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_screen.h"
29 #include "r600_context.h"
30 #include "r600_shader.h"
38 struct r600_shader_tgsi_instruction
;
40 struct r600_shader_ctx
{
41 struct tgsi_shader_info info
;
42 struct tgsi_parse_context parse
;
43 const struct tgsi_token
*tokens
;
45 unsigned file_offset
[TGSI_FILE_COUNT
];
47 struct r600_shader_tgsi_instruction
*inst_info
;
49 struct r600_shader
*shader
;
55 struct r600_shader_tgsi_instruction
{
59 int (*process
)(struct r600_shader_ctx
*ctx
);
62 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[];
63 static int r600_shader_from_tgsi(const struct tgsi_token
*tokens
, struct r600_shader
*shader
);
65 static int r600_shader_update(struct pipe_context
*ctx
, struct r600_shader
*shader
)
67 struct r600_context
*rctx
= r600_context(ctx
);
68 const struct util_format_description
*desc
;
69 enum pipe_format resource_format
[160];
70 unsigned i
, nresources
= 0;
71 struct r600_bc
*bc
= &shader
->bc
;
72 struct r600_bc_cf
*cf
;
73 struct r600_bc_vtx
*vtx
;
75 if (shader
->processor_type
!= TGSI_PROCESSOR_VERTEX
)
77 for (i
= 0; i
< rctx
->vertex_elements
->count
; i
++) {
78 resource_format
[nresources
++] = rctx
->vertex_elements
->elements
[i
].src_format
;
80 LIST_FOR_EACH_ENTRY(cf
, &bc
->cf
, list
) {
82 case V_SQ_CF_WORD1_SQ_CF_INST_VTX
:
83 case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC
:
84 LIST_FOR_EACH_ENTRY(vtx
, &cf
->vtx
, list
) {
85 desc
= util_format_description(resource_format
[vtx
->buffer_id
]);
87 R600_ERR("unknown format %d\n", resource_format
[vtx
->buffer_id
]);
90 vtx
->dst_sel_x
= desc
->swizzle
[0];
91 vtx
->dst_sel_y
= desc
->swizzle
[1];
92 vtx
->dst_sel_z
= desc
->swizzle
[2];
93 vtx
->dst_sel_w
= desc
->swizzle
[3];
100 return r600_bc_build(&shader
->bc
);
103 int r600_pipe_shader_create(struct pipe_context
*ctx
,
104 struct r600_context_state
*rpshader
,
105 const struct tgsi_token
*tokens
)
107 struct r600_screen
*rscreen
= r600_screen(ctx
->screen
);
110 //fprintf(stderr, "--------------------------------------------------------------\n");
111 //tgsi_dump(tokens, 0);
112 if (rpshader
== NULL
)
114 rpshader
->shader
.family
= radeon_get_family(rscreen
->rw
);
115 r
= r600_shader_from_tgsi(tokens
, &rpshader
->shader
);
117 R600_ERR("translation from TGSI failed !\n");
120 r
= r600_bc_build(&rpshader
->shader
.bc
);
122 R600_ERR("building bytecode failed !\n");
125 //fprintf(stderr, "______________________________________________________________\n");
129 static int r600_pipe_shader_vs(struct pipe_context
*ctx
, struct r600_context_state
*rpshader
)
131 struct r600_screen
*rscreen
= r600_screen(ctx
->screen
);
132 struct r600_shader
*rshader
= &rpshader
->shader
;
133 struct radeon_state
*state
;
136 rpshader
->rstate
= radeon_state_decref(rpshader
->rstate
);
137 state
= radeon_state(rscreen
->rw
, R600_VS_SHADER_TYPE
, R600_VS_SHADER
);
140 for (i
= 0; i
< 10; i
++) {
141 state
->states
[R600_VS_SHADER__SPI_VS_OUT_ID_0
+ i
] = 0;
143 /* so far never got proper semantic id from tgsi */
144 for (i
= 0; i
< 32; i
++) {
145 tmp
= i
<< ((i
& 3) * 8);
146 state
->states
[R600_VS_SHADER__SPI_VS_OUT_ID_0
+ i
/ 4] |= tmp
;
148 state
->states
[R600_VS_SHADER__SPI_VS_OUT_CONFIG
] = S_0286C4_VS_EXPORT_COUNT(rshader
->noutput
- 2);
149 state
->states
[R600_VS_SHADER__SQ_PGM_RESOURCES_VS
] = S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
150 S_028868_STACK_SIZE(rshader
->bc
.nstack
);
151 rpshader
->rstate
= state
;
152 rpshader
->rstate
->bo
[0] = radeon_bo_incref(rscreen
->rw
, rpshader
->bo
);
153 rpshader
->rstate
->bo
[1] = radeon_bo_incref(rscreen
->rw
, rpshader
->bo
);
154 rpshader
->rstate
->nbo
= 2;
155 rpshader
->rstate
->placement
[0] = RADEON_GEM_DOMAIN_GTT
;
156 rpshader
->rstate
->placement
[2] = RADEON_GEM_DOMAIN_GTT
;
157 return radeon_state_pm4(state
);
160 static int r600_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_context_state
*rpshader
)
162 const struct pipe_rasterizer_state
*rasterizer
;
163 struct r600_screen
*rscreen
= r600_screen(ctx
->screen
);
164 struct r600_shader
*rshader
= &rpshader
->shader
;
165 struct r600_context
*rctx
= r600_context(ctx
);
166 struct radeon_state
*state
;
167 unsigned i
, tmp
, exports_ps
, num_cout
;
169 rasterizer
= &rctx
->rasterizer
->state
.rasterizer
;
170 rpshader
->rstate
= radeon_state_decref(rpshader
->rstate
);
171 state
= radeon_state(rscreen
->rw
, R600_PS_SHADER_TYPE
, R600_PS_SHADER
);
174 for (i
= 0; i
< rshader
->ninput
; i
++) {
175 tmp
= S_028644_SEMANTIC(i
);
176 tmp
|= S_028644_SEL_CENTROID(1);
177 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_COLOR
||
178 rshader
->input
[i
].name
== TGSI_SEMANTIC_BCOLOR
) {
179 tmp
|= S_028644_FLAT_SHADE(rshader
->flat_shade
);
181 if (rasterizer
->sprite_coord_enable
& (1 << i
)) {
182 tmp
|= S_028644_PT_SPRITE_TEX(1);
184 state
->states
[R600_PS_SHADER__SPI_PS_INPUT_CNTL_0
+ i
] = tmp
;
189 for (i
= 0; i
< rshader
->noutput
; i
++) {
190 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
)
192 else if (rshader
