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 for (i
= 0; i
< 10; i
++) {
50 for (i
= 0; i
< 32; i
++) {
51 tmp
= i
<< ((i
& 3) * 8);
52 spi_vs_out_id
[i
/ 4] |= tmp
;
54 for (i
= 0; i
< 10; i
++) {
55 r600_pipe_state_add_reg(rstate
,
56 R_028614_SPI_VS_OUT_ID_0
+ i
* 4,
57 spi_vs_out_id
[i
], 0xFFFFFFFF, NULL
);
60 r600_pipe_state_add_reg(rstate
,
61 R_0286C4_SPI_VS_OUT_CONFIG
,
62 S_0286C4_VS_EXPORT_COUNT(rshader
->noutput
- 2),
64 r600_pipe_state_add_reg(rstate
,
65 R_028868_SQ_PGM_RESOURCES_VS
,
66 S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
67 S_028868_STACK_SIZE(rshader
->bc
.nstack
),
69 r600_pipe_state_add_reg(rstate
,
70 R_0288A4_SQ_PGM_RESOURCES_FS
,
71 0x00000000, 0xFFFFFFFF, NULL
);
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_0288DC_SQ_PGM_CF_OFFSET_FS
,
77 0x00000000, 0xFFFFFFFF, NULL
);
78 r600_pipe_state_add_reg(rstate
,
79 R_028858_SQ_PGM_START_VS
,
80 r600_bo_offset(shader
->bo
) >> 8, 0xFFFFFFFF, shader
->bo
);
81 r600_pipe_state_add_reg(rstate
,
82 R_028894_SQ_PGM_START_FS
,
83 r600_bo_offset(shader
->bo_fetch
) >> 8, 0xFFFFFFFF, shader
->bo_fetch
);
85 r600_pipe_state_add_reg(rstate
,
86 R_03E200_SQ_LOOP_CONST_0
+ (32 * 4), 0x01000FFF,
91 int r600_find_vs_semantic_index(struct r600_shader
*vs
,
92 struct r600_shader
*ps
, int id
)
94 struct r600_shader_io
*input
= &ps
->input
[id
];
96 for (int i
= 0; i
< vs
->noutput
; i
++) {
97 if (input
->name
== vs
->output
[i
].name
&&
98 input
->sid
== vs
->output
[i
].sid
) {
105 static void r600_pipe_shader_ps(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
107 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
108 struct r600_pipe_state
*rstate
= &shader
->rstate
;
109 struct r600_shader
*rshader
= &shader
->shader
;
110 unsigned i
, tmp
, exports_ps
, num_cout
, spi_ps_in_control_0
, spi_input_z
, spi_ps_in_control_1
;
111 int pos_index
= -1, face_index
= -1;
113 /* clear previous register */
116 for (i
= 0; i
< rshader
->ninput
; i
++) {
117 tmp
= S_028644_SEMANTIC(r600_find_vs_semantic_index(&rctx
->vs_shader
->shader
, rshader
, i
));
118 if (rshader
->input
[i
].centroid
)
119 tmp
|= S_028644_SEL_CENTROID(1);
120 if (rshader
->input
[i
].interpolate
== TGSI_INTERPOLATE_LINEAR
)
121 tmp
|= S_028644_SEL_LINEAR(1);
123 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_POSITION
)
125 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_COLOR
||
126 rshader
->input
[i
].name
== TGSI_SEMANTIC_BCOLOR
||
127 rshader
->input
[i
].name
== TGSI_SEMANTIC_POSITION
) {
128 tmp
|= S_028644_FLAT_SHADE(rshader
->flat_shade
);
130 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_FACE
)
132 if (rshader
->input
[i
].name
== TGSI_SEMANTIC_GENERIC
&&
133 rctx
->sprite_coord_enable
& (1 << rshader
->input
[i
].sid
)) {
134 tmp
|= S_028644_PT_SPRITE_TEX(1);
136 r600_pipe_state_add_reg(rstate
, R_028644_SPI_PS_INPUT_CNTL_0
+ i
* 4, tmp
, 0xFFFFFFFF, NULL
);
138 for (i
= 0; i
< rshader
->noutput
; i
++) {
139 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
)
140 r600_pipe_state_add_reg(rstate
,
141 R_02880C_DB_SHADER_CONTROL
,
142 S_02880C_Z_EXPORT_ENABLE(1),
143 S_02880C_Z_EXPORT_ENABLE(1), NULL
);
144 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
)
145 r600_pipe_state_add_reg(rstate
,
146 R_02880C_DB_SHADER_CONTROL
,
147 S_02880C_STENCIL_REF_EXPORT_ENABLE(1),
148 S_02880C_STENCIL_REF_EXPORT_ENABLE(1), NULL
);
153 for (i
= 0; i
< rshader
->noutput
; i
++) {
154 if (rshader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
|| rshader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
)
156 else if (rshader
->output
[i
].name
== TGSI_SEMANTIC_COLOR
) {
160 exports_ps
|= S_028854_EXPORT_COLORS(num_cout
);
162 /* always at least export 1 component per pixel */
166 spi_ps_in_control_0
= S_0286CC_NUM_INTERP(rshader
->ninput
) |
167 S_0286CC_PERSP_GRADIENT_ENA(1);
169 if (pos_index
!= -1) {
170 spi_ps_in_control_0
|= (S_0286CC_POSITION_ENA(1) |
171 S_0286CC_POSITION_CENTROID(rshader
->input
[pos_index
].centroid
) |
172 S_0286CC_POSITION_ADDR(rshader
->input
[pos_index
].gpr
) |
173 S_0286CC_BARYC_SAMPLE_CNTL(1));
177 spi_ps_in_control_1
= 0;
178 if (face_index
!= -1) {
179 spi_ps_in_control_1
|= S_0286D0_FRONT_FACE_ENA(1) |
180 S_0286D0_FRONT_FACE_ADDR(rshader
->input
[face_index
].gpr
);
183 r600_pipe_state_add_reg(rstate
, R_0286CC_SPI_PS_IN_CONTROL_0
, spi_ps_in_control_0
, 0xFFFFFFFF, NULL
);
184 r600_pipe_state_add_reg(rstate
, R_0286D0_SPI_PS_IN_CONTROL_1
, spi_ps_in_control_1
, 0xFFFFFFFF, NULL
);
185 r600_pipe_state_add_reg(rstate
, R_0286D8_SPI_INPUT_Z
, spi_input_z
, 0xFFFFFFFF, NULL
);
186 r600_pipe_state_add_reg(rstate
,
187 R_028840_SQ_PGM_START_PS
,
188 r600_bo_offset(shader
->bo
) >> 8, 0xFFFFFFFF, shader
->bo
);
189 r600_pipe_state_add_reg(rstate
,
190 R_028850_SQ_PGM_RESOURCES_PS
,
191 S_028868_NUM_GPRS(rshader
->bc
.ngpr
) |
192 S_028868_STACK_SIZE(rshader
->bc
.nstack
),
194 r600_pipe_state_add_reg(rstate
,
195 R_028854_SQ_PGM_EXPORTS_PS
,
196 exports_ps
, 0xFFFFFFFF, NULL
);
197 r600_pipe_state_add_reg(rstate
,
198 R_0288CC_SQ_PGM_CF_OFFSET_PS
,
199 0x00000000, 0xFFFFFFFF, NULL
);
201 if (rshader
->uses_kill
) {
202 /* only set some bits here, the other bits are set in the dsa state */
203 r600_pipe_state_add_reg(rstate
,
204 R_02880C_DB_SHADER_CONTROL
,
205 S_02880C_KILL_ENABLE(1),
206 S_02880C_KILL_ENABLE(1), NULL
);
208 r600_pipe_state_add_reg(rstate
,
209 R_03E200_SQ_LOOP_CONST_0
, 0x01000FFF,
213 static int r600_pipe_shader(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
215 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
216 struct r600_shader
*rshader
= &shader
->shader
;
219 /* copy new shader */
220 if (rshader
->processor_type
== TGSI_PROCESSOR_VERTEX
&& shader
->bo_fetch
== NULL
) {
221 shader
->bo_fetch
= r600_bo(rctx
->radeon
, rshader
->bc_fetch
.ndw
* 4, 4096, 0, 0);
222 if (shader
->bo_fetch
== NULL
) {
225 ptr
= r600_bo_map(rctx
->radeon
, shader
->bo_fetch
, 0, NULL
);
226 memcpy(ptr
, rshader
->bc_fetch
.bytecode
, rshader
->bc_fetch
.ndw
* 4);
227 r600_bo_unmap(rctx
->radeon
, shader
->bo_fetch
);
229 if (shader
->bo
== NULL
) {
230 shader
->bo
= r600_bo(rctx
->radeon
, rshader
->bc
.ndw
* 4, 4096, 0, 0);
231 if (shader
->bo
== NULL
) {
234 ptr
= r600_bo_map(rctx
->radeon
, shader
->bo
, 0, NULL
);
235 memcpy(ptr
, rshader
->bc
.bytecode
, rshader
->bc
.ndw
* 4);
236 r600_bo_unmap(rctx
->radeon
, shader
->bo
);
239 rshader
->flat_shade
= rctx
->flatshade
;
240 switch (rshader
->processor_type
) {
241 case TGSI_PROCESSOR_VERTEX
:
242 if (rshader
->family
>= CHIP_CEDAR
) {
243 evergreen_pipe_shader_vs(ctx
, shader
);
245 r600_pipe_shader_vs(ctx
, shader
);
248 case TGSI_PROCESSOR_FRAGMENT
:
249 if (rshader
->family
>= CHIP_CEDAR
) {
250 evergreen_pipe_shader_ps(ctx
, shader
);
252 r600_pipe_shader_ps(ctx
, shader
);
258 r600_context_pipe_state_set(&rctx
->ctx
, &shader
->rstate
);
262 static int r600_shader_update(struct pipe_context
*ctx
, struct r600_pipe_shader
*rshader
)
264 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
265 struct r600_shader
*shader
= &rshader
->shader
;
266 const struct util_format_description
*desc
;
267 enum pipe_format resource_format
[160];
268 unsigned i
, nresources
= 0;
269 struct r600_bc
*bc
= &shader
->bc_fetch
;
270 struct r600_bc_cf
*cf
;
271 struct r600_bc_vtx
*vtx
;
273 if (shader
->processor_type
!= TGSI_PROCESSOR_VERTEX
)
275 /* doing a full memcmp fell over the refcount */
276 if ((rshader
->vertex_elements
.count
== rctx
->vertex_elements
->count
) &&
277 (!memcmp(&rshader
->vertex_elements
.elements
, &rctx
->vertex_elements
->elements
,
278 rctx
->vertex_elements
->count
* sizeof(struct pipe_vertex_element
)))) {
281 rshader
->vertex_elements
= *rctx
->vertex_elements
;
282 for (i
= 0; i
< rctx
->vertex_elements
->count
; i
++) {
283 resource_format
[nresources
++] = rctx
->vertex_elements
->hw_format
[i
];
285 r600_bo_reference(rctx
->radeon
, &rshader
->bo_fetch
, NULL
);
286 LIST_FOR_EACH_ENTRY(cf
, &bc
->cf
, list
) {
288 case V_SQ_CF_WORD1_SQ_CF_INST_VTX
:
289 case V_SQ_CF_WORD1_SQ_CF_INST_VTX_TC
:
290 LIST_FOR_EACH_ENTRY(vtx
, &cf
->vtx
, list
) {
291 desc
= util_format_description(resource_format
[vtx
->buffer_id
]);
293 R600_ERR("unknown format %d\n", resource_format
[vtx
->buffer_id
]);
296 vtx
->dst_sel_x
= desc
->swizzle
[0];
297 vtx
->dst_sel_y
= desc
->swizzle
[1];
298 vtx
->dst_sel_z
= desc
->swizzle
[2];
299 vtx
->dst_sel_w
= desc
->swizzle
[3];
306 return r600_bc_build(&shader
->bc_fetch
);
309 int r600_pipe_shader_update(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
