projects / mesa.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
nir/spirv: Simplify type copying
[mesa.git] / src / compiler / spirv / spirv_to_nir.c
1 /*
2 * Copyright © 2015 Intel Corporation
3 *
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 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
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
13 * Software.
14 *
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 NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 *
23 * Authors:
24 * Jason Ekstrand (jason@jlekstrand.net)
25 *
26 */
27
28 #include "vtn_private.h"
29 #include "nir/nir_vla.h"
30 #include "nir/nir_control_flow.h"
31 #include "nir/nir_constant_expressions.h"
32 #include "spirv_info.h"
33
34 struct spec_constant_value {
35 bool is_double;
36 union {
37 uint32_t data32;
38 uint64_t data64;
39 };
40 };
41
42 void
43 _vtn_warn(const char *file, int line, const char *msg, ...)
44 {
45 char *formatted;
46 va_list args;
47
48 va_start(args, msg);
49 formatted = ralloc_vasprintf(NULL, msg, args);
50 va_end(args);
51
52 fprintf(stderr, "%s:%d WARNING: %s\n", file, line, formatted);
53
54 ralloc_free(formatted);
55 }
56
57 static struct vtn_ssa_value *
58 vtn_undef_ssa_value(struct vtn_builder *b, const struct glsl_type *type)
59 {
60 struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
61 val->type = type;
62
63 if (glsl_type_is_vector_or_scalar(type)) {
64 unsigned num_components = glsl_get_vector_elements(val->type);
65 unsigned bit_size = glsl_get_bit_size(val->type);
66 val->def = nir_ssa_undef(&b->nb, num_components, bit_size);
67 } else {
68 unsigned elems = glsl_get_length(val->type);
69 val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
70 if (glsl_type_is_matrix(type)) {
71 const struct glsl_type *elem_type =
72 glsl_vector_type(glsl_get_base_type(type),
73 glsl_get_vector_elements(type));
74
75 for (unsigned i = 0; i < elems; i++)
76 val->elems[i] = vtn_undef_ssa_value(b, elem_type);
77 } else if (glsl_type_is_array(type)) {
78 const struct glsl_type *elem_type = glsl_get_array_element(type);
79 for (unsigned i = 0; i < elems; i++)
80 val->elems[i] = vtn_undef_ssa_value(b, elem_type);
81 } else {
82 for (unsigned i = 0; i < elems; i++) {
83 const struct glsl_type *elem_type = glsl_get_struct_field(type, i);
84 val->elems[i] = vtn_undef_ssa_value(b, elem_type);
85 }
86 }
87 }
88
89 return val;
90 }
91
92 static struct vtn_ssa_value *
93 vtn_const_ssa_value(struct vtn_builder *b, nir_constant *constant,
94 const struct glsl_type *type)
95 {
96 struct hash_entry *entry = _mesa_hash_table_search(b->const_table, constant);
97
98 if (entry)
99 return entry->data;
100
101 struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
102 val->type = type;
103
104 switch (glsl_get_base_type(type)) {
105 case GLSL_TYPE_INT:
106 case GLSL_TYPE_UINT:
107 case GLSL_TYPE_INT64:
108 case GLSL_TYPE_UINT64:
109 case GLSL_TYPE_BOOL:
110 case GLSL_TYPE_FLOAT:
111 case GLSL_TYPE_DOUBLE: {
112 int bit_size = glsl_get_bit_size(type);
113 if (glsl_type_is_vector_or_scalar(type)) {
114 unsigned num_components = glsl_get_vector_elements(val->type);
115 nir_load_const_instr *load =
116 nir_load_const_instr_create(b->shader, num_components, bit_size);
117
118 load->value = constant->values[0];
119
120 nir_instr_insert_before_cf_list(&b->impl->body, &load->instr);
121 val->def = &load->def;
122 } else {
123 assert(glsl_type_is_matrix(type));
124 unsigned rows = glsl_get_vector_elements(val->type);
125 unsigned columns = glsl_get_matrix_columns(val->type);
126 val->elems = ralloc_array(b, struct vtn_ssa_value *, columns);
127
128 for (unsigned i = 0; i < columns; i++) {
129 struct vtn_ssa_value *col_val = rzalloc(b, struct vtn_ssa_value);
130 col_val->type = glsl_get_column_type(val->type);
131 nir_load_const_instr *load =
132 nir_load_const_instr_create(b->shader, rows, bit_size);
133
134 load->value = constant->values[i];
135
136 nir_instr_insert_before_cf_list(&b->impl->body, &load->instr);
137 col_val->def = &load->def;
138
139 val->elems[i] = col_val;
140 }
141 }
142 break;
143 }
144
145 case GLSL_TYPE_ARRAY: {
146 unsigned elems = glsl_get_length(val->type);
147 val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
148 const struct glsl_type *elem_type = glsl_get_array_element(val->type);
149 for (unsigned i = 0; i < elems; i++)
150 val->elems[i] = vtn_const_ssa_value(b, constant->elements[i],
151 elem_type);
152 break;
153 }
154
155 case GLSL_TYPE_STRUCT: {
156 unsigned elems = glsl_get_length(val->type);
157 val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
158 for (unsigned i = 0; i < elems; i++) {
159 const struct glsl_type *elem_type =
160 glsl_get_struct_field(val->type, i);
161 val->elems[i] = vtn_const_ssa_value(b, constant->elements[i],
162 elem_type);
163 }
164 break;
165 }
166
167 default:
168 unreachable("bad constant type");
169 }
170
171 return val;
172 }
173
174 struct vtn_ssa_value *
175 vtn_ssa_value(struct vtn_builder *b, uint32_t value_id)
176 {
177 struct vtn_value *val = vtn_untyped_value(b, value_id);
178 switch (val->value_type) {
179 case vtn_value_type_undef:
180 return vtn_undef_ssa_value(b, val->type->type);
181
182 case vtn_value_type_constant:
183 return vtn_const_ssa_value(b, val->constant, val->const_type);
184
185 case vtn_value_type_ssa:
186 return val->ssa;
187
188 case vtn_value_type_pointer:
189 /* This is needed for function parameters */
190 return vtn_variable_load(b, val->pointer);
191
192 default:
193 unreachable("Invalid type for an SSA value");
194 }
195 }
196
197 static char *
198 vtn_string_literal(struct vtn_builder *b, const uint32_t *words,
199 unsigned word_count, unsigned *words_used)
200 {
201 char *dup = ralloc_strndup(b, (char *)words, word_count * sizeof(*words));
202 if (words_used) {
203 /* Ammount of space taken by the string (including the null) */
204 unsigned len = strlen(dup) + 1;
205 *words_used = DIV_ROUND_UP(len, sizeof(*words));
206 }
207 return dup;
208 }
209
210 const uint32_t *
211 vtn_foreach_instruction(struct vtn_builder *b, const uint32_t *start,
212 const uint32_t *end, vtn_instruction_handler handler)
213 {
214 b->file = NULL;
215 b->line = -1;
216 b->col = -1;
217
218 const uint32_t *w = start;
219 while (w < end) {
220 SpvOp opcode = w[0] & SpvOpCodeMask;
221 unsigned count = w[0] >> SpvWordCountShift;
222 assert(count >= 1 && w + count <= end);
223
224 switch (opcode) {
225 case SpvOpNop:
226 break; /* Do nothing */
227
228 case SpvOpLine:
229 b->file = vtn_value(b, w[1], vtn_value_type_string)->str;
230 b->line = w[2];
231 b->col = w[3];
232 break;
233
234 case SpvOpNoLine:
235 b->file = NULL;
236 b->line = -1;
237 b->col = -1;
238 break;
239
240 default:
241 if (!handler(b, opcode, w, count))
242 return w;
243 break;
244 }
245
246 w += count;
247 }
248 assert(w == end);
249 return w;
250 }
251
252 static void
253 vtn_handle_extension(struct vtn_builder *b, SpvOp opcode,
254 const uint32_t *w, unsigned count)
255 {
256 switch (opcode) {
257 case SpvOpExtInstImport: {
258 struct vtn_value *val = vtn_push_value(b, w[1], vtn_value_type_extension);
259 if (strcmp((const char *)&w[2], "GLSL.std.450") == 0) {
260 val->ext_handler = vtn_handle_glsl450_instruction;
261 } else {
262 assert(!"Unsupported extension");
263 }
264 break;
265 }
266
267 case SpvOpExtInst: {
268 struct vtn_value *val = vtn_value(b, w[3], vtn_value_type_extension);
269 bool handled = val->ext_handler(b, w[4], w, count);
270 (void)handled;
271 assert(handled);
272 break;
273 }
274
275 default:
276 unreachable("Unhandled opcode");
277 }
278 }
279
280 static void
281 _foreach_decoration_helper(struct vtn_builder *b,
282 struct vtn_value *base_value,
283 int parent_member,
284 struct vtn_value *value,
285 vtn_decoration_foreach_cb cb, void *data)
286 {
287 for (struct vtn_decoration *dec = value->decoration; dec; dec = dec->next) {
288 int member;
289 if (dec->scope == VTN_DEC_DECORATION) {
290 member = parent_member;
291 } else if (dec->scope >= VTN_DEC_STRUCT_MEMBER0) {
292 assert(parent_member == -1);
293 member = dec->scope - VTN_DEC_STRUCT_MEMBER0;
294 } else {
295 /* Not a decoration */
296 continue;
297 }
298
299 if (dec->group) {
300 assert(dec->group->value_type == vtn_value_type_decoration_group);
301 _foreach_decoration_helper(b, base_value, member, dec->group,
302 cb, data);
303 } else {
304 cb(b, base_value, member, dec, data);
305 }
306 }
307 }
308
309 /** Iterates (recursively if needed) over all of the decorations on a value
310 *
311 * This function iterates over all of the decorations applied to a given
312 * value. If it encounters a decoration group, it recurses into the group
313 * and iterates over all of those decorations as well.