->output
[i
].name
== TGSI_SEMANTIC_COLOR
) {
193 exports_ps
|= (1 << (num_cout
+1));
198 /* always at least export 1 component per pixel */
201 state
->states
[R600_PS_SHADER__SPI_PS_IN_CONTROL_0
] = S_0286CC_NUM_INTERP(rshader
->ninput
) |
202 S_0286CC_PERSP_GRADIENT_ENA(1);
203 state
->states
[R600_PS_SHADER__SPI_PS_IN_CONTROL_1
] = 0x00000000;
204 state
->states
[R600_PS_SHADER__SQ_PGM_RESOURCES_PS
] = S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
205 S_028868_STACK_SIZE(rshader
->bc
.nstack
);
206 state
->states
[R600_PS_SHADER__SQ_PGM_EXPORTS_PS
] = exports_ps
;
207 rpshader
->rstate
= state
;
208 rpshader
->rstate
->bo
[0] = radeon_bo_incref(rscreen
->rw
, rpshader
->bo
);
209 rpshader
->rstate
->nbo
= 1;
210 rpshader
->rstate
->placement
[0] = RADEON_GEM_DOMAIN_GTT
;
211 return radeon_state_pm4(state
);
214 static int r600_pipe_shader(struct pipe_context
*ctx
, struct r600_context_state
*rpshader
)
216 struct r600_screen
*rscreen
= r600_screen(ctx
->screen
);
217 struct r600_context
*rctx
= r600_context(ctx
);
218 struct r600_shader
*rshader
= &rpshader
->shader
;
221 /* copy new shader */
222 radeon_bo_decref(rscreen
->rw
, rpshader
->bo
);
224 rpshader
->bo
= radeon_bo(rscreen
->rw
, 0, rshader
->bc
.ndw
* 4,
226 if (rpshader
->bo
== NULL
) {
229 radeon_bo_map(rscreen
->rw
, rpshader
->bo
);
230 memcpy(rpshader
->bo
->data
, rshader
->bc
.bytecode
, rshader
->bc
.ndw
* 4);
231 radeon_bo_unmap(rscreen
->rw
, rpshader
->bo
);
233 rshader
->flat_shade
= rctx
->flat_shade
;
234 switch (rshader
->processor_type
) {
235 case TGSI_PROCESSOR_VERTEX
:
236 r
= r600_pipe_shader_vs(ctx
, rpshader
);
238 case TGSI_PROCESSOR_FRAGMENT
:
239 r
= r600_pipe_shader_ps(ctx
, rpshader
);
248 int r600_pipe_shader_update(struct pipe_context
*ctx
, struct r600_context_state
*rpshader
)
250 struct r600_context
*rctx
= r600_context(ctx
);
253 if (rpshader
== NULL
)
255 /* there should be enough input */
256 if (rctx
->vertex_elements
->count
< rpshader
->shader
.bc
.nresource
) {
257 R600_ERR("%d resources provided, expecting %d\n",
258 rctx
->vertex_elements
->count
, rpshader
->shader
.bc
.nresource
);
261 r
= r600_shader_update(ctx
, &rpshader
->shader
);
264 return r600_pipe_shader(ctx
, rpshader
);
267 static int tgsi_is_supported(struct r600_shader_ctx
*ctx
)
269 struct tgsi_full_instruction
*i
= &ctx
->parse
.FullToken
.FullInstruction
;
272 if (i
->Instruction
.NumDstRegs
> 1) {
273 R600_ERR("too many dst (%d)\n", i
->Instruction
.NumDstRegs
);
276 if (i
->Instruction
.Predicate
) {
277 R600_ERR("predicate unsupported\n");
281 if (i
->Instruction
.Label
) {
282 R600_ERR("label unsupported\n");
286 for (j
= 0; j
< i
->Instruction
.NumSrcRegs
; j
++) {
287 if (i
->Src
[j
].Register
.Indirect
||
288 i
->Src
[j
].Register
.Dimension
||
289 i
->Src
[j
].Register
.Absolute
) {
290 R600_ERR("unsupported src (indirect|dimension|absolute)\n");
294 for (j
= 0; j
< i
->Instruction
.NumDstRegs
; j
++) {
295 if (i
->Dst
[j
].Register
.Indirect
|| i
->Dst
[j
].Register
.Dimension
) {
296 R600_ERR("unsupported dst (indirect|dimension)\n");
303 static int tgsi_declaration(struct r600_shader_ctx
*ctx
)
305 struct tgsi_full_declaration
*d
= &ctx
->parse
.FullToken
.FullDeclaration
;
306 struct r600_bc_vtx vtx
;
310 switch (d
->Declaration
.File
) {
311 case TGSI_FILE_INPUT
:
312 i
= ctx
->shader
->ninput
++;
313 ctx
->shader
->input
[i
].name
= d
->Semantic
.Name
;
314 ctx
->shader
->input
[i
].sid
= d
->Semantic
.Index
;
315 ctx
->shader
->input
[i
].interpolate
= d
->Declaration
.Interpolate
;
316 ctx
->shader
->input
[i
].gpr
= ctx
->file_offset
[TGSI_FILE_INPUT
] + i
;
317 if (ctx
->type
== TGSI_PROCESSOR_VERTEX
) {
318 /* turn input into fetch */
319 memset(&vtx
, 0, sizeof(struct r600_bc_vtx
));
323 /* register containing the index into the buffer */
326 vtx
.mega_fetch_count
= 0x1F;
327 vtx
.dst_gpr
= ctx
->shader
->input
[i
].gpr
;
332 r
= r600_bc_add_vtx(ctx
->bc
, &vtx
);
337 case TGSI_FILE_OUTPUT
:
338 i
= ctx
->shader
->noutput
++;
339 ctx
->shader
->output
[i
].name
= d
->Semantic
.Name
;
340 ctx
->shader
->output
[i
].sid
= d
->Semantic
.Index
;
341 ctx
->shader
->output
[i
].gpr
= ctx
->file_offset
[TGSI_FILE_OUTPUT
] + i
;
342 ctx
->shader
->output
[i
].interpolate
= d
->Declaration
.Interpolate
;
344 case TGSI_FILE_CONSTANT
:
345 case TGSI_FILE_TEMPORARY
:
346 case TGSI_FILE_SAMPLER
:
349 R600_ERR("unsupported file %d declaration\n", d
->Declaration
.File
);
355 int r600_shader_from_tgsi(const struct tgsi_token
*tokens
, struct r600_shader
*shader
)
357 struct tgsi_full_immediate
*immediate
;
358 struct r600_shader_ctx ctx
;
359 struct r600_bc_output output
[32];
360 unsigned output_done
, noutput
;
364 ctx
.bc
= &shader
->bc
;
366 r
= r600_bc_init(ctx
.bc
, shader
->family
);
370 tgsi_scan_shader(tokens
, &ctx
.info
);
371 tgsi_parse_init(&ctx
.parse
, tokens
);
372 ctx
.type
= ctx
.parse
.FullHeader
.Processor
.Processor
;
373 shader
->processor_type
= ctx
.type
;
375 /* register allocations */
376 /* Values [0,127] correspond to GPR[0..127].