311 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
316 /* there should be enough input */
317 if (rctx
->vertex_elements
->count
< shader
->shader
.bc
.nresource
) {
318 R600_ERR("%d resources provided, expecting %d\n",
319 rctx
->vertex_elements
->count
, shader
->shader
.bc
.nresource
);
322 r
= r600_shader_update(ctx
, shader
);
325 return r600_pipe_shader(ctx
, shader
);
328 int r600_shader_from_tgsi(const struct tgsi_token
*tokens
, struct r600_shader
*shader
);
329 int r600_pipe_shader_create(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
, const struct tgsi_token
*tokens
)
331 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
334 //fprintf(stderr, "--------------------------------------------------------------\n");
335 //tgsi_dump(tokens, 0);
336 shader
->shader
.family
= r600_get_family(rctx
->radeon
);
337 r
= r600_shader_from_tgsi(tokens
, &shader
->shader
);
339 R600_ERR("translation from TGSI failed !\n");
342 r
= r600_bc_build(&shader
->shader
.bc
);
344 R600_ERR("building bytecode failed !\n");
347 if (shader
->shader
.processor_type
== TGSI_PROCESSOR_VERTEX
) {
348 r
= r600_bc_build(&shader
->shader
.bc_fetch
);
350 R600_ERR("building bytecode failed !\n");
354 //fprintf(stderr, "______________________________________________________________\n");
359 r600_pipe_shader_destroy(struct pipe_context
*ctx
, struct r600_pipe_shader
*shader
)
361 struct r600_pipe_context
*rctx
= (struct r600_pipe_context
*)ctx
;
363 r600_bo_reference(rctx
->radeon
, &shader
->bo
, NULL
);
365 r600_bc_clear(&shader
->shader
.bc
);
367 /* FIXME: is there more stuff to free? */
371 * tgsi -> r600 shader
373 struct r600_shader_tgsi_instruction
;
375 struct r600_shader_ctx
{
376 struct tgsi_shader_info info
;
377 struct tgsi_parse_context parse
;
378 const struct tgsi_token
*tokens
;
380 unsigned file_offset
[TGSI_FILE_COUNT
];
382 struct r600_shader_tgsi_instruction
*inst_info
;
384 struct r600_bc
*bc_fetch
;
385 struct r600_shader
*shader
;
389 u32 max_driver_temp_used
;
390 /* needed for evergreen interpolation */
391 boolean input_centroid
;
392 boolean input_linear
;
393 boolean input_perspective
;
397 struct r600_shader_tgsi_instruction
{
398 unsigned tgsi_opcode
;
400 unsigned r600_opcode
;
401 int (*process
)(struct r600_shader_ctx
*ctx
);
404 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[], eg_shader_tgsi_instruction
[];
405 static int tgsi_helper_tempx_replicate(struct r600_shader_ctx
*ctx
);
407 static int tgsi_is_supported(struct r600_shader_ctx
*ctx
)
409 struct tgsi_full_instruction
*i
= &ctx
->parse
.FullToken
.FullInstruction
;
412 if (i
->Instruction
.NumDstRegs
> 1) {
413 R600_ERR("too many dst (%d)\n", i
->Instruction
.NumDstRegs
);
416 if (i
->Instruction
.Predicate
) {
417 R600_ERR("predicate unsupported\n");
421 if (i
->Instruction
.Label
) {
422 R600_ERR("label unsupported\n");
426 for (j
= 0; j
< i
->Instruction
.NumSrcRegs
; j
++) {
427 if (i
->Src
[j
].Register
.Dimension
) {
428 R600_ERR("unsupported src %d (dimension %d)\n", j
,
429 i
->Src
[j
].Register
.Dimension
);
433 for (j
= 0; j
< i
->Instruction
.NumDstRegs
; j
++) {
434 if (i
->Dst
[j
].Register
.Dimension
) {
435 R600_ERR("unsupported dst (dimension)\n");
442 static int evergreen_interp_alu(struct r600_shader_ctx
*ctx
, int input
)
445 struct r600_bc_alu alu
;
446 int gpr
= 0, base_chan
= 0;
449 if (ctx
->shader
->input
[input
].interpolate
== TGSI_INTERPOLATE_PERSPECTIVE
) {
451 if (ctx
->shader
->input
[input
].centroid
)
453 } else if (ctx
->shader
->input
[input
].interpolate
== TGSI_INTERPOLATE_LINEAR
) {
455 /* if we have perspective add one */
456 if (ctx
->input_perspective
) {
458 /* if we have perspective centroid */
459 if (ctx
->input_centroid
)
462 if (ctx
->shader
->input
[input
].centroid
)
466 /* work out gpr and base_chan from index */
468 base_chan
= (2 * (ij_index
% 2)) + 1;
470 for (i
= 0; i
< 8; i
++) {
471 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
474 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_ZW
;
476 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_XY
;
478 if ((i
> 1) && (i
< 6)) {
479 alu
.dst
.sel
= ctx
->shader
->input
[input
].gpr
;
483 alu
.dst
.chan
= i
% 4;
485 alu
.src
[0].sel
= gpr
;
486 alu
.src
[0].chan
= (base_chan
- (i
% 2));
488 alu
.src
[1].sel
= V_SQ_ALU_SRC_PARAM_BASE
+ ctx
->shader
->input
[input
].lds_pos
;
490 alu
.bank_swizzle_force
= SQ_ALU_VEC_210
;
493 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
501 static int tgsi_declaration(struct r600_shader_ctx
*ctx
)
503 struct tgsi_full_declaration
*d
= &ctx
->parse
.FullToken
.FullDeclaration
;
504 struct r600_bc_vtx vtx
;
508 switch (d
->Declaration
.File
) {
509 case TGSI_FILE_INPUT
:
510 i
= ctx
->shader
->ninput
++;
511 ctx
->shader
->input
[i
].name
= d
->Semantic
.Name
;
512 ctx
->shader
->input
[i
].sid
= d
->Semantic
.Index
;
513 ctx
->shader
->input
[i
].interpolate
= d
->Declaration
.Interpolate
;
514 ctx
->shader
->input
[i
].centroid
= d
->Declaration
.Centroid
;
515 ctx
->shader
->input
[i
].gpr
= ctx
->file_offset
[TGSI_FILE_INPUT
] + i
;
516 if (ctx
->type
== TGSI_PROCESSOR_VERTEX
) {
517 /* turn input into fetch */
518 memset(&vtx
, 0, sizeof(struct r600_bc_vtx
));
522 /* register containing the index into the buffer */
525 vtx
.mega_fetch_count
= 0x1F;
526 vtx
.dst_gpr
= ctx
->shader
->input
[i
].gpr
;
531 vtx
.use_const_fields
= 1;
532 r
= r600_bc_add_vtx(ctx
->bc_fetch
, &vtx
);
536 if (ctx
->type
== TGSI_PROCESSOR_FRAGMENT
&& ctx
->bc
->chiprev
== CHIPREV_EVERGREEN
) {
537 /* turn input into interpolate on EG */
538 if (ctx
->shader
->input
[i
].name
!= TGSI_SEMANTIC_POSITION
) {
539 if (ctx
->shader
->input
[i
].interpolate
> 0) {
540 ctx
->shader
->input
[i
].lds_pos
= ctx
->shader
->nlds
++;
541 evergreen_interp_alu(ctx
, i
);
546 case TGSI_FILE_OUTPUT
:
547 i
= ctx
->shader
->noutput
++;
548 ctx
->shader
->output
[i
].name
= d
->Semantic
.Name
;
549 ctx
->shader
->output
[i
].sid
= d
->Semantic
.Index
;
550 ctx
->shader
->output
[i
].gpr
= ctx
->file_offset
[TGSI_FILE_OUTPUT
] + i
;
551 ctx
->shader
->output
[i
].interpolate
= d
->Declaration
.Interpolate
;
553 case TGSI_FILE_CONSTANT
:
554 case TGSI_FILE_TEMPORARY
:
555 case TGSI_FILE_SAMPLER
:
556 case TGSI_FILE_ADDRESS
:
559 R600_ERR("unsupported file %d declaration\n", d
->Declaration
.File
);
565 static int r600_get_temp(struct r600_shader_ctx
*ctx
)
567 return ctx
->temp_reg
+ ctx
->max_driver_temp_used
++;
571 * for evergreen we need to scan the shader to find the number of GPRs we need to
572 * reserve for interpolation.
574 * we need to know if we are going to emit
575 * any centroid inputs
576 * if perspective and linear are required
578 static int evergreen_gpr_count(struct r600_shader_ctx
*ctx
)
583 ctx
->input_linear
= FALSE
;
584 ctx
->input_perspective
= FALSE
;
585 ctx
->input_centroid
= FALSE
;
586 ctx
->num_interp_gpr
= 1;
588 /* any centroid inputs */
589 for (i
= 0; i
< ctx
->info
.num_inputs
; i
++) {
590 /* skip position/face */
591 if (ctx
->info
.input_semantic_name
[i
] == TGSI_SEMANTIC_POSITION
||
592 ctx
->info
.input_semantic_name
[i
] == TGSI_SEMANTIC_FACE
)
594 if (ctx
->info
.input_interpolate
[i
] == TGSI_INTERPOLATE_LINEAR
)
595 ctx
->input_linear
= TRUE
;
596 if (ctx
->info
.input_interpolate
[i
] == TGSI_INTERPOLATE_PERSPECTIVE
)
597 ctx
->input_perspective
= TRUE
;
598 if (ctx
->info
.input_centroid
[i
])
599 ctx
->input_centroid
= TRUE
;
603 /* ignoring sample for now */
604 if (ctx
->input_perspective
)
606 if (ctx
->input_linear
)
608 if (ctx
->input_centroid
)
611 ctx
->num_interp_gpr
+= (num_baryc
+ 1) >> 1;
613 /* TODO PULL MODEL and LINE STIPPLE, FIXED PT POS */
614 return ctx
->num_interp_gpr
;
617 int r600_shader_from_tgsi(const struct tgsi_token
*tokens
, struct r600_shader
*shader
)
619 struct tgsi_full_immediate
*immediate
;
620 struct r600_shader_ctx ctx
;
621 struct r600_bc_output output
[32];
622 unsigned output_done
, noutput
;
626 ctx
.bc
= &shader
->bc
;
627 ctx
.bc_fetch
= &shader
->bc_fetch
;
629 r
= r600_bc_init(ctx
.bc
, shader
->family
);
633 tgsi_scan_shader(tokens
, &ctx
.info
);
634 tgsi_parse_init(&ctx
.parse
, tokens
);
635 ctx
.type
= ctx
.parse
.FullHeader
.Processor
.Processor
;
636 shader
->processor_type
= ctx
.type
;
637 if (shader
->processor_type
== TGSI_PROCESSOR_VERTEX
) {
638 r
= r600_bc_init(ctx
.bc_fetch
, shader
->family
);
641 ctx
.bc_fetch
->type
= -1;
643 ctx
.bc
->type
= shader
->processor_type
;
645 /* register allocations */
646 /* Values [0,127] correspond to GPR[0..127].
