode
;
939 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
940 alu
.src
[j
] = r600_src
[j
];
941 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
944 alu
.src
[0] = r600_src
[1];
945 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], i
);
947 alu
.src
[1] = r600_src
[0];
948 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
950 /* handle some special cases */
951 switch (ctx
->inst_info
->tgsi_opcode
) {
952 case TGSI_OPCODE_SUB
:
955 case TGSI_OPCODE_ABS
:
964 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
971 static int tgsi_op2(struct r600_shader_ctx
*ctx
)
973 return tgsi_op2_s(ctx
, 0);
976 static int tgsi_op2_swap(struct r600_shader_ctx
*ctx
)
978 return tgsi_op2_s(ctx
, 1);
982 * r600 - trunc to -PI..PI range
983 * r700 - normalize by dividing by 2PI
986 static int tgsi_setup_trig(struct r600_shader_ctx
*ctx
,
987 struct r600_bc_alu_src r600_src
[3])
989 static float half_inv_pi
= 1.0 /(3.1415926535 * 2);
990 static float double_pi
= 3.1415926535 * 2;
991 static float neg_pi
= -3.1415926535;
993 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
995 struct r600_bc_alu alu
;
997 r
= tgsi_split_constant(ctx
, r600_src
);
1000 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1004 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1005 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1009 alu
.dst
.sel
= ctx
->temp_reg
;
1012 alu
.src
[0] = r600_src
[0];
1013 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1015 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
1016 alu
.src
[1].chan
= 0;
1017 alu
.src
[1].value
= (uint32_t *)&half_inv_pi
;
1018 alu
.src
[2].sel
= V_SQ_ALU_SRC_0_5
;
1019 alu
.src
[2].chan
= 0;
1021 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1025 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1026 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
);
1029 alu
.dst
.sel
= ctx
->temp_reg
;
1032 alu
.src
[0].sel
= ctx
->temp_reg
;
1033 alu
.src
[0].chan
= 0;
1035 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1039 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1040 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1044 alu
.dst
.sel
= ctx
->temp_reg
;
1047 alu
.src
[0].sel
= ctx
->temp_reg
;
1048 alu
.src
[0].chan
= 0;
1050 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
1051 alu
.src
[1].chan
= 0;
1052 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1053 alu
.src
[2].chan
= 0;
1055 if (ctx
->bc
->chiprev
== CHIPREV_R600
) {
1056 alu
.src
[1].value
= (uint32_t *)&double_pi
;
1057 alu
.src
[2].value
= (uint32_t *)&neg_pi
;
1059 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1060 alu
.src
[2].sel
= V_SQ_ALU_SRC_0_5
;
1065 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1071 static int tgsi_trig(struct r600_shader_ctx
*ctx
)
1073 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1074 struct r600_bc_alu_src r600_src
[3];
1075 struct r600_bc_alu alu
;
1077 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1079 r
= tgsi_setup_trig(ctx
, r600_src
);
1083 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1084 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1086 alu
.dst
.sel
= ctx
->temp_reg
;
1089 alu
.src
[0].sel
= ctx
->temp_reg
;
1090 alu
.src
[0].chan
= 0;
1092 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1096 /* replicate result */
1097 for (i
= 0; i
< lasti
+ 1; i
++) {
1098 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1101 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1102 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1104 alu
.src
[0].sel
= ctx
->temp_reg
;
1105 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1110 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1117 static int tgsi_scs(struct r600_shader_ctx
*ctx
)
1119 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1120 struct r600_bc_alu_src r600_src
[3];
1121 struct r600_bc_alu alu
;
1124 /* We'll only need the trig stuff if we are going to write to the
1125 * X or Y components of the destination vector.
1127 if (likely(inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_XY
)) {
1128 r
= tgsi_setup_trig(ctx
, r600_src
);
1134 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_X
) {
1135 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1136 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
);
1137 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
1141 alu
.src
[0].sel
= ctx
->temp_reg
;
1142 alu
.src
[0].chan
= 0;
1144 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1150 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_Y
) {
1151 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1152 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
);
1153 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
1157 alu
.src
[0].sel
= ctx
->temp_reg
;
1158 alu
.src
[0].chan
= 0;
1160 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1166 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_Z
) {
1167 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1169 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1171 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1175 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1176 alu
.src
[0].chan
= 0;
1180 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1186 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_W
) {
1187 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1189 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1191 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
1195 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1196 alu
.src
[0].chan
= 0;
1200 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1208 static int tgsi_kill(struct r600_shader_ctx
*ctx
)
1210 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1211 struct r600_bc_alu alu
;
1214 for (i
= 0; i
< 4; i
++) {
1215 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1216 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1220 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1222 if (ctx
->inst_info
->tgsi_opcode
== TGSI_OPCODE_KILP
) {
1223 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1226 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1229 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1234 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1239 /* kill must be last in ALU */
1240 ctx
->bc
->force_add_cf
= 1;
1241 ctx
->shader
->uses_kill
= TRUE
;
1245 static int tgsi_lit(struct r600_shader_ctx
*ctx
)
1247 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1248 struct r600_bc_alu alu
;
1249 struct r600_bc_alu_src r600_src
[3];
1252 r
= tgsi_split_constant(ctx
, r600_src
);
1255 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1260 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1261 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1262 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
; /*1.0*/
1263 alu
.src
[0].chan
= 0;
1264 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
1267 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 0) & 1;
1268 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1272 /* dst.y = max(src.x, 0.0) */
1273 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1274 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
);
1275 alu
.src
[0] = r600_src
[0];
1276 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
; /*0.0*/
1277 alu
.src
[1].chan
= 0;
1278 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
1281 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 1) & 1;
1282 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1287 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1288 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1289 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1290 alu
.src
[0].chan
= 0;
1291 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
1294 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 3) & 1;
1296 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1300 if (inst
->Dst
[0].Register
.WriteMask
& (1 << 2))
1305 /* dst.z = log(src.y) */
1306 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1307 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED
);
1308 alu
.src
[0] = r600_src
[0];
1309 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
1310 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1314 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1318 chan
= alu
.dst
.chan
;
1321 /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */
1322 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1323 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT
);
1324 alu
.src
[0] = r600_src
[0];
1325 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1326 alu
.src
[1].sel
= sel
;
1327 alu
.src
[1].chan
= chan
;
1329 alu
.src
[2] = r600_src
[0];
1330 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], 0);
1331 alu
.dst
.sel
= ctx
->temp_reg
;
1336 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1340 /* dst.z = exp(tmp.x) */
1341 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1342 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
1343 alu
.src
[0].sel
= ctx
->temp_reg
;
1344 alu
.src
[0].chan
= 0;
1345 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1349 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1356 static int tgsi_rsq(struct r600_shader_ctx
*ctx
)
1358 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1359 struct r600_bc_alu alu
;
1362 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1365 * For state trackers other than OpenGL, we'll want to use
1366 * _RECIPSQRT_IEEE instead.