314 */
315 void
316 vtn_foreach_decoration(struct vtn_builder *b, struct vtn_value *value,
317 vtn_decoration_foreach_cb cb, void *data)
318 {
319 _foreach_decoration_helper(b, value, -1, value, cb, data);
320 }
321
322 void
323 vtn_foreach_execution_mode(struct vtn_builder *b, struct vtn_value *value,
324 vtn_execution_mode_foreach_cb cb, void *data)
325 {
326 for (struct vtn_decoration *dec = value->decoration; dec; dec = dec->next) {
327 if (dec->scope != VTN_DEC_EXECUTION_MODE)
328 continue;
329
330 assert(dec->group == NULL);
331 cb(b, value, dec, data);
332 }
333 }
334
335 static void
336 vtn_handle_decoration(struct vtn_builder *b, SpvOp opcode,
337 const uint32_t *w, unsigned count)
338 {
339 const uint32_t *w_end = w + count;
340 const uint32_t target = w[1];
341 w += 2;
342
343 switch (opcode) {
344 case SpvOpDecorationGroup:
345 vtn_push_value(b, target, vtn_value_type_decoration_group);
346 break;
347
348 case SpvOpDecorate:
349 case SpvOpMemberDecorate:
350 case SpvOpExecutionMode: {
351 struct vtn_value *val = &b->values[target];
352
353 struct vtn_decoration *dec = rzalloc(b, struct vtn_decoration);
354 switch (opcode) {
355 case SpvOpDecorate:
356 dec->scope = VTN_DEC_DECORATION;
357 break;
358 case SpvOpMemberDecorate:
359 dec->scope = VTN_DEC_STRUCT_MEMBER0 + *(w++);
360 break;
361 case SpvOpExecutionMode:
362 dec->scope = VTN_DEC_EXECUTION_MODE;
363 break;
364 default:
365 unreachable("Invalid decoration opcode");
366 }
367 dec->decoration = *(w++);
368 dec->literals = w;
369
370 /* Link into the list */
371 dec->next = val->decoration;
372 val->decoration = dec;
373 break;
374 }
375
376 case SpvOpGroupMemberDecorate:
377 case SpvOpGroupDecorate: {
378 struct vtn_value *group =
379 vtn_value(b, target, vtn_value_type_decoration_group);
380
381 for (; w < w_end; w++) {
382 struct vtn_value *val = vtn_untyped_value(b, *w);
383 struct vtn_decoration *dec = rzalloc(b, struct vtn_decoration);
384
385 dec->group = group;
386 if (opcode == SpvOpGroupDecorate) {
387 dec->scope = VTN_DEC_DECORATION;
388 } else {
389 dec->scope = VTN_DEC_STRUCT_MEMBER0 + *(++w);
390 }
391
392 /* Link into the list */
393 dec->next = val->decoration;
394 val->decoration = dec;
395 }
396 break;
397 }
398
399 default:
400 unreachable("Unhandled opcode");
401 }
402 }
403
404 struct member_decoration_ctx {
405 unsigned num_fields;
406 struct glsl_struct_field *fields;
407 struct vtn_type *type;
408 };
409
410 /* does a shallow copy of a vtn_type */
411
412 static struct vtn_type *
413 vtn_type_copy(struct vtn_builder *b, struct vtn_type *src)
414 {
415 struct vtn_type *dest = ralloc(b, struct vtn_type);
416 *dest = *src;
417
418 if (!glsl_type_is_scalar(src->type)) {
419 switch (glsl_get_base_type(src->type)) {
420 case GLSL_TYPE_INT:
421 case GLSL_TYPE_UINT:
422 case GLSL_TYPE_INT64:
423 case GLSL_TYPE_UINT64:
424 case GLSL_TYPE_BOOL:
425 case GLSL_TYPE_FLOAT:
426 case GLSL_TYPE_DOUBLE:
427 case GLSL_TYPE_ARRAY:
428 break;
429
430 case GLSL_TYPE_STRUCT: {
431 unsigned elems = glsl_get_length(src->type);
432
433 dest->members = ralloc_array(b, struct vtn_type *, elems);
434 memcpy(dest->members, src->members, elems * sizeof(struct vtn_type *));
435
436 dest->offsets = ralloc_array(b, unsigned, elems);
437 memcpy(dest->offsets, src->offsets, elems * sizeof(unsigned));
438 break;
439 }
440
441 default:
442 unreachable("unhandled type");
443 }
444 }
445
446 return dest;
447 }
448
449 static struct vtn_type *
450 mutable_matrix_member(struct vtn_builder *b, struct vtn_type *type, int member)
451 {
452 type->members[member] = vtn_type_copy(b, type->members[member]);
453 type = type->members[member];
454
455 /* We may have an array of matrices.... Oh, joy! */
456 while (glsl_type_is_array(type->type)) {
457 type->array_element = vtn_type_copy(b, type->array_element);
458 type = type->array_element;
459 }
460
461 assert(glsl_type_is_matrix(type->type));
462
463 return type;
464 }
465
466 static void
467 struct_member_decoration_cb(struct vtn_builder *b,
468 struct vtn_value *val, int member,
469 const struct vtn_decoration *dec, void *void_ctx)
470 {
471 struct member_decoration_ctx *ctx = void_ctx;
472
473 if (member < 0)
474 return;
475
476 assert(member < ctx->num_fields);
477
478 switch (dec->decoration) {
479 case SpvDecorationNonWritable:
480 case SpvDecorationNonReadable:
481 case SpvDecorationRelaxedPrecision:
482 case SpvDecorationVolatile:
483 case SpvDecorationCoherent:
484 case SpvDecorationUniform:
485 break; /* FIXME: Do nothing with this for now. */
486 case SpvDecorationNoPerspective:
487 ctx->fields[member].interpolation = INTERP_MODE_NOPERSPECTIVE;
488 break;
489 case SpvDecorationFlat:
490 ctx->fields[member].interpolation = INTERP_MODE_FLAT;
491 break;
492 case SpvDecorationCentroid:
493 ctx->fields[member].centroid = true;
494 break;
495 case SpvDecorationSample:
496 ctx->fields[member].sample = true;
497 break;
498 case SpvDecorationStream:
499 /* Vulkan only allows one GS stream */
500 assert(dec->literals[0] == 0);
501 break;
502 case SpvDecorationLocation:
503 ctx->fields[member].location = dec->literals[0];
504 break;
505 case SpvDecorationComponent:
506 break; /* FIXME: What should we do with these? */
507 case SpvDecorationBuiltIn:
508 ctx->type->members[member] = vtn_type_copy(b, ctx->type->members[member]);
509 ctx->type->members[member]->is_builtin = true;
510 ctx->type->members[member]->builtin = dec->literals[0];
511 ctx->type->builtin_block = true;
512 break;
513 case SpvDecorationOffset:
514 ctx->type->offsets[member] = dec->literals[0];
515 break;
516 case SpvDecorationMatrixStride:
517 /* Handled as a second pass */
518 break;
519 case SpvDecorationColMajor:
520 break; /* Nothing to do here. Column-major is the default. */
521 case SpvDecorationRowMajor:
522 mutable_matrix_member(b, ctx->type, member)->row_major = true;
523 break;
524
525 case SpvDecorationPatch:
526 break;
527
528 case SpvDecorationSpecId:
529 case SpvDecorationBlock:
530 case SpvDecorationBufferBlock:
531 case SpvDecorationArrayStride:
532 case SpvDecorationGLSLShared:
533 case SpvDecorationGLSLPacked:
534 case SpvDecorationInvariant:
535 case SpvDecorationRestrict:
536 case SpvDecorationAliased:
537 case SpvDecorationConstant:
538 case SpvDecorationIndex:
539 case SpvDecorationBinding:
540 case SpvDecorationDescriptorSet:
541 case SpvDecorationLinkageAttributes:
542 case SpvDecorationNoContraction:
543 case SpvDecorationInputAttachmentIndex:
544 vtn_warn("Decoration not allowed on struct members: %s",
545 spirv_decoration_to_string(dec->decoration));
546 break;
547
548 case SpvDecorationXfbBuffer:
549 case SpvDecorationXfbStride:
550 vtn_warn("Vulkan does not have transform feedback");
551 break;
552
553 case SpvDecorationCPacked:
554 case SpvDecorationSaturatedConversion:
555 case SpvDecorationFuncParamAttr:
556 case SpvDecorationFPRoundingMode:
557 case SpvDecorationFPFastMathMode:
558 case SpvDecorationAlignment:
559 vtn_warn("Decoration only allowed for CL-style kernels: %s",
560 spirv_decoration_to_string(dec->decoration));
561 break;
562
563 default:
564 unreachable("Unhandled decoration");
565 }
566 }
567
568 /* Matrix strides are handled as a separate pass because we need to know
569 * whether the matrix is row-major or not first.
570 */
571 static void
572 struct_member_matrix_stride_cb(struct vtn_builder *b,
573 struct vtn_value *val, int member,
574 const struct vtn_decoration *dec,
575 void *void_ctx)
576 {
577 if (dec->decoration != SpvDecorationMatrixStride)
578 return;
579 assert(member >= 0);
580
581 struct member_decoration_ctx *ctx = void_ctx;
582
583 struct vtn_type *mat_type = mutable_matrix_member(b, ctx->type, member);
584 if (mat_type->row_major) {
585 mat_type->array_element = vtn_type_copy(b, mat_type->array_element);
586 mat_type->stride = mat_type->array_element->stride;
587 mat_type->array_element->stride = dec->literals[0];
588 } else {
589 assert(mat_type->array_element->stride > 0);
590 mat_type->stride = dec->literals[0];
591 }
592 }
593
594 static void
595 type_decoration_cb(struct vtn_builder *b,
596 struct vtn_value *val, int member,
597 const struct vtn_decoration *dec, void *ctx)
598 {
599 struct vtn_type *type = val->type;
600
601 if (member != -1)
602 return;
603
604 switch (dec->decoration) {
605 case SpvDecorationArrayStride:
606 type->stride = dec->literals[0];
607 break;
608 case SpvDecorationBlock:
609 type->block = true;
610 break;
611 case SpvDecorationBufferBlock:
612 type->buffer_block = true;
613 break;
614 case SpvDecorationGLSLShared:
615 case SpvDecorationGLSLPacked:
616 /* Ignore these, since we get explicit offsets anyways */
617 break;
618
619 case SpvDecorationRowMajor:
620 case SpvDecorationColMajor:
621 case SpvDecorationMatrixStride:
622 case SpvDecorationBuiltIn:
623 case SpvDecorationNoPerspective:
624 case SpvDecorationFlat:
625 case SpvDecorationPatch:
626 case SpvDecorationCentroid:
627 case SpvDecorationSample:
628 case SpvDecorationVolatile:
629 case SpvDecorationCoherent:
630 case SpvDecorationNonWritable:
631 case SpvDecorationNonReadable:
632 case SpvDecorationUniform:
633 case SpvDecorationStream:
634 case SpvDecorationLocation:
635 case SpvDecorationComponent:
636 case SpvDecorationOffset:
637 case SpvDecorationXfbBuffer:
638 case SpvDecorationXfbStride:
639 vtn_warn("Decoration only allowed for struct members: %s",
640 spirv_decoration_to_string(dec->decoration));
641 break;
642
643 case SpvDecorationRelaxedPrecision:
644 case SpvDecorationSpecId:
645 case SpvDecorationInvariant:
646 case SpvDecorationRestrict:
647 case SpvDecorationAliased:
648 case SpvDecorationConstant:
649 case SpvDecorationIndex:
650 case SpvDecorationBinding:
651 case SpvDecorationDescriptorSet:
652 case SpvDecorationLinkageAttributes:
653 case SpvDecorationNoContraction:
654 case SpvDecorationInputAttachmentIndex:
655 vtn_warn("Decoration not allowed on types: %s",
656 spirv_decoration_to_string(dec->decoration));
657 break;
658
659 case SpvDecorationCPacked:
660 case SpvDecorationSaturatedConversion:
661 case SpvDecorationFuncParamAttr:
662 case SpvDecorationFPRoundingMode:
663 case SpvDecorationFPFastMathMode:
664 case SpvDecorationAlignment:
665 vtn_warn("Decoration only allowed for CL-style kernels: %s",
666 spirv_decoration_to_string(dec->decoration));
667 break;
668
669 default:
670 unreachable("Unhandled decoration");
671 }
672 }
673
674 static unsigned
675 translate_image_format(SpvImageFormat format)
676 {
677 switch (format) {
678 case SpvImageFormatUnknown: return 0; /* GL_NONE */
679 case SpvImageFormatRgba32f: return 0x8814; /* GL_RGBA32F */
680 case SpvImageFormatRgba16f: return 0x881A; /* GL_RGBA16F */
681 case SpvImageFormatR32f: return 0x822E; /* GL_R32F */
682 case SpvImageFormatRgba8: return 0x8058; /* GL_RGBA8 */
683 case SpvImageFormatRgba8Snorm: return 0x8F97; /* GL_RGBA8_SNORM */
684 case SpvImageFormatRg32f: return 0x8230; /* GL_RG32F */
685 case SpvImageFormatRg16f: return 0x822F; /* GL_RG16F */
686 case SpvImageFormatR11fG11fB10f: return 0x8C3A; /* GL_R11F_G11F_B10F */
687 case SpvImageFormatR16f: return 0x822D; /* GL_R16F */
688 case SpvImageFormatRgba16: return 0x805B; /* GL_RGBA16 */
689 case SpvImageFormatRgb10A2: return 0x8059; /* GL_RGB10_A2 */
690 case SpvImageFormatRg16: return 0x822C; /* GL_RG16 */
691 case SpvImageFormatRg8: return 0x822B; /* GL_RG8 */
692 case SpvImageFormatR16: return 0x822A; /* GL_R16 */
693 case SpvImageFormatR8: return 0x8229; /* GL_R8 */
694 case SpvImageFormatRgba16Snorm: return 0x8F9B; /* GL_RGBA16_SNORM */
695 case SpvImageFormatRg16Snorm: return 0x8F99; /* GL_RG16_SNORM */
696 case SpvImageFormatRg8Snorm: return 0x8F95; /* GL_RG8_SNORM */
697 case SpvImageFormatR16Snorm: return 0x8F98; /* GL_R16_SNORM */
698 case SpvImageFormatR8Snorm: return 0x8F94; /* GL_R8_SNORM */
699 case SpvImageFormatRgba32i: return 0x8D82; /* GL_RGBA32I */
700 case SpvImageFormatRgba16i: return 0x8D88; /* GL_RGBA16I */
701 case SpvImageFormatRgba8i: return 0x8D8E; /* GL_RGBA8I */
702 case SpvImageFormatR32i: return 0x8235; /* GL_R32I */
703 case SpvImageFormatRg32i: return 0x823B; /* GL_RG32I */
704 case SpvImageFormatRg16i: return 0x8239; /* GL_RG16I */
705 case SpvImageFormatRg8i: return 0x8237; /* GL_RG8I */
706 case SpvImageFormatR16i: return 0x8233; /* GL_R16I */
707 case SpvImageFormatR8i: return 0x8231; /* GL_R8I */
708 case SpvImageFormatRgba32ui: return 0x8D70; /* GL_RGBA32UI */
709 case SpvImageFormatRgba16ui: return 0x8D76; /* GL_RGBA16UI */
710 case SpvImageFormatRgba8ui: return 0x8D7C; /* GL_RGBA8UI */
711 case SpvImageFormatR32ui: return 0x8236; /* GL_R32UI */
712 case SpvImageFormatRgb10a2ui: return 0x906F; /* GL_RGB10_A2UI */
713 case SpvImageFormatRg32ui: return 0x823C; /* GL_RG32UI */
714 case SpvImageFormatRg16ui: return 0x823A; /* GL_RG16UI */
715 case SpvImageFormatRg8ui: return 0x8238; /* GL_RG8UI */
716 case SpvImageFormatR16ui: return 0x823A; /* GL_RG16UI */
717 case SpvImageFormatR8ui: return 0x8232; /* GL_R8UI */
718 default:
719 assert(!"Invalid image format");
720 return 0;
721 }
722 }
723
724 static void
725 vtn_handle_type(struct vtn_builder *b, SpvOp opcode,
726 const uint32_t *w, unsigned count)
727 {
728 struct vtn_value *val = vtn_push_value(b, w[1], vtn_value_type_type);
729
730 val->type = rzalloc(b, struct vtn_type);
731 val->type->is_builtin = false;
732 val->type->val = val;
733
734 switch (opcode) {
735 case SpvOpTypeVoid:
736 val->type->type = glsl_void_type();
737 break;
738 case SpvOpTypeBool:
739 val->type->type = glsl_bool_type();
740 break;
741 case SpvOpTypeInt: {
742 int bit_size = w[2];
743 const bool signedness = w[3];
744 if (bit_size == 64)
745 val->type->type = (signedness ? glsl_int64_t_type() : glsl_uint64_t_type());
746 else
747 val->type->type = (signedness ? glsl_int_type() : glsl_uint_type());
748 break;
749 }
750 case SpvOpTypeFloat: {
751 int bit_size = w[2];
752 val->type->type = bit_size == 64 ? glsl_double_type() : glsl_float_type();
753 break;
754 }
755
756 case SpvOpTypeVector: {
757 struct vtn_type *base = vtn_value(b, w[2], vtn_value_type_type)->type;
758 unsigned elems = w[3];
759
760 assert(glsl_type_is_scalar(base->type));
761 val->type->type = glsl_vector_type(glsl_get_base_type(base->type), elems);
762
763 /* Vectors implicitly have sizeof(base_type) stride. For now, this
764 * is always 4 bytes. This will have to change if we want to start
765 * supporting doubles or half-floats.
766 */
767 val->type->stride = glsl_get_bit_size(base->type) / 8;
768 val->type->array_element = base;
769 break;
770 }
771
772 case SpvOpTypeMatrix: {
773 struct vtn_type *base = vtn_value(b, w[2], vtn_value_type_type)->type;
774 unsigned columns = w[3];
775
776 assert(glsl_type_is_vector(base->type));
777 val->type->type = glsl_matrix_type(glsl_get_base_type(base->type),
778 glsl_get_vector_elements(base->type),
779 columns);
780 assert(!glsl_type_is_error(val->type->type));
781 val->type->array_element = base;
782 val->type->row_major = false;
783 val->type->stride = 0;
784 break;
785 }
786
787 case SpvOpTypeRuntimeArray:
788 case SpvOpTypeArray: {
789 struct vtn_type *array_element =
790 vtn_value(b, w[2], vtn_value_type_type)->type;
791
792 unsigned length;
793 if (opcode == SpvOpTypeRuntimeArray) {
794 /* A length of 0 is used to denote unsized arrays */
795 length = 0;
796 } else {
797 length =
798 vtn_value(b, w[3], vtn_value_type_constant)->constant->values[0].u32[0];
799 }
800
801 val->type->type = glsl_array_type(array_element->type, length);
802 val->type->array_element = array_element;
803 val->type->stride = 0;
804 break;
805 }
806
807 case SpvOpTypeStruct: {
808 unsigned num_fields = count - 2;
809 val->type->members = ralloc_array(b, struct vtn_type *, num_fields);
810 val->type->offsets = ralloc_array(b, unsigned, num_fields);
811
812 NIR_VLA(struct glsl_struct_field, fields, count);
813 for (unsigned i = 0; i < num_fields; i++) {
814 val->type->members[i] =
815 vtn_value(b, w[i + 2], vtn_value_type_type)->type;
816 fields[i] = (struct glsl_struct_field) {
817 .type = val->type->members[i]->type,
818 .name = ralloc_asprintf(b, "field%d", i),
819 .location = -1,
820 };
821 }
822
823 struct member_decoration_ctx ctx = {
824 .num_fields = num_fields,
825 .fields = fields,
826 .type = val->type
827 };
828
829 vtn_foreach_decoration(b, val, struct_member_decoration_cb, &ctx);
830 vtn_foreach_decoration(b, val, struct_member_matrix_stride_cb, &ctx);
831
832 const char *name = val->name ? val->name : "struct";
833
834 val->type->type = glsl_struct_type(fields, num_fields, name);
835 break;
836 }
837
838 case SpvOpTypeFunction: {
839 const struct glsl_type *return_type =
840 vtn_value(b, w[2], vtn_value_type_type)->type->type;
841 NIR_VLA(struct glsl_function_param, params, count - 3);
842 for (unsigned i = 0; i < count - 3; i++) {
843 params[i].type = vtn_value(b, w[i + 3], vtn_value_type_type)->type->type;
844
845 /* FIXME: */
846 params[i].in = true;
847 params[i].out = true;
848 }
849 val->type->type = glsl_function_type(return_type, params, count - 3);
850 break;
851 }
852
853 case SpvOpTypePointer:
854 /* FIXME: For now, we'll just do the really lame thing and return
855 * the same type. The validator should ensure that the proper number
856 * of dereferences happen
857 */
858 val->type = vtn_value(b, w[3], vtn_value_type_type)->type;
859 break;
860
861 case SpvOpTypeImage: {
862 const struct glsl_type *sampled_type =
863 vtn_value(b, w[2], vtn_value_type_type)->type->type;
864
865 assert(glsl_type_is_vector_or_scalar(sampled_type));
866
867 enum glsl_sampler_dim dim;
868 switch ((SpvDim)w[3]) {
869 case SpvDim1D: dim = GLSL_SAMPLER_DIM_1D; break;
870 case SpvDim2D: dim = GLSL_SAMPLER_DIM_2D; break;
871 case SpvDim3D: dim = GLSL_SAMPLER_DIM_3D; break;
872 case SpvDimCube: dim = GLSL_SAMPLER_DIM_CUBE; break;
873 case SpvDimRect: dim = GLSL_SAMPLER_DIM_RECT; break;
874 case SpvDimBuffer: dim = GLSL_SAMPLER_DIM_BUF; break;
875 case SpvDimSubpassData: dim = GLSL_SAMPLER_DIM_SUBPASS; break;
876 default:
877 unreachable("Invalid SPIR-V Sampler dimension");
878 }
879
880 bool is_shadow = w[4];
881 bool is_array = w[5];
882 bool multisampled = w[6];
883 unsigned sampled = w[7];
884 SpvImageFormat format = w[8];
885
886 if (count > 9)
887 val->type->access_qualifier = w[9];
888 else
889 val->type->access_qualifier = SpvAccessQualifierReadWrite;
890
891 if (multisampled) {
892 if (dim == GLSL_SAMPLER_DIM_2D)
893 dim = GLSL_SAMPLER_DIM_MS;
894 else if (dim == GLSL_SAMPLER_DIM_SUBPASS)
895 dim = GLSL_SAMPLER_DIM_SUBPASS_MS;
896 else
897 assert(!"Unsupported multisampled image type");
898 }
899
900 val->type->image_format = translate_image_format(format);
901
902 if (sampled == 1) {
903 val->type->type = glsl_sampler_type(dim, is_shadow, is_array,
904 glsl_get_base_type(sampled_type));
905 } else if (sampled == 2) {
906 assert(!is_shadow);
907 val->type->type = glsl_image_type(dim, is_array,
908 glsl_get_base_type(sampled_type));
909 } else {
910 assert(!"We need to know if the image will be sampled");
911 }
912 break;
913 }
914
915 case SpvOpTypeSampledImage:
916 val->type = vtn_value(b, w[2], vtn_value_type_type)->type;
917 break;
918
919 case SpvOpTypeSampler:
920 /* The actual sampler type here doesn't really matter. It gets
921 * thrown away the moment you combine it with an image. What really
922 * matters is that it's a sampler type as opposed to an integer type
923 * so the backend knows what to do.