647 * Values [128,159] correspond to constant buffer bank 0
648 * Values [160,191] correspond to constant buffer bank 1
649 * Values [256,511] correspond to cfile constants c[0..255].
650 * Other special values are shown in the list below.
651 * 244 ALU_SRC_1_DBL_L: special constant 1.0 double-float, LSW. (RV670+)
652 * 245 ALU_SRC_1_DBL_M: special constant 1.0 double-float, MSW. (RV670+)
653 * 246 ALU_SRC_0_5_DBL_L: special constant 0.5 double-float, LSW. (RV670+)
654 * 247 ALU_SRC_0_5_DBL_M: special constant 0.5 double-float, MSW. (RV670+)
655 * 248 SQ_ALU_SRC_0: special constant 0.0.
656 * 249 SQ_ALU_SRC_1: special constant 1.0 float.
657 * 250 SQ_ALU_SRC_1_INT: special constant 1 integer.
658 * 251 SQ_ALU_SRC_M_1_INT: special constant -1 integer.
659 * 252 SQ_ALU_SRC_0_5: special constant 0.5 float.
660 * 253 SQ_ALU_SRC_LITERAL: literal constant.
661 * 254 SQ_ALU_SRC_PV: previous vector result.
662 * 255 SQ_ALU_SRC_PS: previous scalar result.
664 for (i
= 0; i
< TGSI_FILE_COUNT
; i
++) {
665 ctx
.file_offset
[i
] = 0;
667 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
668 ctx
.file_offset
[TGSI_FILE_INPUT
] = 1;
669 if (ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
) {
670 r600_bc_add_cfinst(ctx
.bc
, EG_V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS
);
672 r600_bc_add_cfinst(ctx
.bc
, V_SQ_CF_WORD1_SQ_CF_INST_CALL_FS
);
675 if (ctx
.type
== TGSI_PROCESSOR_FRAGMENT
&& ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
) {
676 ctx
.file_offset
[TGSI_FILE_INPUT
] = evergreen_gpr_count(&ctx
);
678 ctx
.file_offset
[TGSI_FILE_OUTPUT
] = ctx
.file_offset
[TGSI_FILE_INPUT
] +
679 ctx
.info
.file_count
[TGSI_FILE_INPUT
];
680 ctx
.file_offset
[TGSI_FILE_TEMPORARY
] = ctx
.file_offset
[TGSI_FILE_OUTPUT
] +
681 ctx
.info
.file_count
[TGSI_FILE_OUTPUT
];
683 ctx
.file_offset
[TGSI_FILE_CONSTANT
] = 128;
685 ctx
.file_offset
[TGSI_FILE_IMMEDIATE
] = 253;
686 ctx
.temp_reg
= ctx
.file_offset
[TGSI_FILE_TEMPORARY
] +
687 ctx
.info
.file_count
[TGSI_FILE_TEMPORARY
];
692 while (!tgsi_parse_end_of_tokens(&ctx
.parse
)) {
693 tgsi_parse_token(&ctx
.parse
);
694 switch (ctx
.parse
.FullToken
.Token
.Type
) {
695 case TGSI_TOKEN_TYPE_IMMEDIATE
:
696 immediate
= &ctx
.parse
.FullToken
.FullImmediate
;
697 ctx
.literals
= realloc(ctx
.literals
, (ctx
.nliterals
+ 1) * 16);
698 if(ctx
.literals
== NULL
) {
702 ctx
.literals
[ctx
.nliterals
* 4 + 0] = immediate
->u
[0].Uint
;
703 ctx
.literals
[ctx
.nliterals
* 4 + 1] = immediate
->u
[1].Uint
;
704 ctx
.literals
[ctx
.nliterals
* 4 + 2] = immediate
->u
[2].Uint
;
705 ctx
.literals
[ctx
.nliterals
* 4 + 3] = immediate
->u
[3].Uint
;
708 case TGSI_TOKEN_TYPE_DECLARATION
:
709 r
= tgsi_declaration(&ctx
);
713 case TGSI_TOKEN_TYPE_INSTRUCTION
:
714 r
= tgsi_is_supported(&ctx
);
717 ctx
.max_driver_temp_used
= 0;
718 /* reserve first tmp for everyone */
720 opcode
= ctx
.parse
.FullToken
.FullInstruction
.Instruction
.Opcode
;
721 if (ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
)
722 ctx
.inst_info
= &eg_shader_tgsi_instruction
[opcode
];
724 ctx
.inst_info
= &r600_shader_tgsi_instruction
[opcode
];
725 r
= ctx
.inst_info
->process(&ctx
);
728 r
= r600_bc_add_literal(ctx
.bc
, ctx
.value
);
733 R600_ERR("unsupported token type %d\n", ctx
.parse
.FullToken
.Token
.Type
);
739 noutput
= shader
->noutput
;
740 for (i
= 0, pos0
= 0; i
< noutput
; i
++) {
741 memset(&output
[i
], 0, sizeof(struct r600_bc_output
));
742 output
[i
].gpr
= shader
->output
[i
].gpr
;
743 output
[i
].elem_size
= 3;
744 output
[i
].swizzle_x
= 0;
745 output
[i
].swizzle_y
= 1;
746 output
[i
].swizzle_z
= 2;
747 output
[i
].swizzle_w
= 3;
748 output
[i
].barrier
= 1;
749 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
;
750 output
[i
].array_base
= i
- pos0
;
751 output
[i
].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
);
753 case TGSI_PROCESSOR_VERTEX
:
754 if (shader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
) {
755 output
[i
].array_base
= 60;
756 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS
;
757 /* position doesn't count in array_base */
760 if (shader
->output
[i
].name
== TGSI_SEMANTIC_PSIZE
) {
761 output
[i
].array_base
= 61;
762 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_POS
;
763 /* position doesn't count in array_base */
767 case TGSI_PROCESSOR_FRAGMENT
:
768 if (shader
->output
[i
].name
== TGSI_SEMANTIC_COLOR
) {
769 output
[i
].array_base
= shader
->output
[i
].sid
;
770 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
771 } else if (shader
->output
[i
].name
== TGSI_SEMANTIC_POSITION
) {
772 output
[i
].array_base
= 61;
773 output
[i
].swizzle_x
= 2;
774 output
[i
].swizzle_y
= 7;
775 output
[i
].swizzle_z
= output
[i
].swizzle_w
= 7;
776 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
777 } else if (shader
->output
[i
].name
== TGSI_SEMANTIC_STENCIL
) {
778 output
[i
].array_base
= 61;
779 output
[i
].swizzle_x
= 7;
780 output
[i
].swizzle_y
= 1;
781 output
[i
].swizzle_z
= output
[i
].swizzle_w
= 7;
782 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
784 R600_ERR("unsupported fragment output name %d\n", shader
->output
[i
].name
);
790 R600_ERR("unsupported processor type %d\n", ctx
.type
);
795 /* add fake param output for vertex shader if no param is exported */
796 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
797 for (i
= 0, pos0
= 0; i
< noutput
; i
++) {
798 if (output
[i
].type
== V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
) {
804 memset(&output
[i
], 0, sizeof(struct r600_bc_output
));
806 output
[i
].elem_size
= 3;
807 output
[i
].swizzle_x
= 0;
808 output
[i
].swizzle_y
= 1;
809 output
[i
].swizzle_z
= 2;
810 output
[i
].swizzle_w
= 3;
811 output
[i
].barrier
= 1;
812 output
[i
].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PARAM
;
813 output
[i
].array_base
= 0;
814 output
[i
].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
);
818 /* add fake pixel export */
819 if (ctx
.type
== TGSI_PROCESSOR_FRAGMENT
&& !noutput
) {
820 memset(&output
[0], 0, sizeof(struct r600_bc_output
));
822 output
[0].elem_size
= 3;
823 output
[0].swizzle_x
= 7;
824 output
[0].swizzle_y
= 7;
825 output
[0].swizzle_z
= 7;
826 output
[0].swizzle_w
= 7;
827 output
[0].barrier
= 1;
828 output
[0].type
= V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL
;
829 output
[0].array_base
= 0;
830 output
[0].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT
);
833 /* set export done on last export of each type */
834 for (i
= noutput
- 1, output_done
= 0; i
>= 0; i
--) {
835 if (i
== (noutput
- 1)) {
836 output
[i
].end_of_program
= 1;
838 if (!(output_done
& (1 << output
[i
].type
))) {
839 output_done
|= (1 << output
[i
].type
);
840 output
[i
].inst
= BC_INST(ctx
.bc
, V_SQ_CF_ALLOC_EXPORT_WORD1_SQ_CF_INST_EXPORT_DONE
);
843 /* add return to fetch shader */
844 if (ctx
.type
== TGSI_PROCESSOR_VERTEX
) {
845 if (ctx
.bc
->chiprev
== CHIPREV_EVERGREEN
) {
846 r600_bc_add_cfinst(ctx
.bc_fetch
, EG_V_SQ_CF_WORD1_SQ_CF_INST_RETURN
);
848 r600_bc_add_cfinst(ctx
.bc_fetch
, V_SQ_CF_WORD1_SQ_CF_INST_RETURN
);
851 /* add output to bytecode */
852 for (i
= 0; i
< noutput
; i
++) {
853 r
= r600_bc_add_output(ctx
.bc
, &output
[i
]);
858 tgsi_parse_free(&ctx
.parse
);
862 tgsi_parse_free(&ctx
.parse
);
866 static int tgsi_unsupported(struct r600_shader_ctx
*ctx
)
868 R600_ERR("%d tgsi opcode unsupported\n", ctx
->inst_info
->tgsi_opcode
);
872 static int tgsi_end(struct r600_shader_ctx
*ctx
)
877 static int tgsi_src(struct r600_shader_ctx
*ctx
,
878 const struct tgsi_full_src_register
*tgsi_src
,
879 struct r600_bc_alu_src
*r600_src
)
882 memset(r600_src
, 0, sizeof(struct r600_bc_alu_src
));
883 r600_src
->sel
= tgsi_src
->Register
.Index
;
884 if (tgsi_src
->Register
.File
== TGSI_FILE_IMMEDIATE
) {
886 index
= tgsi_src
->Register
.Index
;
887 ctx
->value
[0] = ctx
->literals
[index
* 4 + 0];
888 ctx
->value
[1] = ctx
->literals
[index
* 4 + 1];
889 ctx
->value
[2] = ctx
->literals
[index
* 4 + 2];
890 ctx
->value
[3] = ctx
->literals
[index
* 4 + 3];
892 if (tgsi_src
->Register
.Indirect
)
893 r600_src
->rel
= V_SQ_REL_RELATIVE
;
894 r600_src
->neg
= tgsi_src
->Register
.Negate
;
895 r600_src
->abs
= tgsi_src
->Register
.Absolute
;
896 r600_src
->sel
+= ctx
->file_offset
[tgsi_src
->Register
.File
];
900 static int tgsi_dst(struct r600_shader_ctx
*ctx
,
901 const struct tgsi_full_dst_register
*tgsi_dst
,
903 struct r600_bc_alu_dst
*r600_dst
)
905 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
907 r600_dst
->sel
= tgsi_dst
->Register
.Index
;
908 r600_dst
->sel
+= ctx
->file_offset
[tgsi_dst
->Register
.File
];
909 r600_dst
->chan
= swizzle
;
911 if (tgsi_dst
->Register
.Indirect
)
912 r600_dst
->rel
= V_SQ_REL_RELATIVE
;
913 if (inst
->Instruction
.Saturate
) {
919 static unsigned tgsi_chan(const struct tgsi_full_src_register
*tgsi_src
, unsigned swizzle
)
923 return tgsi_src
->Register
.SwizzleX
;
925 return tgsi_src
->Register
.SwizzleY
;
927 return tgsi_src
->Register
.SwizzleZ
;
929 return tgsi_src
->Register
.SwizzleW
;
935 static int tgsi_split_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
937 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
938 struct r600_bc_alu alu
;
939 int i
, j
, k
, nconst
, r
;
941 for (i
= 0, nconst
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
942 if (inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
945 r
= tgsi_src(ctx
, &inst
->Src
[i
], &r600_src
[i
]);
950 for (i
= 0, j
= nconst
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
951 if (j
> 0 && inst
->Src
[i
].Register
.File
== TGSI_FILE_CONSTANT
) {
952 int treg
= r600_get_temp(ctx
);
953 for (k
= 0; k
< 4; k
++) {
954 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
955 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
956 alu
.src
[0].sel
= r600_src
[i
].sel
;
958 alu
.src
[0].rel
= r600_src
[i
].rel
;
964 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
968 r600_src
[i
].sel
= treg
;
976 /* need to move any immediate into a temp - for trig functions which use literal for PI stuff */
977 static int tgsi_split_literal_constant(struct r600_shader_ctx
*ctx
, struct r600_bc_alu_src r600_src
[3])
979 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
980 struct r600_bc_alu alu
;
981 int i
, j
, k
, nliteral
, r
;
983 for (i
= 0, nliteral
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
984 if (inst
->Src
[i
].Register
.File
== TGSI_FILE_IMMEDIATE
) {
988 for (i
= 0, j
= nliteral
- 1; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
989 if (j
> 0 && inst
->Src
[i
].Register
.File
== TGSI_FILE_IMMEDIATE
) {
990 int treg
= r600_get_temp(ctx
);
991 for (k
= 0; k
< 4; k
++) {
992 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
993 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
994 alu
.src
[0].sel
= r600_src
[i
].sel
;