1368 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_CLAMPED
);
1370 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1371 r
= tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1374 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1377 alu
.dst
.sel
= ctx
->temp_reg
;
1380 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1383 /* replicate result */
1384 return tgsi_helper_tempx_replicate(ctx
);
1387 static int tgsi_helper_tempx_replicate(struct r600_shader_ctx
*ctx
)
1389 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1390 struct r600_bc_alu alu
;
1393 for (i
= 0; i
< 4; i
++) {
1394 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1395 alu
.src
[0].sel
= ctx
->temp_reg
;
1396 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1398 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1401 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
1404 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1411 static int tgsi_trans_srcx_replicate(struct r600_shader_ctx
*ctx
)
1413 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1414 struct r600_bc_alu alu
;
1417 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1418 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1419 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1420 r
= tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1423 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1425 alu
.dst
.sel
= ctx
->temp_reg
;
1428 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1431 /* replicate result */
1432 return tgsi_helper_tempx_replicate(ctx
);
1435 static int tgsi_pow(struct r600_shader_ctx
*ctx
)
1437 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1438 struct r600_bc_alu alu
;
1442 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1443 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
1444 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1447 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1448 alu
.dst
.sel
= ctx
->temp_reg
;
1451 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1455 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1456 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1457 r
= tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[0]);
1460 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], 0);
1461 alu
.src
[1].sel
= ctx
->temp_reg
;
1462 alu
.dst
.sel
= ctx
->temp_reg
;
1465 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1468 /* POW(a,b) = EXP2(b * LOG2(a))*/
1469 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1470 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
1471 alu
.src
[0].sel
= ctx
->temp_reg
;
1472 alu
.dst
.sel
= ctx
->temp_reg
;
1475 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1478 return tgsi_helper_tempx_replicate(ctx
);
1481 static int tgsi_ssg(struct r600_shader_ctx
*ctx
)
1483 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1484 struct r600_bc_alu alu
;
1485 struct r600_bc_alu_src r600_src
[3];
1488 r
= tgsi_split_constant(ctx
, r600_src
);
1491 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1495 /* tmp = (src > 0 ? 1 : src) */
1496 for (i
= 0; i
< 4; i
++) {
1497 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1498 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
);
1501 alu
.dst
.sel
= ctx
->temp_reg
;
1504 alu
.src
[0] = r600_src
[0];
1505 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1507 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1509 alu
.src
[2] = r600_src
[0];
1510 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], i
);
1513 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1518 /* dst = (-tmp > 0 ? -1 : tmp) */
1519 for (i
= 0; i
< 4; i
++) {
1520 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1521 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
);
1523 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1527 alu
.src
[0].sel
= ctx
->temp_reg
;
1528 alu
.src
[0].chan
= i
;
1531 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1534 alu
.src
[2].sel
= ctx
->temp_reg
;
1535 alu
.src
[2].chan
= i
;
1539 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1546 static int tgsi_helper_copy(struct r600_shader_ctx
*ctx
, struct tgsi_full_instruction
*inst
)
1548 struct r600_bc_alu alu
;
1551 for (i
= 0; i
< 4; i
++) {
1552 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1553 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
))) {
1554 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
);
1557 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1558 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1561 alu
.src
[0].sel
= ctx
->temp_reg
;
1562 alu
.src
[0].chan
= i
;
1567 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1574 static int tgsi_op3(struct r600_shader_ctx
*ctx
)
1576 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1577 struct r600_bc_alu_src r600_src
[3];
1578 struct r600_bc_alu alu
;
1580 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1582 r
= tgsi_split_constant(ctx
, r600_src
);
1585 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1588 for (i
= 0; i
< lasti
+ 1; i
++) {
1589 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1592 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1593 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1594 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1595 alu
.src
[j
] = r600_src
[j
];
1596 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1599 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1609 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1616 static int tgsi_dp(struct r600_shader_ctx
*ctx
)
1618 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1619 struct r600_bc_alu_src r600_src
[3];
1620 struct r600_bc_alu alu
;
1623 r
= tgsi_split_constant(ctx
, r600_src
);
1626 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1629 for (i
= 0; i
< 4; i
++) {
1630 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1631 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1632 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1633 alu
.src
[j
] = r600_src
[j
];
1634 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1637 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1642 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
1643 /* handle some special cases */
1644 switch (ctx
->inst_info
->tgsi_opcode
) {
1645 case TGSI_OPCODE_DP2
:
1647 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1648 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1651 case TGSI_OPCODE_DP3
:
1653 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1654 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1657 case TGSI_OPCODE_DPH
:
1659 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1660 alu
.src
[0].chan
= 0;
1670 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1677 static int tgsi_tex(struct r600_shader_ctx
*ctx
)
1679 static float one_point_five
= 1.5f
;
1680 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1681 struct r600_bc_tex tex
;
1682 struct r600_bc_alu alu
;
1686 boolean src_not_temp
=
1687 inst
->Src
[0].Register
.File
!= TGSI_FILE_TEMPORARY
&&
1688 inst
->Src
[0].Register
.File
!= TGSI_FILE_INPUT
;
1690 src_gpr
= ctx
->file_offset
[inst
->Src
[0].Register
.File
] + inst
->Src
[0].Register
.Index
;
1692 if (inst
->Instruction
.Opcode
== TGSI_OPCODE_TXP
) {
1693 /* Add perspective divide */
1694 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1695 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
1696 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1700 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1701 alu
.dst
.sel
= ctx
->temp_reg
;
1705 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1709 for (i
= 0; i
< 3; i
++) {
1710 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1711 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1712 alu
.src
[0].sel
= ctx
->temp_reg
;
1713 alu
.src
[0].chan