924 */
925 val->type->type = glsl_bare_sampler_type();
926 break;
927
928 case SpvOpTypeOpaque:
929 case SpvOpTypeEvent:
930 case SpvOpTypeDeviceEvent:
931 case SpvOpTypeReserveId:
932 case SpvOpTypeQueue:
933 case SpvOpTypePipe:
934 default:
935 unreachable("Unhandled opcode");
936 }
937
938 vtn_foreach_decoration(b, val, type_decoration_cb, NULL);
939 }
940
941 static nir_constant *
942 vtn_null_constant(struct vtn_builder *b, const struct glsl_type *type)
943 {
944 nir_constant *c = rzalloc(b, nir_constant);
945
946 switch (glsl_get_base_type(type)) {
947 case GLSL_TYPE_INT:
948 case GLSL_TYPE_UINT:
949 case GLSL_TYPE_INT64:
950 case GLSL_TYPE_UINT64:
951 case GLSL_TYPE_BOOL:
952 case GLSL_TYPE_FLOAT:
953 case GLSL_TYPE_DOUBLE:
954 /* Nothing to do here. It's already initialized to zero */
955 break;
956
957 case GLSL_TYPE_ARRAY:
958 assert(glsl_get_length(type) > 0);
959 c->num_elements = glsl_get_length(type);
960 c->elements = ralloc_array(b, nir_constant *, c->num_elements);
961
962 c->elements[0] = vtn_null_constant(b, glsl_get_array_element(type));
963 for (unsigned i = 1; i < c->num_elements; i++)
964 c->elements[i] = c->elements[0];
965 break;
966
967 case GLSL_TYPE_STRUCT:
968 c->num_elements = glsl_get_length(type);
969 c->elements = ralloc_array(b, nir_constant *, c->num_elements);
970
971 for (unsigned i = 0; i < c->num_elements; i++) {
972 c->elements[i] = vtn_null_constant(b, glsl_get_struct_field(type, i));
973 }
974 break;
975
976 default:
977 unreachable("Invalid type for null constant");
978 }
979
980 return c;
981 }
982
983 static void
984 spec_constant_decoration_cb(struct vtn_builder *b, struct vtn_value *v,
985 int member, const struct vtn_decoration *dec,
986 void *data)
987 {
988 assert(member == -1);
989 if (dec->decoration != SpvDecorationSpecId)
990 return;
991
992 struct spec_constant_value *const_value = data;
993
994 for (unsigned i = 0; i < b->num_specializations; i++) {
995 if (b->specializations[i].id == dec->literals[0]) {
996 if (const_value->is_double)
997 const_value->data64 = b->specializations[i].data64;
998 else
999 const_value->data32 = b->specializations[i].data32;
1000 return;
1001 }
1002 }
1003 }
1004
1005 static uint32_t
1006 get_specialization(struct vtn_builder *b, struct vtn_value *val,
1007 uint32_t const_value)
1008 {
1009 struct spec_constant_value data;
1010 data.is_double = false;
1011 data.data32 = const_value;
1012 vtn_foreach_decoration(b, val, spec_constant_decoration_cb, &data);
1013 return data.data32;
1014 }
1015
1016 static uint64_t
1017 get_specialization64(struct vtn_builder *b, struct vtn_value *val,
1018 uint64_t const_value)
1019 {
1020 struct spec_constant_value data;
1021 data.is_double = true;
1022 data.data64 = const_value;
1023 vtn_foreach_decoration(b, val, spec_constant_decoration_cb, &data);
1024 return data.data64;
1025 }
1026
1027 static void
1028 handle_workgroup_size_decoration_cb(struct vtn_builder *b,
1029 struct vtn_value *val,
1030 int member,
1031 const struct vtn_decoration *dec,
1032 void *data)
1033 {
1034 assert(member == -1);
1035 if (dec->decoration != SpvDecorationBuiltIn ||
1036 dec->literals[0] != SpvBuiltInWorkgroupSize)
1037 return;
1038
1039 assert(val->const_type == glsl_vector_type(GLSL_TYPE_UINT, 3));
1040
1041 b->shader->info.cs.local_size[0] = val->constant->values[0].u32[0];
1042 b->shader->info.cs.local_size[1] = val->constant->values[0].u32[1];
1043 b->shader->info.cs.local_size[2] = val->constant->values[0].u32[2];
1044 }
1045
1046 static void
1047 vtn_handle_constant(struct vtn_builder *b, SpvOp opcode,
1048 const uint32_t *w, unsigned count)
1049 {
1050 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_constant);
1051 val->const_type = vtn_value(b, w[1], vtn_value_type_type)->type->type;
1052 val->constant = rzalloc(b, nir_constant);
1053 switch (opcode) {
1054 case SpvOpConstantTrue:
1055 assert(val->const_type == glsl_bool_type());
1056 val->constant->values[0].u32[0] = NIR_TRUE;
1057 break;
1058 case SpvOpConstantFalse:
1059 assert(val->const_type == glsl_bool_type());
1060 val->constant->values[0].u32[0] = NIR_FALSE;
1061 break;
1062
1063 case SpvOpSpecConstantTrue:
1064 case SpvOpSpecConstantFalse: {
1065 assert(val->const_type == glsl_bool_type());
1066 uint32_t int_val =
1067 get_specialization(b, val, (opcode == SpvOpSpecConstantTrue));
1068 val->constant->values[0].u32[0] = int_val ? NIR_TRUE : NIR_FALSE;
1069 break;
1070 }
1071
1072 case SpvOpConstant: {
1073 assert(glsl_type_is_scalar(val->const_type));
1074 int bit_size = glsl_get_bit_size(val->const_type);
1075 if (bit_size == 64) {
1076 val->constant->values->u32[0] = w[3];
1077 val->constant->values->u32[1] = w[4];
1078 } else {
1079 assert(bit_size == 32);
1080 val->constant->values->u32[0] = w[3];
1081 }
1082 break;
1083 }
1084 case SpvOpSpecConstant: {
1085 assert(glsl_type_is_scalar(val->const_type));
1086 val->constant->values[0].u32[0] = get_specialization(b, val, w[3]);
1087 int bit_size = glsl_get_bit_size(val->const_type);
1088 if (bit_size == 64)
1089 val->constant->values[0].u64[0] =
1090 get_specialization64(b, val, vtn_u64_literal(&w[3]));
1091 else
1092 val->constant->values[0].u32[0] = get_specialization(b, val, w[3]);
1093 break;
1094 }
1095 case SpvOpSpecConstantComposite:
1096 case SpvOpConstantComposite: {
1097 unsigned elem_count = count - 3;
1098 nir_constant **elems = ralloc_array(b, nir_constant *, elem_count);
1099 for (unsigned i = 0; i < elem_count; i++)
1100 elems[i] = vtn_value(b, w[i + 3], vtn_value_type_constant)->constant;
1101
1102 switch (glsl_get_base_type(val->const_type)) {
1103 case GLSL_TYPE_UINT:
1104 case GLSL_TYPE_INT:
1105 case GLSL_TYPE_UINT64:
1106 case GLSL_TYPE_INT64:
1107 case GLSL_TYPE_FLOAT:
1108 case GLSL_TYPE_BOOL:
1109 case GLSL_TYPE_DOUBLE: {
1110 int bit_size = glsl_get_bit_size(val->const_type);
1111 if (glsl_type_is_matrix(val->const_type)) {
1112 assert(glsl_get_matrix_columns(val->const_type) == elem_count);
1113 for (unsigned i = 0; i < elem_count; i++)
1114 val->constant->values[i] = elems[i]->values[0];
1115 } else {
1116 assert(glsl_type_is_vector(val->const_type));
1117 assert(glsl_get_vector_elements(val->const_type) == elem_count);
1118 for (unsigned i = 0; i < elem_count; i++) {
1119 if (bit_size == 64) {
1120 val->constant->values[0].u64[i] = elems[i]->values[0].u64[0];
1121 } else {
1122 assert(bit_size == 32);
1123 val->constant->values[0].u32[i] = elems[i]->values[0].u32[0];
1124 }
1125 }
1126 }
1127 ralloc_free(elems);
1128 break;
1129 }
1130 case GLSL_TYPE_STRUCT:
1131 case GLSL_TYPE_ARRAY:
1132 ralloc_steal(val->constant, elems);
1133 val->constant->num_elements = elem_count;
1134 val->constant->elements = elems;
1135 break;
1136
1137 default:
1138 unreachable("Unsupported type for constants");
1139 }
1140 break;
1141 }
1142
1143 case SpvOpSpecConstantOp: {
1144 SpvOp opcode = get_specialization(b, val, w[3]);
1145 switch (opcode) {
1146 case SpvOpVectorShuffle: {
1147 struct vtn_value *v0 = &b->values[w[4]];
1148 struct vtn_value *v1 = &b->values[w[5]];
1149
1150 assert(v0->value_type == vtn_value_type_constant ||
1151 v0->value_type == vtn_value_type_undef);
1152 assert(v1->value_type == vtn_value_type_constant ||
1153 v1->value_type == vtn_value_type_undef);
1154
1155 unsigned len0 = v0->value_type == vtn_value_type_constant ?
1156 glsl_get_vector_elements(v0->const_type) :
1157 glsl_get_vector_elements(v0->type->type);
1158 unsigned len1 = v1->value_type == vtn_value_type_constant ?
1159 glsl_get_vector_elements(v1->const_type) :
1160 glsl_get_vector_elements(v1->type->type);
1161
1162 assert(len0 + len1 < 16);
1163
1164 unsigned bit_size = glsl_get_bit_size(val->const_type);
1165 unsigned bit_size0 = v0->value_type == vtn_value_type_constant ?
1166 glsl_get_bit_size(v0->const_type) :
1167 glsl_get_bit_size(v0->type->type);
1168 unsigned bit_size1 = v1->value_type == vtn_value_type_constant ?
1169 glsl_get_bit_size(v1->const_type) :
1170 glsl_get_bit_size(v1->type->type);
1171
1172 assert(bit_size == bit_size0 && bit_size == bit_size1);
1173 (void)bit_size0; (void)bit_size1;
1174
1175 if (bit_size == 64) {
1176 uint64_t u64[8];
1177 if (v0->value_type == vtn_value_type_constant) {
1178 for (unsigned i = 0; i < len0; i++)
1179 u64[i] = v0->constant->values[0].u64[i];
1180 }
1181 if (v1->value_type == vtn_value_type_constant) {
1182 for (unsigned i = 0; i < len1; i++)
1183 u64[len0 + i] = v1->constant->values[0].u64[i];
1184 }
1185
1186 for (unsigned i = 0, j = 0; i < count - 6; i++, j++) {
1187 uint32_t comp = w[i + 6];
1188 /* If component is not used, set the value to a known constant
1189 * to detect if it is wrongly used.
1190 */
1191 if (comp == (uint32_t)-1)
1192 val->constant->values[0].u64[j] = 0xdeadbeefdeadbeef;
1193 else
1194 val->constant->values[0].u64[j] = u64[comp];
1195 }
1196 } else {
1197 uint32_t u32[8];
1198 if (v0->value_type == vtn_value_type_constant) {
1199 for (unsigned i = 0; i < len0; i++)
1200 u32[i] = v0->constant->values[0].u32[i];
1201 }
1202 if (v1->value_type == vtn_value_type_constant) {
1203 for (unsigned i = 0; i < len1; i++)
1204 u32[len0 + i] = v1->constant->values[0].u32[i];
1205 }
1206
1207 for (unsigned i = 0, j = 0; i < count - 6; i++, j++) {
1208 uint32_t comp = w[i + 6];
1209 /* If component is not used, set the value to a known constant
1210 * to detect if it is wrongly used.