377 * Values [128,159] correspond to constant buffer bank 0
378 * Values [160,191] correspond to constant buffer bank 1
379 * Values [256,511] correspond to cfile constants c[0..255].
380 * Other special values are shown in the list below.
381 * 244 ALU_SRC_1_DBL_L: special constant 1.0 double-float, LSW. (RV670+)
382 * 245 ALU_SRC_1_DBL_M: special constant 1.0 double-float, MSW. (RV670+)
383 * 246 ALU_SRC_0_5_DBL_L: special constant 0.5 double-float, LSW. (RV670+)
384 * 247 ALU_SRC_0_5_DBL_M: special constant 0.5 double-float, MSW. (RV670+)
385 * 248 SQ_ALU_SRC_0: special constant 0.0.
386 * 249 SQ_ALU_SRC_1: special constant 1.0 float.
387 * 250 SQ_ALU_SRC_1_INT: special constant 1 integer.
388 * 251 SQ_ALU_SRC_M_1_INT: special constant -1 integer.
389 * 252 SQ_ALU_SRC_0_5: special constant 0.5 float.
390 * 253 SQ_ALU_SRC_LITERAL: literal constant.
391 * 254 SQ_ALU_SRC_PV: previous vector result.
392 * 255 SQ_ALU_SRC_PS: previous scalar result.
394 for (i
= 0; i
< TGSI_FILE_COUNT
; i
++) {
395 ctx
.file_offset
[i
] = 0;
397 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
398 ctx
.file_offset
[TGSI_FILE_INPUT
] = 1;
400 ctx
.file_offset
[TGSI_FILE_OUTPUT
] = ctx
.file_offset
[TGSI_FILE_INPUT
] +
401 ctx
.info
.file_count
[TGSI_FILE_INPUT
];
402 ctx
.file_offset
[TGSI_FILE_TEMPORARY
] = ctx
.file_offset
[TGSI_FILE_OUTPUT
] +
403 ctx
.info
.file_count
[TGSI_FILE_OUTPUT
];
404 ctx
.file_offset
[TGSI_FILE_CONSTANT
] = 256;
405 ctx
.file_offset
[TGSI_FILE_IMMEDIATE
] = 253;
406 ctx
.temp_reg
= ctx
.file_offset
[TGSI_FILE_TEMPORARY
] +
407 ctx
.info
.file_count
[TGSI_FILE_TEMPORARY
];
412 while (!tgsi_parse_end_of_tokens(&ctx
.parse
)) {
413 tgsi_parse_token(&ctx
.parse
);
414 switch (ctx
.parse
.FullToken
.Token
.Type
) {
415 case TGSI_TOKEN_TYPE_IMMEDIATE
:
416 immediate
= &ctx
.parse
.FullToken
.FullImmediate
;
417 ctx
.literals
= realloc(ctx
.literals
, (ctx
.nliterals
+ 1) * 16);
418 if(ctx
.literals
== NULL
) {
422 ctx
.literals
[ctx
.nliterals
* 4 + 0] = immediate
->u
[0].Uint
;
423 ctx
.literals
[ctx
.nliterals
* 4 + 1] = immediate
->u
[1].Uint
;
424 ctx
.literals
[ctx
.nliterals
* 4 + 2] = immediate
->u
[2].Uint
;
425 ctx
.literals
[ctx
.nliterals
* 4 + 3] = immediate
->u
[3].Uint
;
428 case TGSI_TOKEN_TYPE_DECLARATION
:
429 r
= tgsi_declaration(&ctx
);
433 case TGSI_TOKEN_TYPE_INSTRUCTION
:
434 r
= tgsi_is_supported(&ctx
);
437 opcode
= ctx
.parse
.FullToken
.FullInstruction
.Instruction
.Opcode
;
438 ctx
.inst_info
= &r600_shader_tgsi_instruction
[opcode
];
439 r
= ctx
.inst_info
->process(&ctx
);
442 r
= r600_bc_add_literal(ctx
.bc
, ctx
.value
);
447 R600_ERR("unsupported token type %d\n", ctx
.parse
.FullToken
.Token
.Type
);
453 noutput
= shader
->noutput
;
454 for (i
= 0, pos0
= 0; i
< noutput
; i
++) {
455 memset(&output
[i
], 0, sizeof(struct r600_bc_output
));
456 output
[i
].gpr
= shader
->output
[i
].gpr
;
457 output
[i
].elem_size
= 3;
458 output
[i
].swizzle_x
= 0;
459 output
[i
].swizzle_y
= 1;
460 output
[i
].swizzle_z
= 2;
461 output
[i
].swizzle_w
= 3;
462 output
[i
].barrier
= 1;
463 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
;
464 output
[i
].array_base
= i
- pos0
;
465 output
[i
].inst
= V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
;
467 case TGSI_PROCESSOR_VERTEX
:
468 if (shader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
) {
469 output
[i
].array_base
= 60;
470 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS
;
471 /* position doesn't count in array_base */
474 if (shader
->output
[i
].name
== TGSI_SEMANTIC_PSIZE
) {
475 output
[i
].array_base
= 61;
476 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS
;
477 /* position doesn't count in array_base */
481 case TGSI_PROCESSOR_FRAGMENT
:
482 if (shader
->output
[i
].name
== TGSI_SEMANTIC_COLOR
) {
483 output
[i
].array_base
= shader
->output
[i
].sid
;
484 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
485 } else if (shader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
) {
486 output
[i
].array_base
= 61;
487 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
489 R600_ERR("unsupported fragment output name %d\n", shader
->output
[i
].name
);
495 R600_ERR("unsupported processor type %d\n", ctx
.type
);
500 /* add fake param output for vertex shader if no param is exported */
501 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
502 for (i
= 0, pos0
= 0; i
< noutput
; i
++) {
503 if (output
[i
].type
== V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
) {
509 memset(&output
[i
], 0, sizeof(struct r600_bc_output
));
511 output
[i
].elem_size
= 3;
512 output
[i
].swizzle_x
= 0;
513 output
[i
].swizzle_y
= 1;
514 output
[i
].swizzle_z
= 2;
515 output
[i
].swizzle_w
= 3;
516 output
[i
].barrier
= 1;
517 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
;
518 output
[i
].array_base
= 0;
519 output
[i
].inst
= V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
;
523 /* add fake pixel export */
524 if (ctx
.type
== TGSI_PROCESSOR_FRAGMENT
&& !noutput
) {
525 memset(&output
[0], 0, sizeof(struct r600_bc_output
));
527 output
[0].elem_size
= 3;
528 output
[0].swizzle_x
= 7;
529 output
[0].swizzle_y
= 7;
530 output
[0].swizzle_z
= 7;
531 output
[0].swizzle_w
= 7;
532 output
[0].barrier
= 1;
533 output
[0].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
534 output
[0].array_base
= 0;
535 output
[0].inst
= V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
;
538 /* set export done on last export of each type */
539 for (i
= noutput
- 1, output_done
= 0; i
>= 0; i
--) {
540 if (i
== (noutput
- 1)) {
541 output
[i
].end_of_program
= 1;
543 if (!(output_done
& (1 << output
[i
].type
))) {
544 output_done
|= (1 << output
[i
].type
);
545 output
[i
].inst
= V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE
;
548 /* add output to bytecode */
549 for (i
= 0; i
< noutput
; i
++) {
550 r
= r600_bc_add_output(ctx
.bc
, &output
[i
]);
555 tgsi_parse_free(&ctx
.parse
);
559 tgsi_parse_free(&ctx
.parse
);
563 static int tgsi_unsupported(struct r600_shader_ctx
*ctx
)
565 R600_ERR("%d tgsi opcode unsupported\n", ctx
->inst_info
->tgsi_opcode
);
569 static int tgsi_end(struct r600_shader_ctx
*ctx
)
574 static int tgsi_src(struct r600_shader_ctx
*ctx
,
575 const struct tgsi_full_src_register
*tgsi_src
,
576 struct r600_bc_alu_src
*r600_src
)
579 memset(r600_src
, 0, sizeof(struct r600_bc_alu_src
));
580 r600_src
->sel
= tgsi_src
->Register
.Index
;
581 if (tgsi_src
->Register
.File
== TGSI_FILE_IMMEDIATE
) {
583 index
= tgsi_src
->Register
.Index
;
584 ctx
->value
[0] = ctx
->literals
[index
* 4 + 0];
585 ctx
->value
[1] = ctx
->literals
[index
* 4 + 1];
586 ctx
->value
[2] = ctx
->literals
[index
* 4 + 2];
587 ctx
->value
[3] = ctx
->literals
[index
* 4 + 3];
589 r600_src
->neg
= tgsi_src
->Register
.Negate
;
590 r600_src
->sel
+= ctx
->file_offset
[tgsi_src
->Register
.File
];
594 static int tgsi_dst(struct r600_shader_ctx
*ctx
,
595 const struct tgsi_full_dst_register
*tgsi_dst
,
597 struct r600_bc_alu_dst
*r600_dst
)
599 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
601 r600_dst
->sel
= tgsi_dst
->Register
.Index
;
602 r600_dst
->sel
+= ctx
->file_offset
[tgsi_dst
->Register
.File
];
603 r600_dst
->chan
= swizzle
;
605 if (inst
->Instruction
.Saturate
) {
611 static unsigned tgsi_chan(const struct tgsi_full_src_register
*tgsi_src
, unsigned swizzle
)
615 return tgsi_src
->Register
.SwizzleX
;
617 return tgsi_src
->Register
.SwizzleY
;
619 return tgsi_src
->Register
.SwizzleZ
;
621 return tgsi_src