= 3;
1714 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1717 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1718 alu
.dst
.sel
= ctx
->temp_reg
;
1721 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1725 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1726 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1727 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1728 alu
.src
[0].chan
= 0;
1729 alu
.dst
.sel
= ctx
->temp_reg
;
1733 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1736 src_not_temp
= FALSE
;
1737 src_gpr
= ctx
->temp_reg
;
1740 if (inst
->Texture
.Texture
== TGSI_TEXTURE_CUBE
) {
1741 int src_chan
, src2_chan
;
1743 /* tmp1.xyzw = CUBE(R0.zzxy, R0.yxzz) */
1744 for (i
= 0; i
< 4; i
++) {
1745 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1746 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE
);
1770 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1773 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], src_chan
);
1774 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1777 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], src2_chan
);
1778 alu
.dst
.sel
= ctx
->temp_reg
;
1783 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1788 /* tmp1.z = RCP_e(|tmp1.z|) */
1789 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1790 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
1791 alu
.src
[0].sel
= ctx
->temp_reg
;
1792 alu
.src
[0].chan
= 2;
1794 alu
.dst
.sel
= ctx
->temp_reg
;
1798 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1802 /* MULADD R0.x, R0.x, PS1, (0x3FC00000, 1.5f).x
1803 * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x
1804 * muladd has no writemask, have to use another temp
1806 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1807 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1810 alu
.src
[0].sel
= ctx
->temp_reg
;
1811 alu
.src
[0].chan
= 0;
1812 alu
.src
[1].sel
= ctx
->temp_reg
;
1813 alu
.src
[1].chan
= 2;
1815 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1816 alu
.src
[2].chan
= 0;
1817 alu
.src
[2].value
= (u32
*)&one_point_five
;
1819 alu
.dst
.sel
= ctx
->temp_reg
;
1823 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1827 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1828 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1831 alu
.src
[0].sel
= ctx
->temp_reg
;
1832 alu
.src
[0].chan
= 1;
1833 alu
.src
[1].sel
= ctx
->temp_reg
;
1834 alu
.src
[1].chan
= 2;
1836 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1837 alu
.src
[2].chan
= 0;
1838 alu
.src
[2].value
= (u32
*)&one_point_five
;
1840 alu
.dst
.sel
= ctx
->temp_reg
;
1845 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1849 src_not_temp
= FALSE
;
1850 src_gpr
= ctx
->temp_reg
;
1854 for (i
= 0; i
< 4; i
++) {
1855 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1856 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1857 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1860 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1861 alu
.dst
.sel
= ctx
->temp_reg
;
1866 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1870 src_gpr
= ctx
->temp_reg
;
1873 opcode
= ctx
->inst_info
->r600_opcode
;
1874 if (opcode
== SQ_TEX_INST_SAMPLE
&&
1875 (inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW1D
|| inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW2D
))
1876 opcode
= SQ_TEX_INST_SAMPLE_C
;
1878 memset(&tex
, 0, sizeof(struct r600_bc_tex
));
1880 tex
.sampler_id
= ctx
->file_offset
[inst
->Src
[1].Register
.File
] + inst
->Src
[1].Register
.Index
;
1881 tex
.resource_id
= tex
.sampler_id
;
1882 tex
.src_gpr
= src_gpr
;
1883 tex
.dst_gpr
= ctx
->file_offset
[inst
->Dst
[0].Register
.File
] + inst
->Dst
[0].Register
.Index
;
1884 tex
.dst_sel_x
= (inst
->Dst
[0].Register
.WriteMask
& 1) ? 0 : 7;
1885 tex
.dst_sel_y
= (inst
->Dst
[0].Register
.WriteMask
& 2) ? 1 : 7;
1886 tex
.dst_sel_z
= (inst
->Dst
[0].Register
.WriteMask
& 4) ? 2 : 7;
1887 tex
.dst_sel_w
= (inst
->Dst
[0].Register
.WriteMask
& 8) ? 3 : 7;
1893 if (inst
->Texture
.Texture
== TGSI_TEXTURE_CUBE
) {
1900 if (inst
->Texture
.Texture
!= TGSI_TEXTURE_RECT
) {
1901 tex
.coord_type_x
= 1;
1902 tex
.coord_type_y
= 1;
1903 tex
.coord_type_z
= 1;
1904 tex
.coord_type_w
= 1;
1907 if (inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW1D
|| inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW2D
)
1910 r
= r600_bc_add_tex(ctx
->bc
, &tex
);
1914 /* add shadow ambient support - gallium doesn't do it yet */
1918 static int tgsi_lrp(struct r600_shader_ctx
*ctx
)
1920 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1921 struct r600_bc_alu_src r600_src
[3];
1922 struct r600_bc_alu alu
;
1923 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
1927 r
= tgsi_split_constant(ctx
, r600_src
);
1930 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1934 /* optimize if it's just an equal balance */
1935 if(r600_src
[0].sel
== V_SQ_ALU_SRC_0_5
) {
1936 for (i
= 0; i
< lasti
+ 1; i
++) {
1937 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1940 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1941 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
);
1942 alu
.src
[0] = r600_src
[1];
1943 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], i
);
1944 alu
.src
[1] = r600_src
[2];
1945 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1947 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1955 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1963 for (i
= 0; i
< lasti
+ 1; i
++) {
1964 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1967 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1968 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
);
1969 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1970 alu
.src
[0].chan
= 0;
1971 alu
.src
[1] = r600_src
[0];
1972 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1974 alu
.dst
.sel
= ctx
->temp_reg
;
1980 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1985 /* (1 - src0) * src2 */
1986 for (i
= 0; i
< lasti
+ 1; i
++) {
1987 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1990 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1991 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1992 alu
.src
[0].sel
= ctx
->temp_reg
;
1993 alu
.src
[0].chan
= i
;
1994 alu
.src
[1] = r600_src
[2];
1995 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1996 alu
.dst
.sel
= ctx
->temp_reg
;
2002 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2007 /* src0 * src1 + (1 - src0) * src2 */
2008 for (i
= 0; i
< lasti
+ 1; i
++) {
2009 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
2012 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2013 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
2015 alu
.src
[0] = r600_src
[0];
2016 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2017 alu
.src
[1] = r600_src
[1];
2018 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
2019 alu
.src
[2].sel
= ctx
->temp_reg
;
2020 alu
.src
[2].chan
= i
;
2022 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2030 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2037 static int tgsi_cmp(struct r600_shader_ctx
*ctx
)
2039 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2040 struct r600_bc_alu_src r600_src
[3];
2041 struct r600_bc_alu alu
;
2043 int lasti
= tgsi_last_instruction(inst
->Dst
[0].Register
.WriteMask
);
2045 r
= tgsi_split_constant(ctx
, r600_src
);
2048 r
= tgsi_split_literal_constant(ctx
, r600_src
);
2052 for (i
= 0; i
< lasti
+ 1; i
++) {
2053 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
2056 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2057 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE
);
2058 alu
.src
[0] = r600_src
[0];
2059 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2061 alu
.src
[1] = r600_src
[2];
2062 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
2064 alu