1001 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1005 r
= r600_bc_add_literal(ctx
->bc
, &ctx
->literals
[inst
->Src
[i
].Register
.Index
* 4]);
1008 r600_src
[i
].sel
= treg
;
1015 static int tgsi_op2_s(struct r600_shader_ctx
*ctx
, int swap
)
1017 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1018 struct r600_bc_alu_src r600_src
[3];
1019 struct r600_bc_alu alu
;
1023 for (i
= 0; i
< 4; i
++) {
1024 if (inst
->Dst
[0].Register
.WriteMask
& (1 << i
)) {
1029 r
= tgsi_split_constant(ctx
, r600_src
);
1032 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1035 for (i
= 0; i
< lasti
+ 1; i
++) {
1036 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1039 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1040 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1044 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1046 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1047 alu
.src
[j
] = r600_src
[j
];
1048 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1051 alu
.src
[0] = r600_src
[1];
1052 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], i
);
1054 alu
.src
[1] = r600_src
[0];
1055 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1057 /* handle some special cases */
1058 switch (ctx
->inst_info
->tgsi_opcode
) {
1059 case TGSI_OPCODE_SUB
:
1062 case TGSI_OPCODE_ABS
:
1071 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1078 static int tgsi_op2(struct r600_shader_ctx
*ctx
)
1080 return tgsi_op2_s(ctx
, 0);
1083 static int tgsi_op2_swap(struct r600_shader_ctx
*ctx
)
1085 return tgsi_op2_s(ctx
, 1);
1089 * r600 - trunc to -PI..PI range
1090 * r700 - normalize by dividing by 2PI
1093 static int tgsi_setup_trig(struct r600_shader_ctx
*ctx
,
1094 struct r600_bc_alu_src r600_src
[3])
1096 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1098 uint32_t lit_vals
[4];
1099 struct r600_bc_alu alu
;
1101 memset(lit_vals
, 0, 4*4);
1102 r
= tgsi_split_constant(ctx
, r600_src
);
1105 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1109 lit_vals
[0] = fui(1.0 /(3.1415926535 * 2));
1110 lit_vals
[1] = fui(0.5f
);
1112 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1113 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1117 alu
.dst
.sel
= ctx
->temp_reg
;
1120 alu
.src
[0] = r600_src
[0];
1121 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1123 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
1124 alu
.src
[1].chan
= 0;
1125 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1126 alu
.src
[2].chan
= 1;
1128 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1131 r
= r600_bc_add_literal(ctx
->bc
, lit_vals
);
1135 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1136 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
);
1139 alu
.dst
.sel
= ctx
->temp_reg
;
1142 alu
.src
[0].sel
= ctx
->temp_reg
;
1143 alu
.src
[0].chan
= 0;
1145 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1149 if (ctx
->bc
->chiprev
== CHIPREV_R600
) {
1150 lit_vals
[0] = fui(3.1415926535897f
* 2.0f
);
1151 lit_vals
[1] = fui(-3.1415926535897f
);
1153 lit_vals
[0] = fui(1.0f
);
1154 lit_vals
[1] = fui(-0.5f
);
1157 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1158 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1162 alu
.dst
.sel
= ctx
->temp_reg
;
1165 alu
.src
[0].sel
= ctx
->temp_reg
;
1166 alu
.src
[0].chan
= 0;
1168 alu
.src
[1].sel
= V_SQ_ALU_SRC_LITERAL
;
1169 alu
.src
[1].chan
= 0;
1170 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1171 alu
.src
[2].chan
= 1;
1173 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1176 r
= r600_bc_add_literal(ctx
->bc
, lit_vals
);
1182 static int tgsi_trig(struct r600_shader_ctx
*ctx
)
1184 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1185 struct r600_bc_alu_src r600_src
[3];
1186 struct r600_bc_alu alu
;
1190 r
= tgsi_setup_trig(ctx
, r600_src
);
1194 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1195 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1197 alu
.dst
.sel
= ctx
->temp_reg
;
1200 alu
.src
[0].sel
= ctx
->temp_reg
;
1201 alu
.src
[0].chan
= 0;
1203 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1207 /* replicate result */
1208 for (i
= 0; i
< 4; i
++) {
1209 if (inst
->Dst
[0].Register
.WriteMask
& (1 << i
))
1212 for (i
= 0; i
< lasti
+ 1; i
++) {
1213 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
)))
1216 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1217 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1219 alu
.src
[0].sel
= ctx
->temp_reg
;
1220 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1225 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1232 static int tgsi_scs(struct r600_shader_ctx
*ctx
)
1234 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1235 struct r600_bc_alu_src r600_src
[3];
1236 struct r600_bc_alu alu
;
1239 /* We'll only need the trig stuff if we are going to write to the
1240 * X or Y components of the destination vector.
1242 if (likely(inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_XY
)) {
1243 r
= tgsi_setup_trig(ctx
, r600_src
);
1249 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_X
) {
1250 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1251 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
);
1252 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
1256 alu
.src
[0].sel
= ctx
->temp_reg
;
1257 alu
.src
[0].chan
= 0;
1259 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1265 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_Y
) {
1266 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1267 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
);
1268 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
1272 alu
.src
[0].sel
= ctx
->temp_reg
;
1273 alu
.src
[0].chan
= 0;
1275 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1281 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_Z
) {
1282 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1284 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1286 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1290 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1291 alu
.src
[0].chan
= 0;
1295 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1299 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1305 if (inst
->Dst
[0].Register
.WriteMask
& TGSI_WRITEMASK_W
) {
1306 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1308 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1310 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
1314 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1315 alu
.src
[0].chan
= 0;
1319 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1323 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1331 static int tgsi_kill(struct r600_shader_ctx
*ctx
)
1333 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1334 struct r600_bc_alu alu
;
1337 for (i
= 0; i
< 4; i
++) {
1338 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1339 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1343 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
1345 if (ctx
->inst_info
->tgsi_opcode
== TGSI_OPCODE_KILP
) {
1346 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1349 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1352 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1357 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1361 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1365 /* kill must be last in ALU */
1366 ctx
->bc
->force_add_cf
= 1;
1367 ctx
->shader
->uses_kill
= TRUE
;
1371 static int tgsi_lit(struct r600_shader_ctx
*ctx
)
1373 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1374 struct r600_bc_alu alu
;
1375 struct r600_bc_alu_src r600_src
[3];
1378 r
= tgsi_split_constant(ctx
, r600_src
);
1381 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1386 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1387 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1388 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
; /*1.0*/
1389 alu
.src
[0].chan
= 0;
1390 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 0, &alu
.dst
);
1393 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 0) & 1;
1394 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1398 /* dst.y = max(src.x, 0.0) */
1399 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1400 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
);
1401 alu
.src
[0] = r600_src
[0];
1402 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
; /*0.0*/
1403 alu
.src
[1].chan
= 0;
1404 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 1, &alu
.dst
);
1407 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 1) & 1;
1408 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1413 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1414 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1415 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1416 alu
.src
[0].chan
= 0;
1417 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 3, &alu
.dst
);
1420 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> 3) & 1;
1422 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1426 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1430 if (inst
->Dst
[0].Register
.WriteMask
& (1 << 2))
1435 /* dst.z = log(src.y) */
1436 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1437 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_CLAMPED
);
1438 alu
.src
[0] = r600_src
[0];
1439 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
1440 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1444 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1448 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1452 chan
= alu
.dst
.chan
;
1455 /* tmp.x = amd MUL_LIT(src.w, dst.z, src.x ) */
1456 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1457 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MUL_LIT
);
1458 alu
.src
[0] = r600_src
[0];
1459 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1460 alu
.src
[1].sel
= sel
;
1461 alu
.src
[1].chan
= chan
;
1463 alu
.src
[2] = r600_src
[0];
1464 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], 0);
1465 alu
.dst
.sel
= ctx
->temp_reg
;
1470 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1474 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1477 /* dst.z = exp(tmp.x) */
1478 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1479 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
1480 alu
.src
[0].sel
= ctx
->temp_reg
;
1481 alu
.src
[0].chan
= 0;
1482 r
= tgsi_dst(ctx
, &inst
->Dst
[0], 2, &alu
.dst
);
1486 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1493 static int tgsi_rsq(struct r600_shader_ctx
*ctx
)
1495 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1496 struct r600_bc_alu alu
;
1499 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1502 * For state trackers other than OpenGL, we'll want to use
1503 * _RECIPSQRT_IEEE instead.