->Register
.SwizzleW
;
627 static int tgsi_split_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
629 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
630 struct r600_bc_alu alu
;
631 int i
, j
, k
, nconst
, r
;
633 for (i
= 0, nconst
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
634 if (inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
637 r
= tgsi_src(ctx
, &inst
->Src
[i
], &r600_src
[i
]);
642 for (i
= 0, j
= nconst
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
643 if (inst
->Src
[j
].Register
.File
== TGSI_FILE_CONSTANT
&& j
> 0) {
644 for (k
= 0; k
< 4; k
++) {
645 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
646 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
647 alu
.src
[0].sel
= r600_src
[0].sel
;
649 alu
.dst
.sel
= ctx
->temp_reg
+ j
;
654 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
658 r600_src
[0].sel
= ctx
->temp_reg
+ j
;
665 static int tgsi_op2_s(struct r600_shader_ctx
*ctx
, int swap
)
667 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
668 struct r600_bc_alu_src r600_src
[3];
669 struct r600_bc_alu alu
;
673 for (i
= 0; i
< 4; i
++) {
674 if (inst
->Dst
[0].Register
.WriteMask
& (1 << i
)) {
679 r
= tgsi_split_constant(ctx
, r600_src
);
682 for (i
= 0; i
< lasti
+ 1; i
++) {
683 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
686 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
687 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
691 alu
.inst
= ctx
->inst_info
->r600_opcode
;
693 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
694 alu
.src
[j
] = r600_src
[j
];
695 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
698 alu
.src
[0] = r600_src
[1];
699 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], i
);
701 alu
.src
[1] = r600_src
[0];
702 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
704 /* handle some special cases */
705 switch (ctx
->inst_info
->tgsi_opcode
) {
706 case TGSI_OPCODE_SUB
:
709 case TGSI_OPCODE_ABS
:
718 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
725 static int tgsi_op2(struct r600_shader_ctx
*ctx
)
727 return tgsi_op2_s(ctx
, 0);
730 static int tgsi_op2_swap(struct r600_shader_ctx
*ctx
)
732 return tgsi_op2_s(ctx
, 1);
736 * r600 - trunc to -PI..PI range
737 * r700 - normalize by dividing by 2PI
740 static int tgsi_trig(struct r600_shader_ctx
*ctx
)
742 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
743 struct r600_bc_alu_src r600_src
[3];
744 struct r600_bc_alu alu
;
746 uint32_t lit_vals
[4];
748 memset(lit_vals
, 0, 4*4);
749 r
= tgsi_split_constant(ctx
, r600_src
);
752 lit_vals
[0] = fui(1.0 /(3.1415926535 * 2));
753 lit_vals
[1] = fui(0.5f
);
755 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
756 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
;
760 alu
.dst
.sel
= ctx
->temp_reg
;
763 alu
.src
[0] = r600_src
[0];
764 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
766 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
768 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
771 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
774 r
= r600_bc_add_literal(ctx
->bc
, lit_vals
);
778 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
779 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
;
782 alu
.dst
.sel
= ctx
->temp_reg
;
785 alu
.src
[0].sel
= ctx
->temp_reg
;
788 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
792 if (ctx
->bc
->chiprev
== 0) {
793 lit_vals
[0] = fui(3.1415926535897f
* 2.0f
);
794 lit_vals
[1] = fui(-3.1415926535897f
);
796 lit_vals
[0] = fui(1.0f
);
797 lit_vals
[1] = fui(-0.5f
);
800 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
801 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
;
805 alu
.dst
.sel
= ctx
->temp_reg
;
808 alu
.src
[0].sel
= ctx
->temp_reg
;
811 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
813 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
816 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
819 r
= r600_bc_add_literal(ctx
->bc
, lit_vals
);
823 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
824 alu
.inst
= ctx
->inst_info
->r600_opcode
;
826 alu
.dst
.sel
= ctx
->temp_reg
;
829 alu
.src
[0].sel
= ctx
->temp_reg
;
832 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
836 /* replicate result */
837 for (i
= 0; i
< 4; i
++) {
838 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
839 alu
.src
[0].sel
= ctx
->temp_reg
;
840 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
842 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
845 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
848 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
855 static int tgsi_kill(struct r600_shader_ctx
*ctx
)
857 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
858 struct r600_bc_alu alu
;
861 for (i
= 0; i
< 4; i
++) {
862 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
863 alu
.inst
= ctx
->inst_info
->r600_opcode
;
865 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
866 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
869 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
873 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
880 static int tgsi_lit(struct r600_shader_ctx
*ctx
)
882 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
883 struct r600_bc_alu alu
;
887 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
888 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
889 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
; /*1.0*/
891 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
894 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 0) & 1;
895 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
899 /* dst.y = max(src.x, 0.0) */
900 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
901 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
;
902 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
905 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
; /*0.0*/
906 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], 0);
907 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
910 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 1) & 1;
911 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
915 /* dst.z = NOP - fill Z slot */
916 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
917 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
;
919 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
924 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
925 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
926 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
928 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
931 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 3) & 1;
933 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
937 if (inst
->Dst
[0].Register
.WriteMask
& (1 << 2))
942 /* dst.z = log(src.y) */
943 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
944 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED
;
945 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
948 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
949 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
953 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
960 /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */
961 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
962 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT
;
963 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
966 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
967 alu
.src