.src
[2] = r600_src
[1];
2065 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[1], i
);
2067 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2076 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2083 static int tgsi_xpd(struct r600_shader_ctx
*ctx
)
2085 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2086 struct r600_bc_alu_src r600_src
[3];
2087 struct r600_bc_alu alu
;
2088 uint32_t use_temp
= 0;
2091 if (inst
->Dst
[0].Register
.WriteMask
!= 0xf)
2094 r
= tgsi_split_constant(ctx
, r600_src
);
2097 r
= tgsi_split_literal_constant(ctx
, r600_src
);
2101 for (i
= 0; i
< 4; i
++) {
2102 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2103 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2105 alu
.src
[0] = r600_src
[0];
2108 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
2111 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2114 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
2117 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
2118 alu
.src
[0].chan
= i
;
2121 alu
.src
[1] = r600_src
[1];
2124 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
2127 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
2130 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
2133 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2134 alu
.src
[1].chan
= i
;
2137 alu
.dst
.sel
= ctx
->temp_reg
;
2143 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2148 for (i
= 0; i
< 4; i
++) {
2149 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2150 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
2152 alu
.src
[0] = r600_src
[0];
2155 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
2158 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
2161 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2164 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
2165 alu
.src
[0].chan
= i
;
2168 alu
.src
[1] = r600_src
[1];
2171 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
2174 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
2177 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
2180 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2181 alu
.src
[1].chan
= i
;
2184 alu
.src
[2].sel
= ctx
->temp_reg
;
2186 alu
.src
[2].chan
= i
;
2189 alu
.dst
.sel
= ctx
->temp_reg
;
2191 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2200 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2205 return tgsi_helper_copy(ctx
, inst
);
2209 static int tgsi_exp(struct r600_shader_ctx
*ctx
)
2211 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2212 struct r600_bc_alu_src r600_src
[3] = { { 0 } };
2213 struct r600_bc_alu alu
;
2216 /* result.x = 2^floor(src); */
2217 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
2218 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2220 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2221 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2225 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2227 alu
.dst
.sel
= ctx
->temp_reg
;
2231 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2235 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2236 alu
.src
[0].sel
= ctx
->temp_reg
;
2237 alu
.src
[0].chan
= 0;
2239 alu
.dst
.sel
= ctx
->temp_reg
;
2243 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2248 /* result.y = tmp - floor(tmp); */
2249 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
2250 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2252 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
);
2253 alu
.src
[0] = r600_src
[0];
2254 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2257 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2259 alu
.dst
.sel
= ctx
->temp_reg
;
2260 // r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
2268 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2273 /* result.z = RoughApprox2ToX(tmp);*/
2274 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 0x1) {
2275 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2276 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2277 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2280 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2282 alu
.dst
.sel
= ctx
->temp_reg
;
2288 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2293 /* result.w = 1.0;*/
2294 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 0x1) {
2295 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2297 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2298 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2299 alu
.src
[0].chan
= 0;
2301 alu
.dst
.sel
= ctx
->temp_reg
;
2305 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2309 return tgsi_helper_copy(ctx
, inst
);
2312 static int tgsi_log(struct r600_shader_ctx
*ctx
)
2314 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2315 struct r600_bc_alu alu
;
2318 /* result.x = floor(log2(src)); */
2319 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
2320 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2322 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2323 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2327 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2329 alu
.dst
.sel
= ctx
->temp_reg
;
2333 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2337 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2338 alu
.src
[0].sel
= ctx
->temp_reg
;
2339 alu
.src
[0].chan
= 0;
2341 alu
.dst
.sel
= ctx
->temp_reg
;
2346 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2351 /* result.y = src.x / (2 ^ floor(log2(src.x))); */
2352 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
2353 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2355 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2356 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2360 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2362 alu
.dst
.sel
= ctx
->temp_reg
;
2367 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2371 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2373 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2374 alu
.src
[0].sel
= ctx
->temp_reg
;
2375 alu
.src
[0].chan
= 1;
2377 alu
.dst
.sel
= ctx
->temp_reg
;
2382 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2386 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2388 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2389 alu
.src
[0].sel
= ctx
->temp_reg
;
2390 alu
.src
[0].chan
= 1;
2392 alu
.dst
.sel
= ctx
->temp_reg
;
2397 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2401 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2403 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
2404 alu
.src
[0].sel
= ctx
->temp_reg
;
2405 alu
.src
[0].chan
= 1;
2407 alu
.dst
.sel
= ctx
->temp_reg
;
2412 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2416 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2418 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2420 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2424 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2426 alu
.src
[1].sel
= ctx
->temp_reg
;
2427 alu
.src
[1].chan
= 1;
2429 alu
.dst
.sel
= ctx
->temp_reg
;
2434 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2439 /* result.z = log2(src);*/
2440 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 1) {
2441 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2443 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2444 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2448 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2450 alu
.dst
.sel
= ctx
->temp_reg
;
2455 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2460 /* result.w = 1.0; */
2461 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 1) {
2462 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2464 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2465 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2466 alu
.src
[0].chan
= 0;
2468 alu