1505 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_CLAMPED
);
1507 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1508 r
= tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1511 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1514 alu
.dst
.sel
= ctx
->temp_reg
;
1517 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1520 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1523 /* replicate result */
1524 return tgsi_helper_tempx_replicate(ctx
);
1527 static int tgsi_helper_tempx_replicate(struct r600_shader_ctx
*ctx
)
1529 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1530 struct r600_bc_alu alu
;
1533 for (i
= 0; i
< 4; i
++) {
1534 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1535 alu
.src
[0].sel
= ctx
->temp_reg
;
1536 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1538 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1541 alu
.dst
.write
= (inst
->Dst
[0].Register
.WriteMask
>> i
) & 1;
1544 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1551 static int tgsi_trans_srcx_replicate(struct r600_shader_ctx
*ctx
)
1553 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1554 struct r600_bc_alu alu
;
1557 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1558 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1559 for (i
= 0; i
< inst
->Instruction
.NumSrcRegs
; i
++) {
1560 r
= tgsi_src(ctx
, &inst
->Src
[i
], &alu
.src
[i
]);
1563 alu
.src
[i
].chan
= tgsi_chan(&inst
->Src
[i
], 0);
1565 alu
.dst
.sel
= ctx
->temp_reg
;
1568 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1571 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1574 /* replicate result */
1575 return tgsi_helper_tempx_replicate(ctx
);
1578 static int tgsi_pow(struct r600_shader_ctx
*ctx
)
1580 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1581 struct r600_bc_alu alu
;
1585 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1586 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
1587 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1590 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
1591 alu
.dst
.sel
= ctx
->temp_reg
;
1594 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1597 r
= r600_bc_add_literal(ctx
->bc
,ctx
->value
);
1601 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1602 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL_IEEE
);
1603 r
= tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[0]);
1606 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[1], 0);
1607 alu
.src
[1].sel
= ctx
->temp_reg
;
1608 alu
.dst
.sel
= ctx
->temp_reg
;
1611 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1614 r
= r600_bc_add_literal(ctx
->bc
,ctx
->value
);
1617 /* POW(a,b) = EXP2(b * LOG2(a))*/
1618 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1619 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
1620 alu
.src
[0].sel
= ctx
->temp_reg
;
1621 alu
.dst
.sel
= ctx
->temp_reg
;
1624 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1627 r
= r600_bc_add_literal(ctx
->bc
,ctx
->value
);
1630 return tgsi_helper_tempx_replicate(ctx
);
1633 static int tgsi_ssg(struct r600_shader_ctx
*ctx
)
1635 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1636 struct r600_bc_alu alu
;
1637 struct r600_bc_alu_src r600_src
[3];
1640 r
= tgsi_split_constant(ctx
, r600_src
);
1643 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1647 /* tmp = (src > 0 ? 1 : src) */
1648 for (i
= 0; i
< 4; i
++) {
1649 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1650 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
);
1653 alu
.dst
.sel
= ctx
->temp_reg
;
1656 alu
.src
[0] = r600_src
[0];
1657 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
1659 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1661 alu
.src
[2] = r600_src
[0];
1662 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[0], i
);
1665 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1669 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1673 /* dst = (-tmp > 0 ? -1 : tmp) */
1674 for (i
= 0; i
< 4; i
++) {
1675 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1676 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGT
);
1678 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1682 alu
.src
[0].sel
= ctx
->temp_reg
;
1683 alu
.src
[0].chan
= i
;
1686 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
1689 alu
.src
[2].sel
= ctx
->temp_reg
;
1690 alu
.src
[2].chan
= i
;
1694 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1701 static int tgsi_helper_copy(struct r600_shader_ctx
*ctx
, struct tgsi_full_instruction
*inst
)
1703 struct r600_bc_alu alu
;
1706 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
1709 for (i
= 0; i
< 4; i
++) {
1710 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1711 if (!(inst
->Dst
[0].Register
.WriteMask
& (1 << i
))) {
1712 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
);
1715 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1716 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
1719 alu
.src
[0].sel
= ctx
->temp_reg
;
1720 alu
.src
[0].chan
= i
;
1725 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1732 static int tgsi_op3(struct r600_shader_ctx
*ctx
)
1734 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1735 struct r600_bc_alu_src r600_src
[3];
1736 struct r600_bc_alu alu
;
1739 r
= tgsi_split_constant(ctx
, r600_src
);
1742 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1745 /* do it in 2 step as op3 doesn't support writemask */
1746 for (i
= 0; i
< 4; i
++) {
1747 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1748 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1749 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1750 alu
.src
[j
] = r600_src
[j
];
1751 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1753 alu
.dst
.sel
= ctx
->temp_reg
;
1760 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1764 return tgsi_helper_copy(ctx
, inst
);
1767 static int tgsi_dp(struct r600_shader_ctx
*ctx
)
1769 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1770 struct r600_bc_alu_src r600_src
[3];
1771 struct r600_bc_alu alu
;
1774 r
= tgsi_split_constant(ctx
, r600_src
);
1777 r
= tgsi_split_literal_constant(ctx
, r600_src
);
1780 for (i
= 0; i
< 4; i
++) {
1781 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1782 alu
.inst
= ctx
->inst_info
->r600_opcode
;
1783 for (j
= 0; j
< inst
->Instruction
.NumSrcRegs
; j
++) {
1784 alu
.src
[j
] = r600_src
[j
];
1785 alu
.src
[j
].chan
= tgsi_chan(&inst
->Src
[j
], i
);
1787 alu
.dst
.sel
= ctx
->temp_reg
;
1790 /* handle some special cases */
1791 switch (ctx
->inst_info
->tgsi_opcode
) {
1792 case TGSI_OPCODE_DP2
:
1794 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1795 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1798 case TGSI_OPCODE_DP3
:
1800 alu
.src
[0].sel
= alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
1801 alu
.src
[0].chan
= alu
.src
[1].chan
= 0;
1804 case TGSI_OPCODE_DPH
:
1806 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1807 alu
.src
[0].chan
= 0;
1817 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1821 return tgsi_helper_copy(ctx
, inst
);
1824 static int tgsi_tex(struct r600_shader_ctx
*ctx
)
1826 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
1827 struct r600_bc_tex tex
;
1828 struct r600_bc_alu alu
;
1832 boolean src_not_temp
= inst
->Src
[0].Register
.File
!= TGSI_FILE_TEMPORARY
;
1833 uint32_t lit_vals
[4];
1835 src_gpr
= ctx
->file_offset
[inst
->Src
[0].Register
.File
] + inst
->Src
[0].Register
.Index
;
1837 if (inst
->Instruction
.Opcode
== TGSI_OPCODE_TXP
) {
1838 /* Add perspective divide */
1839 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1840 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
1841 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1845 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 3);
1846 alu
.dst
.sel
= ctx
->temp_reg
;
1850 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1854 for (i
= 0; i
< 3; i
++) {
1855 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1856 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
1857 alu
.src
[0].sel
= ctx
->temp_reg
;
1858 alu
.src
[0].chan
= 3;
1859 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1862 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
1863 alu
.dst
.sel
= ctx
->temp_reg
;
1866 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1870 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1871 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
1872 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
1873 alu
.src
[0].chan
= 0;
1874 alu
.dst
.sel
= ctx
->temp_reg
;
1878 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1881 src_not_temp
= FALSE
;
1882 src_gpr
= ctx
->temp_reg
;
1885 if (inst
->Texture
.Texture
== TGSI_TEXTURE_CUBE
) {
1886 int src_chan
, src2_chan
;
1888 /* tmp1.xyzw = CUBE(R0.zzxy, R0.yxzz) */
1889 for (i
= 0; i
< 4; i
++) {
1890 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1891 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE
);
1915 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
1918 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], src_chan
);
1919 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[1]);
1922 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], src2_chan
);
1923 alu
.dst
.sel
= ctx
->temp_reg
;
1928 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1933 /* tmp1.z = RCP_e(|tmp1.z|) */
1934 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1935 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
1936 alu
.src
[0].sel
= ctx
->temp_reg
;
1937 alu
.src
[0].chan
= 2;
1939 alu
.dst
.sel
= ctx
->temp_reg
;
1943 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1947 /* MULADD R0.x, R0.x, PS1, (0x3FC00000, 1.5f).x
1948 * MULADD R0.y, R0.y, PS1, (0x3FC00000, 1.5f).x
1949 * muladd has no writemask, have to use another temp
1951 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1952 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1955 alu
.src
[0].sel
= ctx
->temp_reg
;
1956 alu
.src
[0].chan
= 0;
1957 alu
.src
[1].sel
= ctx
->temp_reg
;
1958 alu
.src
[1].chan
= 2;
1960 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1961 alu
.src
[2].chan
= 0;
1963 alu
.dst
.sel
= ctx
->temp_reg
;
1967 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1971 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
1972 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
1975 alu
.src
[0].sel
= ctx
->temp_reg
;
1976 alu
.src
[0].chan
= 1;
1977 alu
.src
[1].sel
= ctx
->temp_reg
;
1978 alu
.src
[1].chan
= 2;
1980 alu
.src
[2].sel
= V_SQ_ALU_SRC_LITERAL
;
1981 alu
.src
[2].chan
= 0;
1983 alu
.dst
.sel
= ctx
->temp_reg
;
1988 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
1992 lit_vals
[0] = fui(1.5f
);
1994 r
= r600_bc_add_literal(ctx
->bc
, lit_vals
);
1997 src_not_temp
= FALSE
;
1998 src_gpr
= ctx
->temp_reg
;
2002 for (i
= 0; i
< 4; i
++) {
2003 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2004 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2005 alu
.src
[0].sel
= src_gpr
;
2006 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2007 alu
.dst
.sel
= ctx
->temp_reg
;
2012 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2016 src_gpr
= ctx
->temp_reg
;
2019 opcode
= ctx
->inst_info
->r600_opcode
;
2020 if (opcode
== SQ_TEX_INST_SAMPLE
&&
2021 (inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW1D
|| inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW2D
))
2022 opcode
= SQ_TEX_INST_SAMPLE_C
;
2024 memset(&tex
, 0, sizeof(struct r600_bc_tex
));
2026 tex
.sampler_id
= ctx
->file_offset
[inst
->Src
[1].Register
.File
] + inst
->Src
[1].Register
.Index
;
2027 tex
.resource_id
= tex
.sampler_id
;
2028 tex
.src_gpr
= src_gpr
;
2029 tex
.dst_gpr
= ctx
->file_offset
[inst
->Dst
[0].Register
.File
] + inst
->Dst
[0].Register
.Index
;
2030 tex
.dst_sel_x
= (inst
->Dst
[0].Register
.WriteMask
& 1) ? 0 : 7;
2031 tex
.dst_sel_y
= (inst
->Dst
[0].Register
.WriteMask
& 2) ? 1 : 7;
2032 tex
.dst_sel_z
= (inst
->Dst
[0].Register
.WriteMask
& 4) ? 2 : 7;
2033 tex
.dst_sel_w
= (inst
->Dst
[0].Register
.WriteMask
& 8) ? 3 : 7;
2039 if (inst
->Texture
.Texture
== TGSI_TEXTURE_CUBE
) {
2046 if (inst
->Texture
.Texture
!= TGSI_TEXTURE_RECT
) {
2047 tex
.coord_type_x
= 1;
2048 tex
.coord_type_y
= 1;
2049 tex
.coord_type_z
= 1;
2050 tex
.coord_type_w
= 1;
2053 if (inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW1D
|| inst
->Texture
.Texture
== TGSI_TEXTURE_SHADOW2D
)
2056 r
= r600_bc_add_tex(ctx
->bc
, &tex
);
2060 /* add shadow ambient support - gallium doesn't do it yet */
2064 static int tgsi_lrp(struct r600_shader_ctx
*ctx
)
2066 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2067 struct r600_bc_alu_src r600_src
[3];
2068 struct r600_bc_alu alu
;
2072 r
= tgsi_split_constant(ctx
, r600_src
);
2075 r
= tgsi_split_literal_constant(ctx
, r600_src
);
2079 for (i
= 0; i
< 4; i
++) {
2080 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2081 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
);
2082 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2083 alu
.src
[0].chan
= 0;
2084 alu
.src
[1] = r600_src
[0];
2085 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[0], i
);
2087 alu
.dst
.sel
= ctx
->temp_reg
;
2093 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2097 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2101 /* (1 - src0) * src2 */
2102 for (i
= 0; i
< 4; i
++) {
2103 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2104 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2105 alu
.src
[0].sel
= ctx
->temp_reg
;
2106 alu
.src
[0].chan
= i
;
2107 alu
.src
[1] = r600_src
[2];
2108 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
2109 alu
.dst
.sel
= ctx
->temp_reg
;