[1].sel
= sel
;
968 alu
.src
[1].chan
= chan
;
969 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[2]);
972 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], 0);
973 alu
.dst
.sel
= ctx
->temp_reg
;
978 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
982 /* dst.z = exp(tmp.x) */
983 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
984 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
;
985 alu
.src
[0].sel
= ctx
->temp_reg
;
987 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
991 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
998 static int tgsi_trans(struct r600_shader_ctx
*ctx
)
1000 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1001 struct r600_bc_alu alu
;
1004 for (i
= 0; i
< 4; i
++) {
1005 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1006 if (inst
->Dst
[0].Register
.WriteMask
& (1 << i
)) {
1007 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1008 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1009 r
= tgsi_src(ctx
, &inst
->Src
[j
], &alu
.src
[j
]);
1012 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1014 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1018 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1026 static int tgsi_helper_tempx_replicate(struct r600_shader_ctx
*ctx
)
1028 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1029 struct r600_bc_alu alu
;
1032 for (i
= 0; i
< 4; i
++) {
1033 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1034 alu
.src
[0].sel
= ctx
->temp_reg
;
1035 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
1037 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1040 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
1043 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1050 static int tgsi_trans_srcx_replicate(struct r600_shader_ctx
*ctx
)
1052 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1053 struct r600_bc_alu alu
;
1056 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1057 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1058 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1059 r
= tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1062 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1064 alu
.dst
.sel
= ctx
->temp_reg
;
1067 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1070 /* replicate result */
1071 return tgsi_helper_tempx_replicate(ctx
);
1074 static int tgsi_pow(struct r600_shader_ctx
*ctx
)
1076 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1077 struct r600_bc_alu alu
;
1081 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1082 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
;
1083 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1086 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1087 alu
.dst
.sel
= ctx
->temp_reg
;
1090 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1093 r
= r600_bc_add_literal(ctx
->bc
,ctx
->value
);
1097 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1098 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL_IEEE
;
1099 r
= tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[0]);
1102 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], 0);
1103 alu
.src
[1].sel
= ctx
->temp_reg
;
1104 alu
.dst
.sel
= ctx
->temp_reg
;
1107 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1110 r
= r600_bc_add_literal(ctx
->bc
,ctx
->value
);
1113 /* POW(a,b) = EXP2(b * LOG2(a))*/
1114 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1115 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
;
1116 alu
.src
[0].sel
= ctx
->temp_reg
;
1117 alu
.dst
.sel
= ctx
->temp_reg
;
1120 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1123 r
= r600_bc_add_literal(ctx
->bc
,ctx
->value
);
1126 return tgsi_helper_tempx_replicate(ctx
);
1129 static int tgsi_ssg(struct r600_shader_ctx
*ctx
)
1131 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1132 struct r600_bc_alu alu
;
1133 struct r600_bc_alu_src r600_src
[3];
1136 r
= tgsi_split_constant(ctx
, r600_src
);
1140 /* tmp = (src > 0 ? 1 : src) */
1141 for (i
= 0; i
< 4; i
++) {
1142 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1143 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
;
1145 alu
.dst
.sel
= ctx
->temp_reg
;
1148 alu
.src
[0] = r600_src
[0];
1149 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1151 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1153 alu
.src
[2] = r600_src
[0];
1154 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], i
);
1157 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1162 /* dst = (-tmp > 0 ? -1 : tmp) */
1163 for (i
= 0; i
< 4; i
++) {
1164 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1165 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
;
1167 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1171 alu
.src
[0].sel
= ctx
->temp_reg
;
1174 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1177 alu
.src
[2].sel
= ctx
->temp_reg
;
1182 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1189 static int tgsi_helper_copy(struct r600_shader_ctx
*ctx
, struct tgsi_full_instruction
*inst
)
1191 struct r600_bc_alu alu
;
1194 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1197 for (i
= 0; i
< 4; i
++) {
1198 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1199 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
))) {
1200 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
;
1203 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
1204 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1207 alu
.src
[0].sel
= ctx
->temp_reg
;
1208 alu
.src
[0].chan
= i
;
1213 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1220 static int tgsi_op3(struct r600_shader_ctx
*ctx
)
1222 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1223 struct r600_bc_alu_src r600_src
[3];
1224 struct r600_bc_alu alu
;
1227 r
= tgsi_split_constant(ctx
, r600_src
);
1230 /* do it in 2 step as op3 doesn't support writemask */
1231 for (i
= 0; i
< 4; i
++) {
1232 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1233 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1234 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1235 alu
.src
[j
] = r600_src
[j
];
1236 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1238 alu
.dst
.sel
= ctx
->temp_reg
;
1245 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1249 return tgsi_helper_copy(ctx
, inst
);
1252 static int tgsi_dp(struct r600_shader_ctx
*ctx
)
1254 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1255 struct r600_bc_alu_src r600_src
[3];
1256 struct r600_bc_alu alu
;
1259 r
= tgsi_split_constant(ctx
, r600_src
);
1262 for (i
= 0; i
< 4; i
++) {
1263 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1264 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1265 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1266 alu
.src
[j
] = r600_src
[j
];
1267 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1269 alu
.dst
.sel
= ctx
->temp_reg
;
1272 /* handle some special cases */
1273 switch (ctx
->inst_info
->tgsi_opcode
) {
1274 case TGSI_OPCODE_DP2
:
1276 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1277 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1280 case TGSI_OPCODE_DP3
:
1282 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1283 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1286 case TGSI_OPCODE_DPH
:
1288 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1289 alu
.src
[0].chan
= 0;
1299 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1303 return tgsi_helper_copy(ctx
, inst