.dst
.sel
= ctx
->temp_reg
;
2473 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2478 return tgsi_helper_copy(ctx
, inst
);
2481 static int tgsi_eg_arl(struct r600_shader_ctx
*ctx
)
2483 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2484 struct r600_bc_alu alu
;
2486 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2488 switch (inst
->Instruction
.Opcode
) {
2489 case TGSI_OPCODE_ARL
:
2490 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT_FLOOR
;
2492 case TGSI_OPCODE_ARR
:
2493 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT
;
2500 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2503 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2506 alu
.dst
.sel
= ctx
->temp_reg
;
2508 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2511 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2512 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT
;
2513 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2516 alu
.src
[0].sel
= ctx
->temp_reg
;
2517 alu
.src
[0].chan
= 0;
2519 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2524 static int tgsi_r600_arl(struct r600_shader_ctx
*ctx
)
2526 /* TODO from r600c, ar values don't persist between clauses */
2527 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2528 struct r600_bc_alu alu
;
2530 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2532 switch (inst
->Instruction
.Opcode
) {
2533 case TGSI_OPCODE_ARL
:
2534 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_FLOOR
;
2536 case TGSI_OPCODE_ARR
:
2537 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA
;
2545 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2548 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2552 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2555 ctx
->bc
->cf_last
->r6xx_uses_waterfall
= 1;
2559 static int tgsi_opdst(struct r600_shader_ctx
*ctx
)
2561 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2562 struct r600_bc_alu alu
;
2565 for (i
= 0; i
< 4; i
++) {
2566 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2568 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2569 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2573 if (i
== 0 || i
== 3) {
2574 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2576 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2579 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2582 if (i
== 0 || i
== 2) {
2583 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
2585 r
= tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[1]);
2588 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
2592 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2599 static int emit_logic_pred(struct r600_shader_ctx
*ctx
, int opcode
)
2601 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2602 struct r600_bc_alu alu
;
2605 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2609 alu
.dst
.sel
= ctx
->temp_reg
;
2613 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2616 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2617 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2618 alu
.src
[1].chan
= 0;
2622 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE
));
2628 static int pops(struct r600_shader_ctx
*ctx
, int pops
)
2631 if (ctx
->bc
->cf_last
) {
2632 if (ctx
->bc
->cf_last
->inst
== CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU
) << 3)
2634 else if (ctx
->bc
->cf_last
->inst
== CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER
) << 3)
2639 ctx
->bc
->cf_last
->inst
= CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP_AFTER
) << 3;
2640 ctx
->bc
->force_add_cf
= 1;
2641 } else if (alu_pop
== 2) {
2642 ctx
->bc
->cf_last
->inst
= CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_POP2_AFTER
) << 3;
2643 ctx
->bc
->force_add_cf
= 1;
2645 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_POP
));
2646 ctx
->bc
->cf_last
->pop_count
= pops
;
2647 ctx
->bc
->cf_last
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2652 static inline void callstack_decrease_current(struct r600_shader_ctx
*ctx
, unsigned reason
)
2656 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
--;
2660 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
-= 4;
2663 /* TOODO : for 16 vp asic should -= 2; */
2664 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
--;
2669 static inline void callstack_check_depth(struct r600_shader_ctx
*ctx
, unsigned reason
, unsigned check_max_only
)
2671 if (check_max_only
) {
2684 if ((ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+ diff
) >
2685 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
) {
2686 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
=
2687 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+ diff
;
2693 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
++;
2697 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+= 4;
2700 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
++;
2704 if ((ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
) >
2705 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
) {
2706 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
=
2707 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
;
2711 static void fc_set_mid(struct r600_shader_ctx
*ctx
, int fc_sp
)
2713 struct r600_cf_stack_entry
*sp
= &ctx
->bc
->fc_stack
[fc_sp
];
2715 sp
->mid
= (struct r600_bc_cf
**)realloc((void *)sp
->mid
,
2716 sizeof(struct r600_bc_cf
*) * (sp
->num_mid
+ 1));
2717 sp
->mid
[sp
->num_mid
] = ctx
->bc
->cf_last
;
2721 static void fc_pushlevel(struct r600_shader_ctx
*ctx
, int type
)
2724 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
= type
;
2725 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
= ctx
->bc
->cf_last
;
2728 static void fc_poplevel(struct r600_shader_ctx
*ctx
)
2730 struct r600_cf_stack_entry
*sp
= &ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
];
2742 static int emit_return(struct r600_shader_ctx
*ctx
)
2744 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_RETURN
);
2748 static int emit_jump_to_offset(struct r600_shader_ctx
*ctx
, int pops
, int offset
)
2751 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_JUMP
);
2752 ctx
->bc
->cf_last
->pop_count
= pops
;
2753 /* TODO work out offset */
2757 static int emit_setret_in_loop_flag(struct r600_shader_ctx
*ctx
, unsigned flag_value
)
2762 static void emit_testflag(struct r600_shader_ctx
*ctx
)
2767 static void emit_return_on_flag(struct r600_shader_ctx
*ctx
, unsigned ifidx
)
2770 emit_jump_to_offset(ctx
, 1, 4);
2771 emit_setret_in_loop_flag(ctx
, V_SQ_ALU_SRC_0
);
2772 pops(ctx
, ifidx
+ 1);
2776 static void break_loop_on_flag(struct r600_shader_ctx
*ctx
, unsigned fc_sp
)
2780 r600_bc_add_cfinst(ctx
->bc
, ctx
->inst_info
->r600_opcode
);
2781 ctx
->bc
->cf_last
->pop_count
= 1;
2783 fc_set_mid(ctx
, fc_sp
);
2789 static int tgsi_if(struct r600_shader_ctx
*ctx
)
2791 emit_logic_pred(ctx
, CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE
));
2793 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_JUMP
));
2795 fc_pushlevel(ctx
, FC_IF
);
2797 callstack_check_depth(ctx
, FC_PUSH_VPM
, 0);
2801 static int tgsi_else(struct r600_shader_ctx
*ctx
)
2803 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_ELSE
));
2804 ctx
->bc
->cf_last
->pop_count
= 1;
2806 fc_set_mid(ctx
, ctx
->bc
->fc_sp
);
2807 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
;
2811 static int tgsi_endif(struct r600_shader_ctx
*ctx
)
2814 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_IF
) {
2815 R600_ERR("if/endif unbalanced in shader\n");
2819 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
== NULL
) {
2820 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2821 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->pop_count
= 1;
2823 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
[0]->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2827 callstack_decrease_current(ctx