1211 */
1212 if (comp == (uint32_t)-1)
1213 val->constant->values[0].u32[j] = 0xdeadbeef;
1214 else
1215 val->constant->values[0].u32[j] = u32[comp];
1216 }
1217 }
1218 break;
1219 }
1220
1221 case SpvOpCompositeExtract:
1222 case SpvOpCompositeInsert: {
1223 struct vtn_value *comp;
1224 unsigned deref_start;
1225 struct nir_constant **c;
1226 if (opcode == SpvOpCompositeExtract) {
1227 comp = vtn_value(b, w[4], vtn_value_type_constant);
1228 deref_start = 5;
1229 c = &comp->constant;
1230 } else {
1231 comp = vtn_value(b, w[5], vtn_value_type_constant);
1232 deref_start = 6;
1233 val->constant = nir_constant_clone(comp->constant,
1234 (nir_variable *)b);
1235 c = &val->constant;
1236 }
1237
1238 int elem = -1;
1239 int col = 0;
1240 const struct glsl_type *type = comp->const_type;
1241 for (unsigned i = deref_start; i < count; i++) {
1242 switch (glsl_get_base_type(type)) {
1243 case GLSL_TYPE_UINT:
1244 case GLSL_TYPE_INT:
1245 case GLSL_TYPE_UINT64:
1246 case GLSL_TYPE_INT64:
1247 case GLSL_TYPE_FLOAT:
1248 case GLSL_TYPE_DOUBLE:
1249 case GLSL_TYPE_BOOL:
1250 /* If we hit this granularity, we're picking off an element */
1251 if (glsl_type_is_matrix(type)) {
1252 assert(col == 0 && elem == -1);
1253 col = w[i];
1254 elem = 0;
1255 type = glsl_get_column_type(type);
1256 } else {
1257 assert(elem <= 0 && glsl_type_is_vector(type));
1258 elem = w[i];
1259 type = glsl_scalar_type(glsl_get_base_type(type));
1260 }
1261 continue;
1262
1263 case GLSL_TYPE_ARRAY:
1264 c = &(*c)->elements[w[i]];
1265 type = glsl_get_array_element(type);
1266 continue;
1267
1268 case GLSL_TYPE_STRUCT:
1269 c = &(*c)->elements[w[i]];
1270 type = glsl_get_struct_field(type, w[i]);
1271 continue;
1272
1273 default:
1274 unreachable("Invalid constant type");
1275 }
1276 }
1277
1278 if (opcode == SpvOpCompositeExtract) {
1279 if (elem == -1) {
1280 val->constant = *c;
1281 } else {
1282 unsigned num_components = glsl_get_vector_elements(type);
1283 unsigned bit_size = glsl_get_bit_size(type);
1284 for (unsigned i = 0; i < num_components; i++)
1285 if (bit_size == 64) {
1286 val->constant->values[0].u64[i] = (*c)->values[col].u64[elem + i];
1287 } else {
1288 assert(bit_size == 32);
1289 val->constant->values[0].u32[i] = (*c)->values[col].u32[elem + i];
1290 }
1291 }
1292 } else {
1293 struct vtn_value *insert =
1294 vtn_value(b, w[4], vtn_value_type_constant);
1295 assert(insert->const_type == type);
1296 if (elem == -1) {
1297 *c = insert->constant;
1298 } else {
1299 unsigned num_components = glsl_get_vector_elements(type);
1300 unsigned bit_size = glsl_get_bit_size(type);
1301 for (unsigned i = 0; i < num_components; i++)
1302 if (bit_size == 64) {
1303 (*c)->values[col].u64[elem + i] = insert->constant->values[0].u64[i];
1304 } else {
1305 assert(bit_size == 32);
1306 (*c)->values[col].u32[elem + i] = insert->constant->values[0].u32[i];
1307 }
1308 }
1309 }
1310 break;
1311 }
1312
1313 default: {
1314 bool swap;
1315 nir_alu_type dst_alu_type = nir_get_nir_type_for_glsl_type(val->const_type);
1316 nir_alu_type src_alu_type = dst_alu_type;
1317 nir_op op = vtn_nir_alu_op_for_spirv_opcode(opcode, &swap, src_alu_type, dst_alu_type);
1318
1319 unsigned num_components = glsl_get_vector_elements(val->const_type);
1320 unsigned bit_size =
1321 glsl_get_bit_size(val->const_type);
1322
1323 nir_const_value src[4];
1324 assert(count <= 7);
1325 for (unsigned i = 0; i < count - 4; i++) {
1326 nir_constant *c =
1327 vtn_value(b, w[4 + i], vtn_value_type_constant)->constant;
1328
1329 unsigned j = swap ? 1 - i : i;
1330 assert(bit_size == 32);
1331 src[j] = c->values[0];
1332 }
1333
1334 val->constant->values[0] =
1335 nir_eval_const_opcode(op, num_components, bit_size, src);
1336 break;
1337 } /* default */
1338 }
1339 break;
1340 }
1341
1342 case SpvOpConstantNull:
1343 val->constant = vtn_null_constant(b, val->const_type);
1344 break;
1345
1346 case SpvOpConstantSampler:
1347 assert(!"OpConstantSampler requires Kernel Capability");
1348 break;
1349
1350 default:
1351 unreachable("Unhandled opcode");
1352 }
1353
1354 /* Now that we have the value, update the workgroup size if needed */
1355 vtn_foreach_decoration(b, val, handle_workgroup_size_decoration_cb, NULL);
1356 }
1357
1358 static void
1359 vtn_handle_function_call(struct vtn_builder *b, SpvOp opcode,
1360 const uint32_t *w, unsigned count)
1361 {
1362 struct nir_function *callee =
1363 vtn_value(b, w[3], vtn_value_type_function)->func->impl->function;
1364
1365 nir_call_instr *call = nir_call_instr_create(b->nb.shader, callee);
1366 for (unsigned i = 0; i < call->num_params; i++) {
1367 unsigned arg_id = w[4 + i];
1368 struct vtn_value *arg = vtn_untyped_value(b, arg_id);
1369 if (arg->value_type == vtn_value_type_pointer) {
1370 nir_deref_var *d = vtn_pointer_to_deref(b, arg->pointer);
1371 call->params[i] = nir_deref_var_clone(d, call);
1372 } else {
1373 struct vtn_ssa_value *arg_ssa = vtn_ssa_value(b, arg_id);
1374
1375 /* Make a temporary to store the argument in */
1376 nir_variable *tmp =
1377 nir_local_variable_create(b->impl, arg_ssa->type, "arg_tmp");
1378 call->params[i] = nir_deref_var_create(call, tmp);
1379
1380 vtn_local_store(b, arg_ssa, call->params[i]);
1381 }
1382 }
1383
1384 nir_variable *out_tmp = NULL;
1385 if (!glsl_type_is_void(callee->return_type)) {
1386 out_tmp = nir_local_variable_create(b->impl, callee->return_type,
1387 "out_tmp");
1388 call->return_deref = nir_deref_var_create(call, out_tmp);
1389 }
1390
1391 nir_builder_instr_insert(&b->nb, &call->instr);
1392
1393 if (glsl_type_is_void(callee->return_type)) {
1394 vtn_push_value(b, w[2], vtn_value_type_undef);
1395 } else {
1396 struct vtn_value *retval = vtn_push_value(b, w[2], vtn_value_type_ssa);
1397 retval->ssa = vtn_local_load(b, call->return_deref);
1398 }
1399 }
1400
1401 struct vtn_ssa_value *
1402 vtn_create_ssa_value(struct vtn_builder *b, const struct glsl_type *type)
1403 {
1404 struct vtn_ssa_value *val = rzalloc(b, struct vtn_ssa_value);
1405 val->type = type;
1406
1407 if (!glsl_type_is_vector_or_scalar(type)) {
1408 unsigned elems = glsl_get_length(type);
1409 val->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
1410 for (unsigned i = 0; i < elems; i++) {
1411 const struct glsl_type *child_type;
1412
1413 switch (glsl_get_base_type(type)) {
1414 case GLSL_TYPE_INT:
1415 case GLSL_TYPE_UINT:
1416 case GLSL_TYPE_INT64:
1417 case GLSL_TYPE_UINT64:
1418 case GLSL_TYPE_BOOL:
1419 case GLSL_TYPE_FLOAT:
1420 case GLSL_TYPE_DOUBLE:
1421 child_type = glsl_get_column_type(type);
1422 break;
1423 case GLSL_TYPE_ARRAY:
1424 child_type = glsl_get_array_element(type);
1425 break;
1426 case GLSL_TYPE_STRUCT:
1427 child_type = glsl_get_struct_field(type, i);
1428 break;
1429 default:
1430 unreachable("unkown base type");
1431 }
1432
1433 val->elems[i] = vtn_create_ssa_value(b, child_type);
1434 }
1435 }
1436
1437 return val;
1438 }
1439
1440 static nir_tex_src
1441 vtn_tex_src(struct vtn_builder *b, unsigned index, nir_tex_src_type type)
1442 {
1443 nir_tex_src src;
1444 src.src = nir_src_for_ssa(vtn_ssa_value(b, index)->def);
1445 src.src_type = type;
1446 return src;
1447 }
1448
1449 static void
1450 vtn_handle_texture(struct vtn_builder *b, SpvOp opcode,
1451 const uint32_t *w, unsigned count)
1452 {
1453 if (opcode == SpvOpSampledImage) {
1454 struct vtn_value *val =
1455 vtn_push_value(b, w[2], vtn_value_type_sampled_image);
1456 val->sampled_image = ralloc(b, struct vtn_sampled_image);
1457 val->sampled_image->image =
1458 vtn_value(b, w[3], vtn_value_type_pointer)->pointer;
1459 val->sampled_image->sampler =
1460 vtn_value(b, w[4], vtn_value_type_pointer)->pointer;
1461 return;
1462 } else if (opcode == SpvOpImage) {
1463 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_pointer);
1464 struct vtn_value *src_val = vtn_untyped_value(b, w[3]);
1465 if (src_val->value_type == vtn_value_type_sampled_image) {
1466 val->pointer = src_val->sampled_image->image;
1467 } else {
1468 assert(src_val->value_type == vtn_value_type_pointer);
1469 val->pointer = src_val->pointer;
1470 }
1471 return;
1472 }
1473
1474 struct vtn_type *ret_type = vtn_value(b, w[1], vtn_value_type_type)->type;
1475 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
1476
1477 struct vtn_sampled_image sampled;
1478 struct vtn_value *sampled_val = vtn_untyped_value(b, w[3]);
1479 if (sampled_val->value_type == vtn_value_type_sampled_image) {
1480 sampled = *sampled_val->sampled_image;
1481 } else {
1482 assert(sampled_val->value_type == vtn_value_type_pointer);
1483 sampled.image = NULL;
1484 sampled.sampler = sampled_val->pointer;
1485 }
1486
1487 const struct glsl_type *image_type;
1488 if (sampled.image) {
1489 image_type = sampled.image->var->var->interface_type;
1490 } else {
1491 image_type = sampled.sampler->var->var->interface_type;
1492 }
1493 const enum glsl_sampler_dim sampler_dim = glsl_get_sampler_dim(image_type);
1494 const bool is_array = glsl_sampler_type_is_array(image_type);