2115 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2119 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2123 /* src0 * src1 + (1 - src0) * src2 */
2124 for (i
= 0; i
< 4; i
++) {
2125 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2126 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
2128 alu
.src
[0] = r600_src
[0];
2129 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2130 alu
.src
[1] = r600_src
[1];
2131 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
2132 alu
.src
[2].sel
= ctx
->temp_reg
;
2133 alu
.src
[2].chan
= i
;
2134 alu
.dst
.sel
= ctx
->temp_reg
;
2139 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2143 return tgsi_helper_copy(ctx
, inst
);
2146 static int tgsi_cmp(struct r600_shader_ctx
*ctx
)
2148 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2149 struct r600_bc_alu_src r600_src
[3];
2150 struct r600_bc_alu alu
;
2154 r
= tgsi_split_constant(ctx
, r600_src
);
2157 r
= tgsi_split_literal_constant(ctx
, r600_src
);
2161 if (inst
->Dst
[0].Register
.WriteMask
!= 0xf)
2164 for (i
= 0; i
< 4; i
++) {
2165 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2166 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_CNDGE
);
2167 alu
.src
[0] = r600_src
[0];
2168 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2170 alu
.src
[1] = r600_src
[2];
2171 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[2], i
);
2173 alu
.src
[2] = r600_src
[1];
2174 alu
.src
[2].chan
= tgsi_chan(&inst
->Src
[1], i
);
2177 alu
.dst
.sel
= ctx
->temp_reg
;
2179 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2188 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2193 return tgsi_helper_copy(ctx
, inst
);
2197 static int tgsi_xpd(struct r600_shader_ctx
*ctx
)
2199 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2200 struct r600_bc_alu_src r600_src
[3];
2201 struct r600_bc_alu alu
;
2202 uint32_t use_temp
= 0;
2205 if (inst
->Dst
[0].Register
.WriteMask
!= 0xf)
2208 r
= tgsi_split_constant(ctx
, r600_src
);
2211 r
= tgsi_split_literal_constant(ctx
, r600_src
);
2215 for (i
= 0; i
< 4; i
++) {
2216 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2217 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2219 alu
.src
[0] = r600_src
[0];
2222 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
2225 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2228 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
2231 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
2232 alu
.src
[0].chan
= i
;
2235 alu
.src
[1] = r600_src
[1];
2238 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
2241 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
2244 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
2247 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2248 alu
.src
[1].chan
= i
;
2251 alu
.dst
.sel
= ctx
->temp_reg
;
2257 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2261 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2266 for (i
= 0; i
< 4; i
++) {
2267 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2268 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
);
2270 alu
.src
[0] = r600_src
[0];
2273 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 1);
2276 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 2);
2279 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2282 alu
.src
[0].sel
= V_SQ_ALU_SRC_0
;
2283 alu
.src
[0].chan
= i
;
2286 alu
.src
[1] = r600_src
[1];
2289 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 2);
2292 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 0);
2295 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], 1);
2298 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2299 alu
.src
[1].chan
= i
;
2302 alu
.src
[2].sel
= ctx
->temp_reg
;
2304 alu
.src
[2].chan
= i
;
2307 alu
.dst
.sel
= ctx
->temp_reg
;
2309 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2318 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2322 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2327 return tgsi_helper_copy(ctx
, inst
);
2331 static int tgsi_exp(struct r600_shader_ctx
*ctx
)
2333 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2334 struct r600_bc_alu_src r600_src
[3] = { { 0 } };
2335 struct r600_bc_alu alu
;
2338 /* result.x = 2^floor(src); */
2339 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
2340 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2342 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2343 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2347 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2349 alu
.dst
.sel
= ctx
->temp_reg
;
2353 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2357 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2361 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2362 alu
.src
[0].sel
= ctx
->temp_reg
;
2363 alu
.src
[0].chan
= 0;
2365 alu
.dst
.sel
= ctx
->temp_reg
;
2369 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2373 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2378 /* result.y = tmp - floor(tmp); */
2379 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
2380 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2382 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
);
2383 alu
.src
[0] = r600_src
[0];
2384 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2387 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2389 alu
.dst
.sel
= ctx
->temp_reg
;
2390 // r = tgsi_dst(ctx, &inst->Dst[0], i, &alu.dst);
2398 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2401 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2406 /* result.z = RoughApprox2ToX(tmp);*/
2407 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 0x1) {
2408 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2409 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2410 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2413 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2415 alu
.dst
.sel
= ctx
->temp_reg
;
2421 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2424 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2429 /* result.w = 1.0;*/
2430 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 0x1) {
2431 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2433 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2434 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2435 alu
.src
[0].chan
= 0;
2437 alu
.dst
.sel
= ctx
->temp_reg
;
2441 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2444 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2448 return tgsi_helper_copy(ctx
, inst
);
2451 static int tgsi_log(struct r600_shader_ctx
*ctx
)
2453 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2454 struct r600_bc_alu alu
;
2457 /* result.x = floor(log2(src)); */
2458 if (inst
->Dst
[0].Register
.WriteMask
& 1) {
2459 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2461 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2462 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2466 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2468 alu
.dst
.sel
= ctx
->temp_reg
;
2472 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2476 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2480 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2481 alu
.src
[0].sel
= ctx
->temp_reg
;
2482 alu
.src
[0].chan
= 0;
2484 alu
.dst
.sel
= ctx
->temp_reg
;
2489 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2493 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2498 /* result.y = src.x / (2 ^ floor(log2(src.x))); */
2499 if ((inst
->Dst
[0].Register
.WriteMask
>> 1) & 1) {
2500 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2502 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2503 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2507 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2509 alu
.dst
.sel
= ctx
->temp_reg
;
2514 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2518 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2522 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2524 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
);
2525 alu
.src
[0].sel
= ctx
->temp_reg
;
2526 alu
.src
[0].chan
= 1;
2528 alu
.dst
.sel
= ctx
->temp_reg
;
2533 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2537 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2541 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2543 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
);
2544 alu
.src
[0].sel
= ctx
->temp_reg
;
2545 alu
.src
[0].chan
= 1;
2547 alu
.dst
.sel
= ctx
->temp_reg
;
2552 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2556 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2560 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2562 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
);
2563 alu
.src
[0].sel
= ctx
->temp_reg
;
2564 alu
.src
[0].chan
= 1;
2566 alu
.dst
.sel
= ctx
->temp_reg
;
2571 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2575 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2579 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2581 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2583 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2587 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2589 alu
.src
[1].sel
= ctx
->temp_reg
;
2590 alu
.src
[1].chan
= 1;
2592 alu
.dst
.sel
= ctx
->temp_reg
;
2597 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2601 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2606 /* result.z = log2(src);*/
2607 if ((inst
->Dst
[0].Register
.WriteMask
>> 2) & 1) {
2608 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2610 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
);
2611 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2615 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2617 alu
.dst
.sel
= ctx
->temp_reg
;
2622 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2626 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2631 /* result.w = 1.0; */
2632 if ((inst
->Dst
[0].Register
.WriteMask
>> 3) & 1) {
2633 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2635 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
);
2636 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2637 alu
.src
[0].chan
= 0;
2639 alu
.dst
.sel
= ctx
->temp_reg
;
2644 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2648 r
= r600_bc_add_literal(ctx
->bc
, ctx
->value
);
2653 return tgsi_helper_copy(ctx
, inst
);
2656 static int tgsi_eg_arl(struct r600_shader_ctx
*ctx
)
2658 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2659 struct r600_bc_alu alu
;
2661 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2663 switch (inst
->Instruction
.Opcode
) {
2664 case TGSI_OPCODE_ARL
:
2665 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT_FLOOR
;
2667 case TGSI_OPCODE_ARR
:
2668 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT
;
2675 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2678 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2681 alu
.dst
.sel
= ctx
->temp_reg
;
2683 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU
));
2686 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2687 alu
.inst
= EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT
;
2688 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2691 alu
.src
[0].sel
= ctx
->temp_reg
;
2692 alu
.src
[0].chan
= 0;
2694 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU
));
2699 static int tgsi_r600_arl(struct r600_shader_ctx
*ctx
)
2701 /* TODO from r600c, ar values don't persist between clauses */
2702 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2703 struct r600_bc_alu alu
;
2705 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2707 switch (inst
->Instruction
.Opcode
) {
2708 case TGSI_OPCODE_ARL
:
2709 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_FLOOR
;
2711 case TGSI_OPCODE_ARR
:
2712 alu
.inst
= V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA
;
2720 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2723 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2727 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU
));
2730 ctx
->bc
->cf_last
->r6xx_uses_waterfall
= 1;
2734 static int tgsi_opdst(struct r600_shader_ctx
*ctx
)
2736 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2737 struct r600_bc_alu alu
;
2740 for (i
= 0; i
< 4; i
++) {
2741 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2743 alu
.inst
= CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
);
2744 r
= tgsi_dst(ctx
, &inst
->Dst
[0], i
, &alu
.dst
);
2748 if (i
== 0 || i
== 3) {
2749 alu
.src
[0].sel
= V_SQ_ALU_SRC_1
;
2751 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2754 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], i
);
2757 if (i
== 0 || i
== 2) {
2758 alu
.src
[1].sel
= V_SQ_ALU_SRC_1
;
2760 r
= tgsi_src(ctx
, &inst
->Src
[1], &alu
.src
[1]);
2763 alu
.src
[1].chan
= tgsi_chan(&inst
->Src
[1], i
);
2767 r
= r600_bc_add_alu(ctx
->bc
, &alu
);
2774 static int emit_logic_pred(struct r600_shader_ctx
*ctx
, int opcode
)
2776 struct tgsi_full_instruction
*inst
= &ctx
->parse
.FullToken
.FullInstruction
;
2777 struct r600_bc_alu alu
;
2780 memset(&alu
, 0, sizeof(struct r600_bc_alu
));
2784 alu
.dst
.sel
= ctx
->temp_reg
;
2788 r
= tgsi_src(ctx
, &inst
->Src
[0], &alu
.src
[0]);
2791 alu
.src
[0].chan
= tgsi_chan(&inst
->Src
[0], 0);
2792 alu
.src
[1].sel
= V_SQ_ALU_SRC_0
;
2793 alu
.src
[1].chan
= 0;
2797 r
= r600_bc_add_alu_type(ctx
->bc
, &alu
, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE
));
2803 static int pops(struct r600_shader_ctx
*ctx
, int pops
)
2805 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_POP
));
2806 ctx
->bc
->cf_last
->pop_count
= pops
;
2810 static inline void callstack_decrease_current(struct r600_shader_ctx
*ctx
, unsigned reason
)
2814 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