);
1306 static int tgsi_tex(struct r600_shader_ctx
*ctx
)
1308 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1309 struct r600_bc_tex tex
;
1310 struct r600_bc_alu alu
;
1314 src_gpr
= ctx
->file_offset
[inst
->Src
[0].Register
.File
] + inst
->Src
[0].Register
.Index
;
1316 if (inst
->Instruction
.Opcode
== TGSI_OPCODE_TXP
) {
1317 /* Add perspective divide */
1318 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1319 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
;
1320 alu
.src
[0].sel
= src_gpr
;
1321 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1322 alu
.dst
.sel
= ctx
->temp_reg
;
1326 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1330 for (i
= 0; i
< 3; i
++) {
1331 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1332 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
;
1333 alu
.src
[0].sel
= ctx
->temp_reg
;
1334 alu
.src
[0].chan
= 3;
1335 alu
.src
[1].sel
= src_gpr
;
1336 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1337 alu
.dst
.sel
= ctx
->temp_reg
;
1340 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1344 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1345 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
1346 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1347 alu
.src
[0].chan
= 0;
1348 alu
.dst
.sel
= ctx
->temp_reg
;
1352 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1355 src_gpr
= ctx
->temp_reg
;
1356 } else if (inst
->Src
[0].Register
.File
!= TGSI_FILE_TEMPORARY
) {
1357 for (i
= 0; i
< 4; i
++) {
1358 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1359 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
1360 alu
.src
[0].sel
= src_gpr
;
1361 alu
.src
[0].chan
= i
;
1362 alu
.dst
.sel
= ctx
->temp_reg
;
1367 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1371 src_gpr
= ctx
->temp_reg
;
1374 memset(&tex
, 0, sizeof(struct r600_bc_tex
));
1375 tex
.inst
= ctx
->inst_info
->r600_opcode
;
1376 tex
.resource_id
= ctx
->file_offset
[inst
->Src
[1].Register
.File
] + inst
->Src
[1].Register
.Index
;
1377 tex
.sampler_id
= tex
.resource_id
;
1378 tex
.src_gpr
= src_gpr
;
1379 tex
.dst_gpr
= ctx
->file_offset
[inst
->Dst
[0].Register
.File
] + inst
->Dst
[0].Register
.Index
;
1389 if (inst
->Texture
.Texture
!= TGSI_TEXTURE_RECT
) {
1390 tex
.coord_type_x
= 1;
1391 tex
.coord_type_y
= 1;
1392 tex
.coord_type_z
= 1;
1393 tex
.coord_type_w
= 1;
1395 return r600_bc_add_tex(ctx
->bc
, &tex
);
1398 static int tgsi_lrp(struct r600_shader_ctx
*ctx
)
1400 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1401 struct r600_bc_alu_src r600_src
[3];
1402 struct r600_bc_alu alu
;
1406 r
= tgsi_split_constant(ctx
, r600_src
);
1410 for (i
= 0; i
< 4; i
++) {
1411 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1412 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
;
1413 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1414 alu
.src
[0].chan
= 0;
1415 alu
.src
[1] = r600_src
[0];
1416 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1418 alu
.dst
.sel
= ctx
->temp_reg
;
1424 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1428 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1432 /* (1 - src0) * src2 */
1433 for (i
= 0; i
< 4; i
++) {
1434 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1435 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
;
1436 alu
.src
[0].sel
= ctx
->temp_reg
;
1437 alu
.src
[0].chan
= i
;
1438 alu
.src
[1] = r600_src
[2];
1439 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1440 alu
.dst
.sel
= ctx
->temp_reg
;
1446 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1450 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1454 /* src0 * src1 + (1 - src0) * src2 */
1455 for (i
= 0; i
< 4; i
++) {
1456 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1457 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
;
1459 alu
.src
[0] = r600_src
[0];
1460 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1461 alu
.src
[1] = r600_src
[1];
1462 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
1463 alu
.src
[2].sel
= ctx
->temp_reg
;
1464 alu
.src
[2].chan
= i
;
1465 alu
.dst
.sel
= ctx
->temp_reg
;
1470 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1474 return tgsi_helper_copy(ctx
, inst
);
1477 static int tgsi_cmp(struct r600_shader_ctx
*ctx
)
1479 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1480 struct r600_bc_alu_src r600_src
[3];
1481 struct r600_bc_alu alu
;
1485 r
= tgsi_split_constant(ctx
, r600_src
);
1489 if (inst
->Dst
[0].Register
.WriteMask
!= 0xf)
1492 for (i
= 0; i
< 4; i
++) {
1493 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1494 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE
;
1495 alu
.src
[0] = r600_src
[0];
1496 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1498 alu
.src
[1] = r600_src
[2];
1499 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
1501 alu
.src
[2] = r600_src
[1];
1502 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[1], i
);
1505 alu
.dst
.sel
= ctx
->temp_reg
;
1507 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1516 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1521 return tgsi_helper_copy(ctx
, inst
);
1525 static int tgsi_xpd(struct r600_shader_ctx
*ctx
)
1527 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1528 struct r600_bc_alu_src r600_src
[3];
1529 struct r600_bc_alu alu
;
1530 uint32_t use_temp
= 0;
1533 if (inst
->Dst
[0].Register
.WriteMask
!= 0xf)
1536 r
= tgsi_split_constant(ctx
, r600_src
);
1540 for (i
= 0; i
< 4; i
++) {
1541 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1542 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
;
1544 alu
.src
[0] = r600_src
[0];
1547 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
1550 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1553 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
1556 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1557 alu
.src
[0].chan
= i
;
1560 alu
.src
[1] = r600_src
[1];
1563 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
1566 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
1569 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
1572 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1573 alu
.src
[1].chan
= i
;
1576 alu
.dst
.sel
= ctx
->temp_reg
;
1582 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1587 for (i
= 0; i
< 4; i
++) {
1588 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1589 alu
.inst
= V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
;
1591 alu
.src
[0] = r600_src
[0];
1594 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
1597 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
1600 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1603 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1604 alu
.src
[0].chan
= i
;
1607 alu
.src
[1] = r600_src
[1];
1610 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
1613 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
1616 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
1619 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1620 alu
.src
[1].chan
= i
;
1623 alu
.src
[2].sel
= ctx
->temp_reg
;
1625 alu
.src
[2].chan
= i
;
1628 alu
.dst
.sel
= ctx
->temp_reg
;
1630 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1639 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1644 return tgsi_helper_copy(ctx
, inst
);
1648 static int tgsi_exp(struct r600_shader_ctx
*ctx
)
1650 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1651 struct r600_bc_alu_src r600_src