1495 const bool is_shadow = glsl_sampler_type_is_shadow(image_type);
1496
1497 /* Figure out the base texture operation */
1498 nir_texop texop;
1499 switch (opcode) {
1500 case SpvOpImageSampleImplicitLod:
1501 case SpvOpImageSampleDrefImplicitLod:
1502 case SpvOpImageSampleProjImplicitLod:
1503 case SpvOpImageSampleProjDrefImplicitLod:
1504 texop = nir_texop_tex;
1505 break;
1506
1507 case SpvOpImageSampleExplicitLod:
1508 case SpvOpImageSampleDrefExplicitLod:
1509 case SpvOpImageSampleProjExplicitLod:
1510 case SpvOpImageSampleProjDrefExplicitLod:
1511 texop = nir_texop_txl;
1512 break;
1513
1514 case SpvOpImageFetch:
1515 if (glsl_get_sampler_dim(image_type) == GLSL_SAMPLER_DIM_MS) {
1516 texop = nir_texop_txf_ms;
1517 } else {
1518 texop = nir_texop_txf;
1519 }
1520 break;
1521
1522 case SpvOpImageGather:
1523 case SpvOpImageDrefGather:
1524 texop = nir_texop_tg4;
1525 break;
1526
1527 case SpvOpImageQuerySizeLod:
1528 case SpvOpImageQuerySize:
1529 texop = nir_texop_txs;
1530 break;
1531
1532 case SpvOpImageQueryLod:
1533 texop = nir_texop_lod;
1534 break;
1535
1536 case SpvOpImageQueryLevels:
1537 texop = nir_texop_query_levels;
1538 break;
1539
1540 case SpvOpImageQuerySamples:
1541 texop = nir_texop_texture_samples;
1542 break;
1543
1544 default:
1545 unreachable("Unhandled opcode");
1546 }
1547
1548 nir_tex_src srcs[8]; /* 8 should be enough */
1549 nir_tex_src *p = srcs;
1550
1551 unsigned idx = 4;
1552
1553 struct nir_ssa_def *coord;
1554 unsigned coord_components;
1555 switch (opcode) {
1556 case SpvOpImageSampleImplicitLod:
1557 case SpvOpImageSampleExplicitLod:
1558 case SpvOpImageSampleDrefImplicitLod:
1559 case SpvOpImageSampleDrefExplicitLod:
1560 case SpvOpImageSampleProjImplicitLod:
1561 case SpvOpImageSampleProjExplicitLod:
1562 case SpvOpImageSampleProjDrefImplicitLod:
1563 case SpvOpImageSampleProjDrefExplicitLod:
1564 case SpvOpImageFetch:
1565 case SpvOpImageGather:
1566 case SpvOpImageDrefGather:
1567 case SpvOpImageQueryLod: {
1568 /* All these types have the coordinate as their first real argument */
1569 switch (sampler_dim) {
1570 case GLSL_SAMPLER_DIM_1D:
1571 case GLSL_SAMPLER_DIM_BUF:
1572 coord_components = 1;
1573 break;
1574 case GLSL_SAMPLER_DIM_2D:
1575 case GLSL_SAMPLER_DIM_RECT:
1576 case GLSL_SAMPLER_DIM_MS:
1577 coord_components = 2;
1578 break;
1579 case GLSL_SAMPLER_DIM_3D:
1580 case GLSL_SAMPLER_DIM_CUBE:
1581 coord_components = 3;
1582 break;
1583 default:
1584 unreachable("Invalid sampler type");
1585 }
1586
1587 if (is_array && texop != nir_texop_lod)
1588 coord_components++;
1589
1590 coord = vtn_ssa_value(b, w[idx++])->def;
1591 p->src = nir_src_for_ssa(nir_channels(&b->nb, coord,
1592 (1 << coord_components) - 1));
1593 p->src_type = nir_tex_src_coord;
1594 p++;
1595 break;
1596 }
1597
1598 default:
1599 coord = NULL;
1600 coord_components = 0;
1601 break;
1602 }
1603
1604 switch (opcode) {
1605 case SpvOpImageSampleProjImplicitLod:
1606 case SpvOpImageSampleProjExplicitLod:
1607 case SpvOpImageSampleProjDrefImplicitLod:
1608 case SpvOpImageSampleProjDrefExplicitLod:
1609 /* These have the projector as the last coordinate component */
1610 p->src = nir_src_for_ssa(nir_channel(&b->nb, coord, coord_components));
1611 p->src_type = nir_tex_src_projector;
1612 p++;
1613 break;
1614
1615 default:
1616 break;
1617 }
1618
1619 unsigned gather_component = 0;
1620 switch (opcode) {
1621 case SpvOpImageSampleDrefImplicitLod:
1622 case SpvOpImageSampleDrefExplicitLod:
1623 case SpvOpImageSampleProjDrefImplicitLod:
1624 case SpvOpImageSampleProjDrefExplicitLod:
1625 case SpvOpImageDrefGather:
1626 /* These all have an explicit depth value as their next source */
1627 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_comparator);
1628 break;
1629
1630 case SpvOpImageGather:
1631 /* This has a component as its next source */
1632 gather_component =
1633 vtn_value(b, w[idx++], vtn_value_type_constant)->constant->values[0].u32[0];
1634 break;
1635
1636 default:
1637 break;
1638 }
1639
1640 /* For OpImageQuerySizeLod, we always have an LOD */
1641 if (opcode == SpvOpImageQuerySizeLod)
1642 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_lod);
1643
1644 /* Now we need to handle some number of optional arguments */
1645 const struct vtn_ssa_value *gather_offsets = NULL;
1646 if (idx < count) {
1647 uint32_t operands = w[idx++];
1648
1649 if (operands & SpvImageOperandsBiasMask) {
1650 assert(texop == nir_texop_tex);
1651 texop = nir_texop_txb;
1652 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_bias);
1653 }
1654
1655 if (operands & SpvImageOperandsLodMask) {
1656 assert(texop == nir_texop_txl || texop == nir_texop_txf ||
1657 texop == nir_texop_txs);
1658 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_lod);
1659 }
1660
1661 if (operands & SpvImageOperandsGradMask) {
1662 assert(texop == nir_texop_txl);
1663 texop = nir_texop_txd;
1664 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ddx);
1665 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ddy);
1666 }
1667
1668 if (operands & SpvImageOperandsOffsetMask ||
1669 operands & SpvImageOperandsConstOffsetMask)
1670 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_offset);
1671
1672 if (operands & SpvImageOperandsConstOffsetsMask) {
1673 gather_offsets = vtn_ssa_value(b, w[idx++]);
1674 (*p++) = (nir_tex_src){};
1675 }
1676
1677 if (operands & SpvImageOperandsSampleMask) {
1678 assert(texop == nir_texop_txf_ms);
1679 texop = nir_texop_txf_ms;
1680 (*p++) = vtn_tex_src(b, w[idx++], nir_tex_src_ms_index);
1681 }
1682 }
1683 /* We should have now consumed exactly all of the arguments */
1684 assert(idx == count);
1685
1686 nir_tex_instr *instr = nir_tex_instr_create(b->shader, p - srcs);
1687 instr->op = texop;
1688
1689 memcpy(instr->src, srcs, instr->num_srcs * sizeof(*instr->src));
1690
1691 instr->coord_components = coord_components;
1692 instr->sampler_dim = sampler_dim;
1693 instr->is_array = is_array;
1694 instr->is_shadow = is_shadow;
1695 instr->is_new_style_shadow =
1696 is_shadow && glsl_get_components(ret_type->type) == 1;
1697 instr->component = gather_component;
1698
1699 switch (glsl_get_sampler_result_type(image_type)) {
1700 case GLSL_TYPE_FLOAT: instr->dest_type = nir_type_float; break;
1701 case GLSL_TYPE_INT: instr->dest_type = nir_type_int; break;
1702 case GLSL_TYPE_UINT: instr->dest_type = nir_type_uint; break;
1703 case GLSL_TYPE_BOOL: instr->dest_type = nir_type_bool; break;
1704 default:
1705 unreachable("Invalid base type for sampler result");
1706 }
1707
1708 nir_deref_var *sampler = vtn_pointer_to_deref(b, sampled.sampler);
1709 nir_deref_var *texture;
1710 if (sampled.image) {
1711 nir_deref_var *image = vtn_pointer_to_deref(b, sampled.image);
1712 texture = image;
1713 } else {
1714 texture = sampler;
1715 }
1716
1717 instr->texture = nir_deref_var_clone(texture, instr);
1718
1719 switch (instr->op) {
1720 case nir_texop_tex:
1721 case nir_texop_txb:
1722 case nir_texop_txl:
1723 case nir_texop_txd:
1724 /* These operations require a sampler */
1725 instr->sampler = nir_deref_var_clone(sampler, instr);
1726 break;
1727 case nir_texop_txf:
1728 case nir_texop_txf_ms:
1729 case nir_texop_txs:
1730 case nir_texop_lod:
1731 case nir_texop_tg4:
1732 case nir_texop_query_levels:
1733 case nir_texop_texture_samples:
1734 case nir_texop_samples_identical:
1735 /* These don't */
1736 instr->sampler = NULL;
1737 break;
1738 case nir_texop_txf_ms_mcs:
1739 unreachable("unexpected nir_texop_txf_ms_mcs");
1740 }
1741
1742 nir_ssa_dest_init(&instr->instr, &instr->dest,
1743 nir_tex_instr_dest_size(instr), 32, NULL);
1744
1745 assert(glsl_get_vector_elements(ret_type->type) ==
1746 nir_tex_instr_dest_size(instr));
1747
1748 nir_ssa_def *def;
1749 nir_instr *instruction;
1750 if (gather_offsets) {
1751 assert(glsl_get_base_type(gather_offsets->type) == GLSL_TYPE_ARRAY);
1752 assert(glsl_get_length(gather_offsets->type) == 4);
1753 nir_tex_instr *instrs[4] = {instr, NULL, NULL, NULL};
1754
1755 /* Copy the current instruction 4x */
1756 for (uint32_t i = 1; i < 4; i++) {
1757 instrs[i] = nir_tex_instr_create(b->shader, instr->num_srcs);
1758 instrs[i]->op = instr->op;
1759 instrs[i]->coord_components = instr->coord_components;
1760 instrs[i]->sampler_dim = instr->sampler_dim;
1761 instrs[i]->is_array = instr->is_array;
1762 instrs[i]->is_shadow = instr->is_shadow;
1763 instrs[i]->is_new_style_shadow = instr->is_new_style_shadow;
1764 instrs[i]->component = instr->component;
1765 instrs[i]->dest_type = instr->dest_type;
1766 instrs[i]->texture = nir_deref_var_clone(texture, instrs[i]);
1767 instrs[i]->sampler = NULL;
1768
1769 memcpy(instrs[i]->src, srcs, instr->num_srcs * sizeof(*instr->src));
1770
1771 nir_ssa_dest_init(&instrs[i]->instr, &instrs[i]->dest,
1772 nir_tex_instr_dest_size(instr), 32, NULL);
1773 }
1774
1775 /* Fill in the last argument with the offset from the passed in offsets
1776 * and insert the instruction into the stream.