, FC_PUSH_VPM
);
2831 static int tgsi_bgnloop(struct r600_shader_ctx
*ctx
)
2833 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL
));
2835 fc_pushlevel(ctx
, FC_LOOP
);
2837 /* check stack depth */
2838 callstack_check_depth(ctx
, FC_LOOP
, 0);
2842 static int tgsi_endloop(struct r600_shader_ctx
*ctx
)
2846 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END
));
2848 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_LOOP
) {
2849 R600_ERR("loop/endloop in shader code are not paired.\n");
2853 /* fixup loop pointers - from r600isa
2854 LOOP END points to CF after LOOP START,
2855 LOOP START point to CF after LOOP END
2856 BRK/CONT point to LOOP END CF
2858 ctx
->bc
->cf_last
->cf_addr
= ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->id
+ 2;
2860 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2862 for (i
= 0; i
< ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].num_mid
; i
++) {
2863 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
[i
]->cf_addr
= ctx
->bc
->cf_last
->id
;
2865 /* TODO add LOOPRET support */
2867 callstack_decrease_current(ctx
, FC_LOOP
);
2871 static int tgsi_loop_brk_cont(struct r600_shader_ctx
*ctx
)
2875 for (fscp
= ctx
->bc
->fc_sp
; fscp
> 0; fscp
--)
2877 if (FC_LOOP
== ctx
->bc
->fc_stack
[fscp
].type
)
2882 R600_ERR("Break not inside loop/endloop pair\n");
2886 r600_bc_add_cfinst(ctx
->bc
, ctx
->inst_info
->r600_opcode
);
2887 ctx
->bc
->cf_last
->pop_count
= 1;
2889 fc_set_mid(ctx
, fscp
);
2892 callstack_check_depth(ctx
, FC_PUSH_VPM
, 1);
2896 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[] = {
2897 {TGSI_OPCODE_ARL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_r600_arl
},
2898 {TGSI_OPCODE_MOV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
2899 {TGSI_OPCODE_LIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
2902 * For state trackers other than OpenGL, we'll want to use
2903 * _RECIP_IEEE instead.
2905 {TGSI_OPCODE_RCP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_CLAMPED
, tgsi_trans_srcx_replicate
},
2907 {TGSI_OPCODE_RSQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_rsq
},
2908 {TGSI_OPCODE_EXP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
2909 {TGSI_OPCODE_LOG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_log
},
2910 {TGSI_OPCODE_MUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
2911 {TGSI_OPCODE_ADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
2912 {TGSI_OPCODE_DP3
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2913 {TGSI_OPCODE_DP4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2914 {TGSI_OPCODE_DST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_opdst
},
2915 {TGSI_OPCODE_MIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
2916 {TGSI_OPCODE_MAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
2917 {TGSI_OPCODE_SLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
2918 {TGSI_OPCODE_SGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
2919 {TGSI_OPCODE_MAD
, 1, V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
2920 {TGSI_OPCODE_SUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
2921 {TGSI_OPCODE_LRP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
2922 {TGSI_OPCODE_CND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2924 {20, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2925 {TGSI_OPCODE_DP2A
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2927 {22, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2928 {23, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2929 {TGSI_OPCODE_FRC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
2930 {TGSI_OPCODE_CLAMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2931 {TGSI_OPCODE_FLR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
2932 {TGSI_OPCODE_ROUND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2933 {TGSI_OPCODE_EX2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
2934 {TGSI_OPCODE_LG2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
2935 {TGSI_OPCODE_POW
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
2936 {TGSI_OPCODE_XPD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
2938 {32, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2939 {TGSI_OPCODE_ABS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
2940 {TGSI_OPCODE_RCC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2941 {TGSI_OPCODE_DPH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2942 {TGSI_OPCODE_COS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
, tgsi_trig
},
2943 {TGSI_OPCODE_DDX
, 0, SQ_TEX_INST_GET_GRADIENTS_H
, tgsi_tex
},
2944 {TGSI_OPCODE_DDY
, 0, SQ_TEX_INST_GET_GRADIENTS_V
, tgsi_tex
},
2945 {TGSI_OPCODE_KILP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* predicated kill */
2946 {TGSI_OPCODE_PK2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2947 {TGSI_OPCODE_PK2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2948 {TGSI_OPCODE_PK4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2949 {TGSI_OPCODE_PK4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2950 {TGSI_OPCODE_RFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2951 {TGSI_OPCODE_SEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
2952 {TGSI_OPCODE_SFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2953 {TGSI_OPCODE_SGT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
2954 {TGSI_OPCODE_SIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
2955 {TGSI_OPCODE_SLE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
2956 {TGSI_OPCODE_SNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
2957 {TGSI_OPCODE_STR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2958 {TGSI_OPCODE_TEX
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
2959 {TGSI_OPCODE_TXD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2960 {TGSI_OPCODE_TXP
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
2961 {TGSI_OPCODE_UP2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2962 {TGSI_OPCODE_UP2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2963 {TGSI_OPCODE_UP4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2964 {TGSI_OPCODE_UP4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2965 {TGSI_OPCODE_X2D
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2966 {TGSI_OPCODE_ARA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2967 {TGSI_OPCODE_ARR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_r600_arl
},
2968 {TGSI_OPCODE_BRA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2969 {TGSI_OPCODE_CAL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2970 {TGSI_OPCODE_RET
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2971 {TGSI_OPCODE_SSG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
2972 {TGSI_OPCODE_CMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
2973 {TGSI_OPCODE_SCS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_scs
},
2974 {TGSI_OPCODE_TXB
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
2975 {TGSI_OPCODE_NRM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2976 {TGSI_OPCODE_DIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2977 {TGSI_OPCODE_DP2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
2978 {TGSI_OPCODE_TXL
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
2979 {TGSI_OPCODE_BRK
, 0, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK
, tgsi_loop_brk_cont
},
2980 {TGSI_OPCODE_IF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
2982 {75, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2983 {76, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2984 {TGSI_OPCODE_ELSE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
2985 {TGSI_OPCODE_ENDIF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