[3];
1652 struct r600_bc_alu alu
;
1653 uint32_t use_temp
= 0;
1656 /* result.x = 2^floor(src); */
1657 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
1658 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1660 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
;
1661 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1665 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1667 alu
.dst
.sel
= ctx
->temp_reg
;
1671 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1675 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
;
1676 alu
.src
[0].sel
= ctx
->temp_reg
;
1677 alu
.src
[0].chan
= 0;
1679 alu
.dst
.sel
= ctx
->temp_reg
;
1683 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1688 /* result.y = tmp - floor(tmp); */
1689 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
1690 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1692 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
;
1693 alu
.src
[0] = r600_src
[0];
1694 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1697 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1699 alu
.dst
.sel
= ctx
->temp_reg
;
1700 // r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
1708 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1713 /* result.z = RoughApprox2ToX(tmp);*/
1714 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 0x1) {
1715 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1716 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
;
1717 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1720 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1722 alu
.dst
.sel
= ctx
->temp_reg
;
1728 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1734 /* result.w = 1.0;*/
1735 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 0x1) {
1736 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1738 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
;
1739 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1740 alu
.src
[0].chan
= 0;
1742 alu
.dst
.sel
= ctx
->temp_reg
;
1746 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1750 return tgsi_helper_copy(ctx
, inst
);
1753 static int emit_logic_pred(struct r600_shader_ctx
*ctx
, int opcode
)
1755 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1756 struct r600_bc_alu alu
, *lalu
;
1757 struct r600_bc_cf
*last
;
1760 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1764 alu
.dst
.sel
= ctx
->temp_reg
;
1768 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1771 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1772 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1773 alu
.src
[1].chan
= 0;
1777 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE
);
1784 static int pops(struct r600_shader_ctx
*ctx
, int pops
)
1786 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_POP
);
1787 ctx
->bc
->cf_last
->pop_count
= pops
;
1791 static int tgsi_if(struct r600_shader_ctx
*ctx
)
1793 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1795 emit_logic_pred(ctx
, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE
);
1798 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
= FC_IF
;
1799 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
= NULL
;
1800 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_JUMP
);
1802 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
= ctx
->bc
->cf_last
;
1806 static int tgsi_else(struct r600_shader_ctx
*ctx
)
1808 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1809 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_ELSE
);
1810 ctx
->bc
->cf_last
->pop_count
= 1;
1813 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
= ctx
->bc
->cf_last
;
1814 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
;
1818 static int tgsi_endif(struct r600_shader_ctx
*ctx
)
1821 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_IF
) {
1822 R600_ERR("if/endif unbalanced in shader\n");
1826 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
== NULL
) {
1827 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
1828 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->pop_count
= 1;
1830 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
1837 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[] = {
1838 {TGSI_OPCODE_ARL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1839 {TGSI_OPCODE_MOV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
1840 {TGSI_OPCODE_LIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
1841 {TGSI_OPCODE_RCP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
, tgsi_trans_srcx_replicate
},
1842 {TGSI_OPCODE_RSQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE
, tgsi_trans_srcx_replicate
},
1843 {TGSI_OPCODE_EXP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
1844 {TGSI_OPCODE_LOG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1845 {TGSI_OPCODE_MUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
1846 {TGSI_OPCODE_ADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
1847 {TGSI_OPCODE_DP3
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
1848 {TGSI_OPCODE_DP4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
1849 {TGSI_OPCODE_DST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1850 {TGSI_OPCODE_MIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
1851 {TGSI_OPCODE_MAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
1852 {TGSI_OPCODE_SLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
1853 {TGSI_OPCODE_SGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
1854 {TGSI_OPCODE_MAD
, 1, V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
1855 {TGSI_OPCODE_SUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
1856 {TGSI_OPCODE_LRP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
1857 {TGSI_OPCODE_CND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1859 {20, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1860 {TGSI_OPCODE_DP2A
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1862 {22, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1863 {23, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1864 {TGSI_OPCODE_FRC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
1865 {TGSI_OPCODE_CLAMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1866 {TGSI_OPCODE_FLR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
1867 {TGSI_OPCODE_ROUND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1868 {TGSI_OPCODE_EX2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
1869 {TGSI_OPCODE_LG2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
1870 {TGSI_OPCODE_POW
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
1871 {TGSI_OPCODE_XPD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
1873 {32, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1874 {TGSI_OPCODE_ABS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
1875 {TGSI_OPCODE_RCC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1876 {TGSI_OPCODE_DPH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
1877 {TGSI_OPCODE_COS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
, tgsi_trig
},
1878 {TGSI_OPCODE_DDX
, 0, SQ_TEX_INST_GET_GRADIENTS_H
, tgsi_tex
},
1879 {TGSI_OPCODE_DDY
, 0, SQ_TEX_INST_GET_GRADIENTS_V
, tgsi_tex
},
1880 {TGSI_OPCODE_KILP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
}, /* predicated kill */
1881 {TGSI_OPCODE_PK2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1882 {TGSI_OPCODE_PK2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1883 {TGSI_OPCODE_PK4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1884 {TGSI_OPCODE_PK4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1885 {TGSI_OPCODE_RFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1886 {TGSI_OPCODE_SEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