1777 */
1778 for (uint32_t i = 0; i < 4; i++) {
1779 nir_tex_src src;
1780 src.src = nir_src_for_ssa(gather_offsets->elems[i]->def);
1781 src.src_type = nir_tex_src_offset;
1782 instrs[i]->src[instrs[i]->num_srcs - 1] = src;
1783 nir_builder_instr_insert(&b->nb, &instrs[i]->instr);
1784 }
1785
1786 /* Combine the results of the 4 instructions by taking their .w
1787 * components
1788 */
1789 nir_alu_instr *vec4 = nir_alu_instr_create(b->shader, nir_op_vec4);
1790 nir_ssa_dest_init(&vec4->instr, &vec4->dest.dest, 4, 32, NULL);
1791 vec4->dest.write_mask = 0xf;
1792 for (uint32_t i = 0; i < 4; i++) {
1793 vec4->src[i].src = nir_src_for_ssa(&instrs[i]->dest.ssa);
1794 vec4->src[i].swizzle[0] = 3;
1795 }
1796 def = &vec4->dest.dest.ssa;
1797 instruction = &vec4->instr;
1798 } else {
1799 def = &instr->dest.ssa;
1800 instruction = &instr->instr;
1801 }
1802
1803 val->ssa = vtn_create_ssa_value(b, ret_type->type);
1804 val->ssa->def = def;
1805
1806 nir_builder_instr_insert(&b->nb, instruction);
1807 }
1808
1809 static void
1810 fill_common_atomic_sources(struct vtn_builder *b, SpvOp opcode,
1811 const uint32_t *w, nir_src *src)
1812 {
1813 switch (opcode) {
1814 case SpvOpAtomicIIncrement:
1815 src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, 1));
1816 break;
1817
1818 case SpvOpAtomicIDecrement:
1819 src[0] = nir_src_for_ssa(nir_imm_int(&b->nb, -1));
1820 break;
1821
1822 case SpvOpAtomicISub:
1823 src[0] =
1824 nir_src_for_ssa(nir_ineg(&b->nb, vtn_ssa_value(b, w[6])->def));
1825 break;
1826
1827 case SpvOpAtomicCompareExchange:
1828 src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[8])->def);
1829 src[1] = nir_src_for_ssa(vtn_ssa_value(b, w[7])->def);
1830 break;
1831
1832 case SpvOpAtomicExchange:
1833 case SpvOpAtomicIAdd:
1834 case SpvOpAtomicSMin:
1835 case SpvOpAtomicUMin:
1836 case SpvOpAtomicSMax:
1837 case SpvOpAtomicUMax:
1838 case SpvOpAtomicAnd:
1839 case SpvOpAtomicOr:
1840 case SpvOpAtomicXor:
1841 src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[6])->def);
1842 break;
1843
1844 default:
1845 unreachable("Invalid SPIR-V atomic");
1846 }
1847 }
1848
1849 static nir_ssa_def *
1850 get_image_coord(struct vtn_builder *b, uint32_t value)
1851 {
1852 struct vtn_ssa_value *coord = vtn_ssa_value(b, value);
1853
1854 /* The image_load_store intrinsics assume a 4-dim coordinate */
1855 unsigned dim = glsl_get_vector_elements(coord->type);
1856 unsigned swizzle[4];
1857 for (unsigned i = 0; i < 4; i++)
1858 swizzle[i] = MIN2(i, dim - 1);
1859
1860 return nir_swizzle(&b->nb, coord->def, swizzle, 4, false);
1861 }
1862
1863 static void
1864 vtn_handle_image(struct vtn_builder *b, SpvOp opcode,
1865 const uint32_t *w, unsigned count)
1866 {
1867 /* Just get this one out of the way */
1868 if (opcode == SpvOpImageTexelPointer) {
1869 struct vtn_value *val =
1870 vtn_push_value(b, w[2], vtn_value_type_image_pointer);
1871 val->image = ralloc(b, struct vtn_image_pointer);
1872
1873 val->image->image = vtn_value(b, w[3], vtn_value_type_pointer)->pointer;
1874 val->image->coord = get_image_coord(b, w[4]);
1875 val->image->sample = vtn_ssa_value(b, w[5])->def;
1876 return;
1877 }
1878
1879 struct vtn_image_pointer image;
1880
1881 switch (opcode) {
1882 case SpvOpAtomicExchange:
1883 case SpvOpAtomicCompareExchange:
1884 case SpvOpAtomicCompareExchangeWeak:
1885 case SpvOpAtomicIIncrement:
1886 case SpvOpAtomicIDecrement:
1887 case SpvOpAtomicIAdd:
1888 case SpvOpAtomicISub:
1889 case SpvOpAtomicLoad:
1890 case SpvOpAtomicSMin:
1891 case SpvOpAtomicUMin:
1892 case SpvOpAtomicSMax:
1893 case SpvOpAtomicUMax:
1894 case SpvOpAtomicAnd:
1895 case SpvOpAtomicOr:
1896 case SpvOpAtomicXor:
1897 image = *vtn_value(b, w[3], vtn_value_type_image_pointer)->image;
1898 break;
1899
1900 case SpvOpAtomicStore:
1901 image = *vtn_value(b, w[1], vtn_value_type_image_pointer)->image;
1902 break;
1903
1904 case SpvOpImageQuerySize:
1905 image.image = vtn_value(b, w[3], vtn_value_type_pointer)->pointer;
1906 image.coord = NULL;
1907 image.sample = NULL;
1908 break;
1909
1910 case SpvOpImageRead:
1911 image.image = vtn_value(b, w[3], vtn_value_type_pointer)->pointer;
1912 image.coord = get_image_coord(b, w[4]);
1913
1914 if (count > 5 && (w[5] & SpvImageOperandsSampleMask)) {
1915 assert(w[5] == SpvImageOperandsSampleMask);
1916 image.sample = vtn_ssa_value(b, w[6])->def;
1917 } else {
1918 image.sample = nir_ssa_undef(&b->nb, 1, 32);
1919 }
1920 break;
1921
1922 case SpvOpImageWrite:
1923 image.image = vtn_value(b, w[1], vtn_value_type_pointer)->pointer;
1924 image.coord = get_image_coord(b, w[2]);
1925
1926 /* texel = w[3] */
1927
1928 if (count > 4 && (w[4] & SpvImageOperandsSampleMask)) {
1929 assert(w[4] == SpvImageOperandsSampleMask);
1930 image.sample = vtn_ssa_value(b, w[5])->def;
1931 } else {
1932 image.sample = nir_ssa_undef(&b->nb, 1, 32);
1933 }
1934 break;
1935
1936 default:
1937 unreachable("Invalid image opcode");
1938 }
1939
1940 nir_intrinsic_op op;
1941 switch (opcode) {
1942 #define OP(S, N) case SpvOp##S: op = nir_intrinsic_image_##N; break;
1943 OP(ImageQuerySize, size)
1944 OP(ImageRead, load)
1945 OP(ImageWrite, store)
1946 OP(AtomicLoad, load)
1947 OP(AtomicStore, store)
1948 OP(AtomicExchange, atomic_exchange)
1949 OP(AtomicCompareExchange, atomic_comp_swap)
1950 OP(AtomicIIncrement, atomic_add)
1951 OP(AtomicIDecrement, atomic_add)
1952 OP(AtomicIAdd, atomic_add)
1953 OP(AtomicISub, atomic_add)
1954 OP(AtomicSMin, atomic_min)
1955 OP(AtomicUMin, atomic_min)
1956 OP(AtomicSMax, atomic_max)
1957 OP(AtomicUMax, atomic_max)
1958 OP(AtomicAnd, atomic_and)
1959 OP(AtomicOr, atomic_or)
1960 OP(AtomicXor, atomic_xor)
1961 #undef OP
1962 default:
1963 unreachable("Invalid image opcode");
1964 }
1965
1966 nir_intrinsic_instr *intrin = nir_intrinsic_instr_create(b->shader, op);
1967
1968 nir_deref_var *image_deref = vtn_pointer_to_deref(b, image.image);
1969 intrin->variables[0] = nir_deref_var_clone(image_deref, intrin);
1970
1971 /* ImageQuerySize doesn't take any extra parameters */
1972 if (opcode != SpvOpImageQuerySize) {
1973 /* The image coordinate is always 4 components but we may not have that
1974 * many. Swizzle to compensate.
1975 */
1976 unsigned swiz[4];
1977 for (unsigned i = 0; i < 4; i++)
1978 swiz[i] = i < image.coord->num_components ? i : 0;
1979 intrin->src[0] = nir_src_for_ssa(nir_swizzle(&b->nb, image.coord,
1980 swiz, 4, false));
1981 intrin->src[1] = nir_src_for_ssa(image.sample);
1982 }
1983
1984 switch (opcode) {
1985 case SpvOpAtomicLoad:
1986 case SpvOpImageQuerySize:
1987 case SpvOpImageRead:
1988 break;
1989 case SpvOpAtomicStore:
1990 intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[4])->def);
1991 break;
1992 case SpvOpImageWrite:
1993 intrin->src[2] = nir_src_for_ssa(vtn_ssa_value(b, w[3])->def);
1994 break;
1995
1996 case SpvOpAtomicIIncrement:
1997 case SpvOpAtomicIDecrement:
1998 case SpvOpAtomicExchange:
1999 case SpvOpAtomicIAdd:
2000 case SpvOpAtomicSMin:
2001 case SpvOpAtomicUMin:
2002 case SpvOpAtomicSMax:
2003 case SpvOpAtomicUMax:
2004 case SpvOpAtomicAnd:
2005 case SpvOpAtomicOr:
2006 case SpvOpAtomicXor:
2007 fill_common_atomic_sources(b, opcode, w, &intrin->src[2]);
2008 break;
2009
2010 default:
2011 unreachable("Invalid image opcode");
2012 }
2013
2014 if (opcode != SpvOpImageWrite) {
2015 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
2016 struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
2017
2018 unsigned dest_components =
2019 nir_intrinsic_infos[intrin->intrinsic].dest_components;
2020 if (intrin->intrinsic == nir_intrinsic_image_size) {
2021 dest_components = intrin->num_components =
2022 glsl_get_vector_elements(type->type);
2023 }
2024
2025 nir_ssa_dest_init(&intrin->instr, &intrin->dest,
2026 dest_components, 32, NULL);
2027
2028 nir_builder_instr_insert(&b->nb, &intrin->instr);
2029
2030 val->ssa = vtn_create_ssa_value(b, type->type);
2031 val->ssa->def = &intrin->dest.ssa;
2032 } else {
2033 nir_builder_instr_insert(&b->nb, &intrin->instr);
2034 }
2035 }
2036
2037 static nir_intrinsic_op
2038 get_ssbo_nir_atomic_op(SpvOp opcode)
2039 {
2040 switch (opcode) {
2041 case SpvOpAtomicLoad: return nir_intrinsic_load_ssbo;
2042 case SpvOpAtomicStore: return nir_intrinsic_store_ssbo;
2043 #define OP(S, N) case SpvOp##S: return nir_intrinsic_ssbo_##N;
2044 OP(AtomicExchange, atomic_exchange)
2045 OP(AtomicCompareExchange, atomic_comp_swap)
2046 OP(AtomicIIncrement, atomic_add)
2047 OP(AtomicIDecrement, atomic_add)
2048 OP(AtomicIAdd, atomic_add)
2049 OP(AtomicISub, atomic_add)
2050 OP(AtomicSMin, atomic_imin)
2051 OP(AtomicUMin, atomic_umin)
2052 OP(AtomicSMax, atomic_imax)
2053 OP(AtomicUMax, atomic_umax)
2054 OP(AtomicAnd, atomic_and)
2055 OP(AtomicOr, atomic_or)
2056 OP(AtomicXor, atomic_xor)
2057 #undef OP
2058 default:
2059 unreachable("Invalid SSBO atomic");