--;
2818 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
-= 4;
2821 /* TOODO : for 16 vp asic should -= 2; */
2822 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
--;
2827 static inline void callstack_check_depth(struct r600_shader_ctx
*ctx
, unsigned reason
, unsigned check_max_only
)
2829 if (check_max_only
) {
2842 if ((ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+ diff
) >
2843 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
) {
2844 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
=
2845 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+ diff
;
2851 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
++;
2855 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
+= 4;
2858 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
++;
2862 if ((ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
) >
2863 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
) {
2864 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].max
=
2865 ctx
->bc
->callstack
[ctx
->bc
->call_sp
].current
;
2869 static void fc_set_mid(struct r600_shader_ctx
*ctx
, int fc_sp
)
2871 struct r600_cf_stack_entry
*sp
= &ctx
->bc
->fc_stack
[fc_sp
];
2873 sp
->mid
= (struct r600_bc_cf
**)realloc((void *)sp
->mid
,
2874 sizeof(struct r600_bc_cf
*) * (sp
->num_mid
+ 1));
2875 sp
->mid
[sp
->num_mid
] = ctx
->bc
->cf_last
;
2879 static void fc_pushlevel(struct r600_shader_ctx
*ctx
, int type
)
2882 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
= type
;
2883 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
= ctx
->bc
->cf_last
;
2886 static void fc_poplevel(struct r600_shader_ctx
*ctx
)
2888 struct r600_cf_stack_entry
*sp
= &ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
];
2900 static int emit_return(struct r600_shader_ctx
*ctx
)
2902 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_RETURN
);
2906 static int emit_jump_to_offset(struct r600_shader_ctx
*ctx
, int pops
, int offset
)
2909 r600_bc_add_cfinst(ctx
->bc
, V_SQ_CF_WORD1_SQ_CF_INST_JUMP
);
2910 ctx
->bc
->cf_last
->pop_count
= pops
;
2911 /* TODO work out offset */
2915 static int emit_setret_in_loop_flag(struct r600_shader_ctx
*ctx
, unsigned flag_value
)
2920 static void emit_testflag(struct r600_shader_ctx
*ctx
)
2925 static void emit_return_on_flag(struct r600_shader_ctx
*ctx
, unsigned ifidx
)
2928 emit_jump_to_offset(ctx
, 1, 4);
2929 emit_setret_in_loop_flag(ctx
, V_SQ_ALU_SRC_0
);
2930 pops(ctx
, ifidx
+ 1);
2934 static void break_loop_on_flag(struct r600_shader_ctx
*ctx
, unsigned fc_sp
)
2938 r600_bc_add_cfinst(ctx
->bc
, ctx
->inst_info
->r600_opcode
);
2939 ctx
->bc
->cf_last
->pop_count
= 1;
2941 fc_set_mid(ctx
, fc_sp
);
2947 static int tgsi_if(struct r600_shader_ctx
*ctx
)
2949 emit_logic_pred(ctx
, CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_PRED_SETNE
));
2951 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_JUMP
));
2953 fc_pushlevel(ctx
, FC_IF
);
2955 callstack_check_depth(ctx
, FC_PUSH_VPM
, 0);
2959 static int tgsi_else(struct r600_shader_ctx
*ctx
)
2961 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_ELSE
));
2962 ctx
->bc
->cf_last
->pop_count
= 1;
2964 fc_set_mid(ctx
, ctx
->bc
->fc_sp
);
2965 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
;
2969 static int tgsi_endif(struct r600_shader_ctx
*ctx
)
2972 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_IF
) {
2973 R600_ERR("if/endif unbalanced in shader\n");
2977 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
== NULL
) {
2978 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2979 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->pop_count
= 1;
2981 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
[0]->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
2985 callstack_decrease_current(ctx
, FC_PUSH_VPM
);
2989 static int tgsi_bgnloop(struct r600_shader_ctx
*ctx
)
2991 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_START_NO_AL
));
2993 fc_pushlevel(ctx
, FC_LOOP
);
2995 /* check stack depth */
2996 callstack_check_depth(ctx
, FC_LOOP
, 0);
3000 static int tgsi_endloop(struct r600_shader_ctx
*ctx
)
3004 r600_bc_add_cfinst(ctx
->bc
, CTX_INST(V_SQ_CF_WORD1_SQ_CF_INST_LOOP_END
));
3006 if (ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].type
!= FC_LOOP
) {
3007 R600_ERR("loop/endloop in shader code are not paired.\n");
3011 /* fixup loop pointers - from r600isa
3012 LOOP END points to CF after LOOP START,
3013 LOOP START point to CF after LOOP END
3014 BRK/CONT point to LOOP END CF
3016 ctx
->bc
->cf_last
->cf_addr
= ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->id
+ 2;
3018 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].start
->cf_addr
= ctx
->bc
->cf_last
->id
+ 2;
3020 for (i
= 0; i
< ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].num_mid
; i
++) {
3021 ctx
->bc
->fc_stack
[ctx
->bc
->fc_sp
].mid
[i
]->cf_addr
= ctx
->bc
->cf_last
->id
;
3023 /* TODO add LOOPRET support */
3025 callstack_decrease_current(ctx
, FC_LOOP
);
3029 static int tgsi_loop_brk_cont(struct r600_shader_ctx
*ctx
)
3033 for (fscp
= ctx
->bc
->fc_sp
; fscp
> 0; fscp
--)
3035 if (FC_LOOP
== ctx
->bc
->fc_stack
[fscp
].type
)
3040 R600_ERR("Break not inside loop/endloop pair\n");
3044 r600_bc_add_cfinst(ctx
->bc
, ctx
->inst_info
->r600_opcode
);
3045 ctx
->bc
->cf_last
->pop_count
= 1;
3047 fc_set_mid(ctx
, fscp
);
3050 callstack_check_depth(ctx
, FC_PUSH_VPM
, 1);
3054 static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction
[] = {
3055 {TGSI_OPCODE_ARL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_r600_arl
},
3056 {TGSI_OPCODE_MOV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
3057 {TGSI_OPCODE_LIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
3060 * For state trackers other than OpenGL, we'll want to use
3061 * _RECIP_IEEE instead.
3063 {TGSI_OPCODE_RCP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_CLAMPED
, tgsi_trans_srcx_replicate
},
3065 {TGSI_OPCODE_RSQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_rsq
},
3066 {TGSI_OPCODE_EXP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
3067 {TGSI_OPCODE_LOG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_log
},
3068 {TGSI_OPCODE_MUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
3069 {TGSI_OPCODE_ADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
3070 {TGSI_OPCODE_DP3
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3071 {TGSI_OPCODE_DP4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3072 {TGSI_OPCODE_DST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_opdst
},
3073 {TGSI_OPCODE_MIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
3074 {TGSI_OPCODE_MAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
3075 {TGSI_OPCODE_SLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
3076 {TGSI_OPCODE_SGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
3077 {TGSI_OPCODE_MAD
, 1, V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
3078 {TGSI_OPCODE_SUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
3079 {TGSI_OPCODE_LRP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
3080 {TGSI_OPCODE_CND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3082 {20, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3083 {TGSI_OPCODE_DP2A
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3085 {22, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3086 {23, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3087 {TGSI_OPCODE_FRC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
3088 {TGSI_OPCODE_CLAMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3089 {TGSI_OPCODE_FLR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
3090 {TGSI_OPCODE_ROUND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3091 {TGSI_OPCODE_EX2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
3092 {TGSI_OPCODE_LG2
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
3093 {TGSI_OPCODE_POW
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
3094 {TGSI_OPCODE_XPD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
3096 {32, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3097 {TGSI_OPCODE_ABS
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
3098 {TGSI_OPCODE_RCC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3099 {TGSI_OPCODE_DPH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3100 {TGSI_OPCODE_COS
, 0, 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, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* predicated kill */
3104 {TGSI_OPCODE_PK2H
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3105 {TGSI_OPCODE_PK2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3106 {TGSI_OPCODE_PK4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3107 {TGSI_OPCODE_PK4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3108 {TGSI_OPCODE_RFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3109 {TGSI_OPCODE_SEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
3110 {TGSI_OPCODE_SFL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3111 {TGSI_OPCODE_SGT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
3112 {TGSI_OPCODE_SIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
3113 {TGSI_OPCODE_SLE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
3114 {TGSI_OPCODE_SNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
3115 {TGSI_OPCODE_STR
, 0, 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, 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, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3120 {TGSI_OPCODE_UP2US
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3121 {TGSI_OPCODE_UP4B
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3122 {TGSI_OPCODE_UP4UB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3123 {TGSI_OPCODE_X2D
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3124 {TGSI_OPCODE_ARA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3125 {TGSI_OPCODE_ARR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_r600_arl
},
3126 {TGSI_OPCODE_BRA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3127 {TGSI_OPCODE_CAL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3128 {TGSI_OPCODE_RET
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3129 {TGSI_OPCODE_SSG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
3130 {TGSI_OPCODE_CMP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
3131 {TGSI_OPCODE_SCS
, 0, 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, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3134 {TGSI_OPCODE_DIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3135 {TGSI_OPCODE_DP2
, 0, 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, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK
, tgsi_loop_brk_cont
},
3138 {TGSI_OPCODE_IF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
3140 {75, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3141 {76, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3142 {TGSI_OPCODE_ELSE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
3143 {TGSI_OPCODE_ENDIF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
3145 {79, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3146 {80, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3147 {TGSI_OPCODE_PUSHA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3148 {TGSI_OPCODE_POPA
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3149 {TGSI_OPCODE_CEIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3150 {TGSI_OPCODE_I2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3151 {TGSI_OPCODE_NOT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3152 {TGSI_OPCODE_TRUNC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
3153 {TGSI_OPCODE_SHL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3155 {88, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3156 {TGSI_OPCODE_AND
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3157 {TGSI_OPCODE_OR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3158 {TGSI_OPCODE_MOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3159 {TGSI_OPCODE_XOR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3160 {TGSI_OPCODE_SAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3161 {TGSI_OPCODE_TXF
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3162 {TGSI_OPCODE_TXQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3163 {TGSI_OPCODE_CONT
, 0, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE
, tgsi_loop_brk_cont
},
3164 {TGSI_OPCODE_EMIT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3165 {TGSI_OPCODE_ENDPRIM
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3166 {TGSI_OPCODE_BGNLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_bgnloop
},
3167 {TGSI_OPCODE_BGNSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3168 {TGSI_OPCODE_ENDLOOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endloop
},
3169 {TGSI_OPCODE_ENDSUB
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3171 {103, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3172 {104, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3173 {105, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3174 {106, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3175 {TGSI_OPCODE_NOP
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3177 {108, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3178 {109, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3179 {110, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3180 {111, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3181 {TGSI_OPCODE_NRM4
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3182 {TGSI_OPCODE_CALLNZ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3183 {TGSI_OPCODE_IFC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3184 {TGSI_OPCODE_BREAKC
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3185 {TGSI_OPCODE_KIL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