1887 {TGSI_OPCODE_SFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1888 {TGSI_OPCODE_SGT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
1889 {TGSI_OPCODE_SIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
1890 {TGSI_OPCODE_SLE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
1891 {TGSI_OPCODE_SNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
1892 {TGSI_OPCODE_STR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1893 {TGSI_OPCODE_TEX
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
1894 {TGSI_OPCODE_TXD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1895 {TGSI_OPCODE_TXP
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
1896 {TGSI_OPCODE_UP2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1897 {TGSI_OPCODE_UP2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1898 {TGSI_OPCODE_UP4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1899 {TGSI_OPCODE_UP4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1900 {TGSI_OPCODE_X2D
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1901 {TGSI_OPCODE_ARA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1902 {TGSI_OPCODE_ARR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1903 {TGSI_OPCODE_BRA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1904 {TGSI_OPCODE_CAL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1905 {TGSI_OPCODE_RET
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1906 {TGSI_OPCODE_SSG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
1907 {TGSI_OPCODE_CMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
1908 {TGSI_OPCODE_SCS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1909 {TGSI_OPCODE_TXB
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
1910 {TGSI_OPCODE_NRM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1911 {TGSI_OPCODE_DIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1912 {TGSI_OPCODE_DP2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
1913 {TGSI_OPCODE_TXL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1914 {TGSI_OPCODE_BRK
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1915 {TGSI_OPCODE_IF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
1917 {75, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1918 {76, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1919 {TGSI_OPCODE_ELSE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
1920 {TGSI_OPCODE_ENDIF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
1922 {79, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1923 {80, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1924 {TGSI_OPCODE_PUSHA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1925 {TGSI_OPCODE_POPA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1926 {TGSI_OPCODE_CEIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1927 {TGSI_OPCODE_I2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1928 {TGSI_OPCODE_NOT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1929 {TGSI_OPCODE_TRUNC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
1930 {TGSI_OPCODE_SHL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1932 {88, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1933 {TGSI_OPCODE_AND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1934 {TGSI_OPCODE_OR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1935 {TGSI_OPCODE_MOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1936 {TGSI_OPCODE_XOR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1937 {TGSI_OPCODE_SAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1938 {TGSI_OPCODE_TXF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1939 {TGSI_OPCODE_TXQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1940 {TGSI_OPCODE_CONT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1941 {TGSI_OPCODE_EMIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1942 {TGSI_OPCODE_ENDPRIM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1943 {TGSI_OPCODE_BGNLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1944 {TGSI_OPCODE_BGNSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1945 {TGSI_OPCODE_ENDLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1946 {TGSI_OPCODE_ENDSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1948 {103, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1949 {104, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1950 {105, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1951 {106, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1952 {TGSI_OPCODE_NOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1954 {108, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1955 {109, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1956 {110, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1957 {111, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1958 {TGSI_OPCODE_NRM4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1959 {TGSI_OPCODE_CALLNZ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1960 {TGSI_OPCODE_IFC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1961 {TGSI_OPCODE_BREAKC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1962 {TGSI_OPCODE_KIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
1963 {TGSI_OPCODE_END
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
1965 {118, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1966 {TGSI_OPCODE_F2I
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1967 {TGSI_OPCODE_IDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1968 {TGSI_OPCODE_IMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1969 {TGSI_OPCODE_IMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1970 {TGSI_OPCODE_INEG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1971 {TGSI_OPCODE_ISGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1972 {TGSI_OPCODE_ISHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1973 {TGSI_OPCODE_ISLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1974 {TGSI_OPCODE_F2U
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1975 {TGSI_OPCODE_U2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1976 {TGSI_OPCODE_UADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1977 {TGSI_OPCODE_UDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1978 {TGSI_OPCODE_UMAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1979 {TGSI_OPCODE_UMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1980 {TGSI_OPCODE_UMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1981 {TGSI_OPCODE_UMOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1982 {TGSI_OPCODE_UMUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1983 {TGSI_OPCODE_USEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1984 {TGSI_OPCODE_USGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1985 {TGSI_OPCODE_USHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1986 {TGSI_OPCODE_USLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1987 {TGSI_OPCODE_USNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1988 {TGSI_OPCODE_SWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1989 {TGSI_OPCODE_CASE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1990 {TGSI_OPCODE_DEFAULT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1991 {TGSI_OPCODE_ENDSWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
1992 {TGSI_OPCODE_LAST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},