2987 {79, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2988 {80, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2989 {TGSI_OPCODE_PUSHA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2990 {TGSI_OPCODE_POPA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2991 {TGSI_OPCODE_CEIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2992 {TGSI_OPCODE_I2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2993 {TGSI_OPCODE_NOT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2994 {TGSI_OPCODE_TRUNC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
2995 {TGSI_OPCODE_SHL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2997 {88, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2998 {TGSI_OPCODE_AND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
2999 {TGSI_OPCODE_OR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3000 {TGSI_OPCODE_MOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3001 {TGSI_OPCODE_XOR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3002 {TGSI_OPCODE_SAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3003 {TGSI_OPCODE_TXF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3004 {TGSI_OPCODE_TXQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3005 {TGSI_OPCODE_CONT
, 0, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE
, tgsi_loop_brk_cont
},
3006 {TGSI_OPCODE_EMIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3007 {TGSI_OPCODE_ENDPRIM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3008 {TGSI_OPCODE_BGNLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_bgnloop
},
3009 {TGSI_OPCODE_BGNSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3010 {TGSI_OPCODE_ENDLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endloop
},
3011 {TGSI_OPCODE_ENDSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3013 {103, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3014 {104, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3015 {105, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3016 {106, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3017 {TGSI_OPCODE_NOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3019 {108, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3020 {109, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3021 {110, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3022 {111, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3023 {TGSI_OPCODE_NRM4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3024 {TGSI_OPCODE_CALLNZ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3025 {TGSI_OPCODE_IFC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3026 {TGSI_OPCODE_BREAKC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3027 {TGSI_OPCODE_KIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
3028 {TGSI_OPCODE_END
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
3030 {118, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3031 {TGSI_OPCODE_F2I
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3032 {TGSI_OPCODE_IDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3033 {TGSI_OPCODE_IMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3034 {TGSI_OPCODE_IMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3035 {TGSI_OPCODE_INEG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3036 {TGSI_OPCODE_ISGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3037 {TGSI_OPCODE_ISHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3038 {TGSI_OPCODE_ISLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3039 {TGSI_OPCODE_F2U
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3040 {TGSI_OPCODE_U2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3041 {TGSI_OPCODE_UADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3042 {TGSI_OPCODE_UDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3043 {TGSI_OPCODE_UMAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3044 {TGSI_OPCODE_UMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3045 {TGSI_OPCODE_UMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3046 {TGSI_OPCODE_UMOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3047 {TGSI_OPCODE_UMUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3048 {TGSI_OPCODE_USEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3049 {TGSI_OPCODE_USGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3050 {TGSI_OPCODE_USHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3051 {TGSI_OPCODE_USLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3052 {TGSI_OPCODE_USNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3053 {TGSI_OPCODE_SWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3054 {TGSI_OPCODE_CASE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3055 {TGSI_OPCODE_DEFAULT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3056 {TGSI_OPCODE_ENDSWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3057 {TGSI_OPCODE_LAST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3060 static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction
[] = {
3061 {TGSI_OPCODE_ARL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_eg_arl
},
3062 {TGSI_OPCODE_MOV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
3063 {TGSI_OPCODE_LIT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
3064 {TGSI_OPCODE_RCP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
, tgsi_trans_srcx_replicate
},
3065 {TGSI_OPCODE_RSQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE
, tgsi_trans_srcx_replicate
},
3066 {TGSI_OPCODE_EXP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
3067 {TGSI_OPCODE_LOG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3068 {TGSI_OPCODE_MUL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
3069 {TGSI_OPCODE_ADD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
3070 {TGSI_OPCODE_DP3
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3071 {TGSI_OPCODE_DP4
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3072 {TGSI_OPCODE_DST
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_opdst
},
3073 {TGSI_OPCODE_MIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
3074 {TGSI_OPCODE_MAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
3075 {TGSI_OPCODE_SLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
3076 {TGSI_OPCODE_SGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
3077 {TGSI_OPCODE_MAD
, 1, EG_V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
3078 {TGSI_OPCODE_SUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
3079 {TGSI_OPCODE_LRP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
3080 {TGSI_OPCODE_CND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3082 {20, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3083 {TGSI_OPCODE_DP2A
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3085 {22, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3086 {23, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3087 {TGSI_OPCODE_FRC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
3088 {TGSI_OPCODE_CLAMP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3089 {TGSI_OPCODE_FLR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
3090 {TGSI_OPCODE_ROUND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3091 {TGSI_OPCODE_EX2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
3092 {TGSI_OPCODE_LG2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
3093 {TGSI_OPCODE_POW
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
3094 {TGSI_OPCODE_XPD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
3096 {32, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3097 {TGSI_OPCODE_ABS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
3098 {TGSI_OPCODE_RCC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3099 {TGSI_OPCODE_DPH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3100 {TGSI_OPCODE_COS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
, tgsi_trig
},
3101 {TGSI_OPCODE_DDX
, 0, SQ_TEX_INST_GET_GRADIENTS_H
, tgsi_tex
},
3102 {TGSI_OPCODE_DDY
, 0, SQ_TEX_INST_GET_GRADIENTS_V
, tgsi_tex
},
3103 {TGSI_OPCODE_KILP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* predicated kill */
3104 {TGSI_OPCODE_PK2H