3186 {TGSI_OPCODE_END
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
3188 {118, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3189 {TGSI_OPCODE_F2I
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3190 {TGSI_OPCODE_IDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3191 {TGSI_OPCODE_IMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3192 {TGSI_OPCODE_IMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3193 {TGSI_OPCODE_INEG
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3194 {TGSI_OPCODE_ISGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3195 {TGSI_OPCODE_ISHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3196 {TGSI_OPCODE_ISLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3197 {TGSI_OPCODE_F2U
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3198 {TGSI_OPCODE_U2F
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3199 {TGSI_OPCODE_UADD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3200 {TGSI_OPCODE_UDIV
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3201 {TGSI_OPCODE_UMAD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3202 {TGSI_OPCODE_UMAX
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3203 {TGSI_OPCODE_UMIN
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3204 {TGSI_OPCODE_UMOD
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3205 {TGSI_OPCODE_UMUL
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3206 {TGSI_OPCODE_USEQ
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3207 {TGSI_OPCODE_USGE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3208 {TGSI_OPCODE_USHR
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3209 {TGSI_OPCODE_USLT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3210 {TGSI_OPCODE_USNE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3211 {TGSI_OPCODE_SWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3212 {TGSI_OPCODE_CASE
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3213 {TGSI_OPCODE_DEFAULT
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3214 {TGSI_OPCODE_ENDSWITCH
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3215 {TGSI_OPCODE_LAST
, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3218 static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction
[] = {
3219 {TGSI_OPCODE_ARL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_eg_arl
},
3220 {TGSI_OPCODE_MOV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
3221 {TGSI_OPCODE_LIT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lit
},
3222 {TGSI_OPCODE_RCP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE
, tgsi_trans_srcx_replicate
},
3223 {TGSI_OPCODE_RSQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_IEEE
, tgsi_trans_srcx_replicate
},
3224 {TGSI_OPCODE_EXP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_exp
},
3225 {TGSI_OPCODE_LOG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3226 {TGSI_OPCODE_MUL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MUL
, tgsi_op2
},
3227 {TGSI_OPCODE_ADD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
3228 {TGSI_OPCODE_DP3
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3229 {TGSI_OPCODE_DP4
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3230 {TGSI_OPCODE_DST
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_opdst
},
3231 {TGSI_OPCODE_MIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MIN
, tgsi_op2
},
3232 {TGSI_OPCODE_MAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MAX
, tgsi_op2
},
3233 {TGSI_OPCODE_SLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2_swap
},
3234 {TGSI_OPCODE_SGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2
},
3235 {TGSI_OPCODE_MAD
, 1, EG_V_SQ_ALU_WORD1_OP3_SQ_OP3_INST_MULADD
, tgsi_op3
},
3236 {TGSI_OPCODE_SUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ADD
, tgsi_op2
},
3237 {TGSI_OPCODE_LRP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_lrp
},
3238 {TGSI_OPCODE_CND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3240 {20, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3241 {TGSI_OPCODE_DP2A
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3243 {22, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3244 {23, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3245 {TGSI_OPCODE_FRC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FRACT
, tgsi_op2
},
3246 {TGSI_OPCODE_CLAMP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3247 {TGSI_OPCODE_FLR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR
, tgsi_op2
},
3248 {TGSI_OPCODE_ROUND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3249 {TGSI_OPCODE_EX2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_EXP_IEEE
, tgsi_trans_srcx_replicate
},
3250 {TGSI_OPCODE_LG2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LOG_IEEE
, tgsi_trans_srcx_replicate
},
3251 {TGSI_OPCODE_POW
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_pow
},
3252 {TGSI_OPCODE_XPD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_xpd
},
3254 {32, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3255 {TGSI_OPCODE_ABS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV
, tgsi_op2
},
3256 {TGSI_OPCODE_RCC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3257 {TGSI_OPCODE_DPH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3258 {TGSI_OPCODE_COS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_COS
, tgsi_trig
},
3259 {TGSI_OPCODE_DDX
, 0, SQ_TEX_INST_GET_GRADIENTS_H
, tgsi_tex
},
3260 {TGSI_OPCODE_DDY
, 0, SQ_TEX_INST_GET_GRADIENTS_V
, tgsi_tex
},
3261 {TGSI_OPCODE_KILP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* predicated kill */
3262 {TGSI_OPCODE_PK2H
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3263 {TGSI_OPCODE_PK2US
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3264 {TGSI_OPCODE_PK4B
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3265 {TGSI_OPCODE_PK4UB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3266 {TGSI_OPCODE_RFL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3267 {TGSI_OPCODE_SEQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETE
, tgsi_op2
},
3268 {TGSI_OPCODE_SFL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3269 {TGSI_OPCODE_SGT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGT
, tgsi_op2
},
3270 {TGSI_OPCODE_SIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SIN
, tgsi_trig
},
3271 {TGSI_OPCODE_SLE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETGE
, tgsi_op2_swap
},
3272 {TGSI_OPCODE_SNE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SETNE
, tgsi_op2
},
3273 {TGSI_OPCODE_STR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3274 {TGSI_OPCODE_TEX
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
3275 {TGSI_OPCODE_TXD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3276 {TGSI_OPCODE_TXP
, 0, SQ_TEX_INST_SAMPLE
, tgsi_tex
},
3277 {TGSI_OPCODE_UP2H
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3278 {TGSI_OPCODE_UP2US
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3279 {TGSI_OPCODE_UP4B
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3280 {TGSI_OPCODE_UP4UB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3281 {TGSI_OPCODE_X2D
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3282 {TGSI_OPCODE_ARA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3283 {TGSI_OPCODE_ARR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_eg_arl
},
3284 {TGSI_OPCODE_BRA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3285 {TGSI_OPCODE_CAL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3286 {TGSI_OPCODE_RET
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3287 {TGSI_OPCODE_SSG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_ssg
},
3288 {TGSI_OPCODE_CMP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_cmp
},
3289 {TGSI_OPCODE_SCS
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_scs
},
3290 {TGSI_OPCODE_TXB
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
3291 {TGSI_OPCODE_NRM
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3292 {TGSI_OPCODE_DIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3293 {TGSI_OPCODE_DP2
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4
, tgsi_dp
},
3294 {TGSI_OPCODE_TXL
, 0, SQ_TEX_INST_SAMPLE_L
, tgsi_tex
},
3295 {TGSI_OPCODE_BRK
, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK
, tgsi_loop_brk_cont
},
3296 {TGSI_OPCODE_IF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_if
},
3298 {75, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3299 {76, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3300 {TGSI_OPCODE_ELSE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_else
},
3301 {TGSI_OPCODE_ENDIF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endif
},
3303 {79, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3304 {80, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3305 {TGSI_OPCODE_PUSHA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3306 {TGSI_OPCODE_POPA
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3307 {TGSI_OPCODE_CEIL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3308 {TGSI_OPCODE_I2F
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3309 {TGSI_OPCODE_NOT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3310 {TGSI_OPCODE_TRUNC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_TRUNC
, tgsi_trans_srcx_replicate
},
3311 {TGSI_OPCODE_SHL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3313 {88, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3314 {TGSI_OPCODE_AND
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3315 {TGSI_OPCODE_OR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3316 {TGSI_OPCODE_MOD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3317 {TGSI_OPCODE_XOR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3318 {TGSI_OPCODE_SAD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3319 {TGSI_OPCODE_TXF
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3320 {TGSI_OPCODE_TXQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3321 {TGSI_OPCODE_CONT
, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_CONTINUE
, tgsi_loop_brk_cont
},
3322 {TGSI_OPCODE_EMIT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3323 {TGSI_OPCODE_ENDPRIM
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3324 {TGSI_OPCODE_BGNLOOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_bgnloop
},
3325 {TGSI_OPCODE_BGNSUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3326 {TGSI_OPCODE_ENDLOOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_endloop
},
3327 {TGSI_OPCODE_ENDSUB
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3329 {103, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3330 {104, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3331 {105, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3332 {106, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3333 {TGSI_OPCODE_NOP
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3335 {108, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3336 {109, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3337 {110, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3338 {111, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3339 {TGSI_OPCODE_NRM4
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3340 {TGSI_OPCODE_CALLNZ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3341 {TGSI_OPCODE_IFC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3342 {TGSI_OPCODE_BREAKC
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3343 {TGSI_OPCODE_KIL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_KILLGT
, tgsi_kill
}, /* conditional kill */
3344 {TGSI_OPCODE_END
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_end
}, /* aka HALT */
3346 {118, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3347 {TGSI_OPCODE_F2I
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3348 {TGSI_OPCODE_IDIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3349 {TGSI_OPCODE_IMAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3350 {TGSI_OPCODE_IMIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3351 {TGSI_OPCODE_INEG
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3352 {TGSI_OPCODE_ISGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3353 {TGSI_OPCODE_ISHR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3354 {TGSI_OPCODE_ISLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3355 {TGSI_OPCODE_F2U
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3356 {TGSI_OPCODE_U2F
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3357 {TGSI_OPCODE_UADD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3358 {TGSI_OPCODE_UDIV
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3359 {TGSI_OPCODE_UMAD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3360 {TGSI_OPCODE_UMAX
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3361 {TGSI_OPCODE_UMIN
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3362 {TGSI_OPCODE_UMOD
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3363 {TGSI_OPCODE_UMUL
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3364 {TGSI_OPCODE_USEQ
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3365 {TGSI_OPCODE_USGE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3366 {TGSI_OPCODE_USHR
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3367 {TGSI_OPCODE_USLT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3368 {TGSI_OPCODE_USNE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3369 {TGSI_OPCODE_SWITCH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3370 {TGSI_OPCODE_CASE
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3371 {TGSI_OPCODE_DEFAULT
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3372 {TGSI_OPCODE_ENDSWITCH
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},
3373 {TGSI_OPCODE_LAST
, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP
, tgsi_unsupported
},