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3105 {TGSI_OPCODE_PK2US
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3106 {TGSI_OPCODE_PK4B
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3107 {TGSI_OPCODE_PK4UB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3108 {TGSI_OPCODE_RFL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3109 {TGSI_OPCODE_SEQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
3110 {TGSI_OPCODE_SFL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3111 {TGSI_OPCODE_SGT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
3112 {TGSI_OPCODE_SIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
3113 {TGSI_OPCODE_SLE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
3114 {TGSI_OPCODE_SNE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
3115 {TGSI_OPCODE_STR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3116 {TGSI_OPCODE_TEX
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
3117 {TGSI_OPCODE_TXD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3118 {TGSI_OPCODE_TXP
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
3119 {TGSI_OPCODE_UP2H
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3120 {TGSI_OPCODE_UP2US
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3121 {TGSI_OPCODE_UP4B
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3122 {TGSI_OPCODE_UP4UB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3123 {TGSI_OPCODE_X2D
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3124 {TGSI_OPCODE_ARA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3125 {TGSI_OPCODE_ARR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_eg_arl
},
3126 {TGSI_OPCODE_BRA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3127 {TGSI_OPCODE_CAL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3128 {TGSI_OPCODE_RET
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3129 {TGSI_OPCODE_SSG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
3130 {TGSI_OPCODE_CMP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
3131 {TGSI_OPCODE_SCS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_scs
},
3132 {TGSI_OPCODE_TXB
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
3133 {TGSI_OPCODE_NRM
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3134 {TGSI_OPCODE_DIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3135 {TGSI_OPCODE_DP2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3136 {TGSI_OPCODE_TXL
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
3137 {TGSI_OPCODE_BRK
, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK
, tgsi_loop_brk_cont
},
3138 {TGSI_OPCODE_IF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
3140 {75, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3141 {76, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3142 {TGSI_OPCODE_ELSE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
3143 {TGSI_OPCODE_ENDIF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
3145 {79, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3146 {80, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3147 {TGSI_OPCODE_PUSHA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3148 {TGSI_OPCODE_POPA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3149 {TGSI_OPCODE_CEIL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3150 {TGSI_OPCODE_I2F
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3151 {TGSI_OPCODE_NOT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3152 {TGSI_OPCODE_TRUNC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
3153 {TGSI_OPCODE_SHL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3155 {88, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3156 {TGSI_OPCODE_AND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3157 {TGSI_OPCODE_OR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3158 {TGSI_OPCODE_MOD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3159 {TGSI_OPCODE_XOR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3160 {TGSI_OPCODE_SAD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3161 {TGSI_OPCODE_TXF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3162 {TGSI_OPCODE_TXQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3163 {TGSI_OPCODE_CONT
, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE
, tgsi_loop_brk_cont
},
3164 {TGSI_OPCODE_EMIT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3165 {TGSI_OPCODE_ENDPRIM
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3166 {TGSI_OPCODE_BGNLOOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_bgnloop
},
3167 {TGSI_OPCODE_BGNSUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3168 {TGSI_OPCODE_ENDLOOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endloop
},
3169 {TGSI_OPCODE_ENDSUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3171 {103, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3172 {104, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3173 {105, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3174 {106, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3175 {TGSI_OPCODE_NOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3177 {108, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3178 {109, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3179 {110, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3180 {111, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3181 {TGSI_OPCODE_NRM4
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3182 {TGSI_OPCODE_CALLNZ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3183 {TGSI_OPCODE_IFC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3184 {TGSI_OPCODE_BREAKC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3185 {TGSI_OPCODE_KIL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
3186 {TGSI_OPCODE_END
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
3188 {118, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3189 {TGSI_OPCODE_F2I
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3190 {TGSI_OPCODE_IDIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3191 {TGSI_OPCODE_IMAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3192 {TGSI_OPCODE_IMIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3193 {TGSI_OPCODE_INEG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3194 {TGSI_OPCODE_ISGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3195 {TGSI_OPCODE_ISHR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3196 {TGSI_OPCODE_ISLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3197 {TGSI_OPCODE_F2U
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3198 {TGSI_OPCODE_U2F
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3199 {TGSI_OPCODE_UADD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3200 {TGSI_OPCODE_UDIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3201 {TGSI_OPCODE_UMAD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3202 {TGSI_OPCODE_UMAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3203 {TGSI_OPCODE_UMIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3204 {TGSI_OPCODE_UMOD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3205 {TGSI_OPCODE_UMUL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3206 {TGSI_OPCODE_USEQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3207 {TGSI_OPCODE_USGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3208 {TGSI_OPCODE_USHR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3209 {TGSI_OPCODE_USLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3210 {TGSI_OPCODE_USNE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3211 {TGSI_OPCODE_SWITCH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3212 {TGSI_OPCODE_CASE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3213 {TGSI_OPCODE_DEFAULT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3214 {TGSI_OPCODE_ENDSWITCH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3215 {TGSI_OPCODE_LAST
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},