2060 }
2061 }
2062
2063 static nir_intrinsic_op
2064 get_shared_nir_atomic_op(SpvOp opcode)
2065 {
2066 switch (opcode) {
2067 case SpvOpAtomicLoad: return nir_intrinsic_load_var;
2068 case SpvOpAtomicStore: return nir_intrinsic_store_var;
2069 #define OP(S, N) case SpvOp##S: return nir_intrinsic_var_##N;
2070 OP(AtomicExchange, atomic_exchange)
2071 OP(AtomicCompareExchange, atomic_comp_swap)
2072 OP(AtomicIIncrement, atomic_add)
2073 OP(AtomicIDecrement, atomic_add)
2074 OP(AtomicIAdd, atomic_add)
2075 OP(AtomicISub, atomic_add)
2076 OP(AtomicSMin, atomic_imin)
2077 OP(AtomicUMin, atomic_umin)
2078 OP(AtomicSMax, atomic_imax)
2079 OP(AtomicUMax, atomic_umax)
2080 OP(AtomicAnd, atomic_and)
2081 OP(AtomicOr, atomic_or)
2082 OP(AtomicXor, atomic_xor)
2083 #undef OP
2084 default:
2085 unreachable("Invalid shared atomic");
2086 }
2087 }
2088
2089 static void
2090 vtn_handle_ssbo_or_shared_atomic(struct vtn_builder *b, SpvOp opcode,
2091 const uint32_t *w, unsigned count)
2092 {
2093 struct vtn_pointer *ptr;
2094 nir_intrinsic_instr *atomic;
2095
2096 switch (opcode) {
2097 case SpvOpAtomicLoad:
2098 case SpvOpAtomicExchange:
2099 case SpvOpAtomicCompareExchange:
2100 case SpvOpAtomicCompareExchangeWeak:
2101 case SpvOpAtomicIIncrement:
2102 case SpvOpAtomicIDecrement:
2103 case SpvOpAtomicIAdd:
2104 case SpvOpAtomicISub:
2105 case SpvOpAtomicSMin:
2106 case SpvOpAtomicUMin:
2107 case SpvOpAtomicSMax:
2108 case SpvOpAtomicUMax:
2109 case SpvOpAtomicAnd:
2110 case SpvOpAtomicOr:
2111 case SpvOpAtomicXor:
2112 ptr = vtn_value(b, w[3], vtn_value_type_pointer)->pointer;
2113 break;
2114
2115 case SpvOpAtomicStore:
2116 ptr = vtn_value(b, w[1], vtn_value_type_pointer)->pointer;
2117 break;
2118
2119 default:
2120 unreachable("Invalid SPIR-V atomic");
2121 }
2122
2123 /*
2124 SpvScope scope = w[4];
2125 SpvMemorySemanticsMask semantics = w[5];
2126 */
2127
2128 if (ptr->mode == vtn_variable_mode_workgroup) {
2129 nir_deref_var *deref = vtn_pointer_to_deref(b, ptr);
2130 const struct glsl_type *deref_type = nir_deref_tail(&deref->deref)->type;
2131 nir_intrinsic_op op = get_shared_nir_atomic_op(opcode);
2132 atomic = nir_intrinsic_instr_create(b->nb.shader, op);
2133 atomic->variables[0] = nir_deref_var_clone(deref, atomic);
2134
2135 switch (opcode) {
2136 case SpvOpAtomicLoad:
2137 atomic->num_components = glsl_get_vector_elements(deref_type);
2138 break;
2139
2140 case SpvOpAtomicStore:
2141 atomic->num_components = glsl_get_vector_elements(deref_type);
2142 nir_intrinsic_set_write_mask(atomic, (1 << atomic->num_components) - 1);
2143 atomic->src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[4])->def);
2144 break;
2145
2146 case SpvOpAtomicExchange:
2147 case SpvOpAtomicCompareExchange:
2148 case SpvOpAtomicCompareExchangeWeak:
2149 case SpvOpAtomicIIncrement:
2150 case SpvOpAtomicIDecrement:
2151 case SpvOpAtomicIAdd:
2152 case SpvOpAtomicISub:
2153 case SpvOpAtomicSMin:
2154 case SpvOpAtomicUMin:
2155 case SpvOpAtomicSMax:
2156 case SpvOpAtomicUMax:
2157 case SpvOpAtomicAnd:
2158 case SpvOpAtomicOr:
2159 case SpvOpAtomicXor:
2160 fill_common_atomic_sources(b, opcode, w, &atomic->src[0]);
2161 break;
2162
2163 default:
2164 unreachable("Invalid SPIR-V atomic");
2165
2166 }
2167 } else {
2168 assert(ptr->mode == vtn_variable_mode_ssbo);
2169 nir_ssa_def *offset, *index;
2170 offset = vtn_pointer_to_offset(b, ptr, &index, NULL);
2171
2172 nir_intrinsic_op op = get_ssbo_nir_atomic_op(opcode);
2173
2174 atomic = nir_intrinsic_instr_create(b->nb.shader, op);
2175
2176 switch (opcode) {
2177 case SpvOpAtomicLoad:
2178 atomic->num_components = glsl_get_vector_elements(ptr->type->type);
2179 atomic->src[0] = nir_src_for_ssa(index);
2180 atomic->src[1] = nir_src_for_ssa(offset);
2181 break;
2182
2183 case SpvOpAtomicStore:
2184 atomic->num_components = glsl_get_vector_elements(ptr->type->type);
2185 nir_intrinsic_set_write_mask(atomic, (1 << atomic->num_components) - 1);
2186 atomic->src[0] = nir_src_for_ssa(vtn_ssa_value(b, w[4])->def);
2187 atomic->src[1] = nir_src_for_ssa(index);
2188 atomic->src[2] = nir_src_for_ssa(offset);
2189 break;
2190
2191 case SpvOpAtomicExchange:
2192 case SpvOpAtomicCompareExchange:
2193 case SpvOpAtomicCompareExchangeWeak:
2194 case SpvOpAtomicIIncrement:
2195 case SpvOpAtomicIDecrement:
2196 case SpvOpAtomicIAdd:
2197 case SpvOpAtomicISub:
2198 case SpvOpAtomicSMin:
2199 case SpvOpAtomicUMin:
2200 case SpvOpAtomicSMax:
2201 case SpvOpAtomicUMax:
2202 case SpvOpAtomicAnd:
2203 case SpvOpAtomicOr:
2204 case SpvOpAtomicXor:
2205 atomic->src[0] = nir_src_for_ssa(index);
2206 atomic->src[1] = nir_src_for_ssa(offset);
2207 fill_common_atomic_sources(b, opcode, w, &atomic->src[2]);
2208 break;
2209
2210 default:
2211 unreachable("Invalid SPIR-V atomic");
2212 }
2213 }
2214
2215 if (opcode != SpvOpAtomicStore) {
2216 struct vtn_type *type = vtn_value(b, w[1], vtn_value_type_type)->type;
2217
2218 nir_ssa_dest_init(&atomic->instr, &atomic->dest,
2219 glsl_get_vector_elements(type->type),
2220 glsl_get_bit_size(type->type), NULL);
2221
2222 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
2223 val->ssa = rzalloc(b, struct vtn_ssa_value);
2224 val->ssa->def = &atomic->dest.ssa;
2225 val->ssa->type = type->type;
2226 }
2227
2228 nir_builder_instr_insert(&b->nb, &atomic->instr);
2229 }
2230
2231 static nir_alu_instr *
2232 create_vec(nir_shader *shader, unsigned num_components, unsigned bit_size)
2233 {
2234 nir_op op;
2235 switch (num_components) {
2236 case 1: op = nir_op_fmov; break;
2237 case 2: op = nir_op_vec2; break;
2238 case 3: op = nir_op_vec3; break;
2239 case 4: op = nir_op_vec4; break;
2240 default: unreachable("bad vector size");
2241 }
2242
2243 nir_alu_instr *vec = nir_alu_instr_create(shader, op);
2244 nir_ssa_dest_init(&vec->instr, &vec->dest.dest, num_components,
2245 bit_size, NULL);
2246 vec->dest.write_mask = (1 << num_components) - 1;
2247
2248 return vec;
2249 }
2250
2251 struct vtn_ssa_value *
2252 vtn_ssa_transpose(struct vtn_builder *b, struct vtn_ssa_value *src)
2253 {
2254 if (src->transposed)
2255 return src->transposed;
2256
2257 struct vtn_ssa_value *dest =
2258 vtn_create_ssa_value(b, glsl_transposed_type(src->type));
2259
2260 for (unsigned i = 0; i < glsl_get_matrix_columns(dest->type); i++) {
2261 nir_alu_instr *vec = create_vec(b->shader,
2262 glsl_get_matrix_columns(src->type),
2263 glsl_get_bit_size(src->type));
2264 if (glsl_type_is_vector_or_scalar(src->type)) {
2265 vec->src[0].src = nir_src_for_ssa(src->def);
2266 vec->src[0].swizzle[0] = i;
2267 } else {
2268 for (unsigned j = 0; j < glsl_get_matrix_columns(src->type); j++) {
2269 vec->src[j].src = nir_src_for_ssa(src->elems[j]->def);
2270 vec->src[j].swizzle[0] = i;
2271 }
2272 }
2273 nir_builder_instr_insert(&b->nb, &vec->instr);
2274 dest->elems[i]->def = &vec->dest.dest.ssa;
2275 }
2276
2277 dest->transposed = src;
2278
2279 return dest;
2280 }
2281
2282 nir_ssa_def *
2283 vtn_vector_extract(struct vtn_builder *b, nir_ssa_def *src, unsigned index)
2284 {
2285 unsigned swiz[4] = { index };
2286 return nir_swizzle(&b->nb, src, swiz, 1, true);
2287 }
2288
2289 nir_ssa_def *
2290 vtn_vector_insert(struct vtn_builder *b, nir_ssa_def *src, nir_ssa_def *insert,
2291 unsigned index)
2292 {
2293 nir_alu_instr *vec = create_vec(b->shader, src->num_components,
2294 src->bit_size);
2295
2296 for (unsigned i = 0; i < src->num_components; i++) {
2297 if (i == index) {
2298 vec->src[i].src = nir_src_for_ssa(insert);
2299 } else {
2300 vec->src[i].src = nir_src_for_ssa(src);
2301 vec->src[i].swizzle[0] = i;
2302 }
2303 }
2304
2305 nir_builder_instr_insert(&b->nb, &vec->instr);
2306
2307 return &vec->dest.dest.ssa;
2308 }
2309
2310 nir_ssa_def *
2311 vtn_vector_extract_dynamic(struct vtn_builder *b, nir_ssa_def *src,
2312 nir_ssa_def *index)
2313 {
2314 nir_ssa_def *dest = vtn_vector_extract(b, src, 0);
2315 for (unsigned i = 1; i < src->num_components; i++)
2316 dest = nir_bcsel(&b->nb, nir_ieq(&b->nb, index, nir_imm_int(&b->nb, i)),
2317 vtn_vector_extract(b, src, i), dest);
2318
2319 return dest;
2320 }
2321
2322 nir_ssa_def *
2323 vtn_vector_insert_dynamic(struct vtn_builder *b, nir_ssa_def *src,
2324 nir_ssa_def *insert, nir_ssa_def *index)
2325 {
2326 nir_ssa_def *dest = vtn_vector_insert(b, src, insert, 0);
2327 for (unsigned i = 1; i < src->num_components; i++)
2328 dest = nir_bcsel(&b->nb, nir_ieq(&b->nb, index, nir_imm_int(&b->nb, i)),
2329 vtn_vector_insert(b, src, insert, i), dest);
2330
2331 return dest;
2332 }
2333
2334 static nir_ssa_def *
2335 vtn_vector_shuffle(struct vtn_builder *b, unsigned num_components,
2336 nir_ssa_def *src0, nir_ssa_def *src1,
2337 const uint32_t *indices)
2338 {
2339 nir_alu_instr *vec = create_vec(b->shader, num_components, src0->bit_size);
2340
2341 for (unsigned i = 0; i < num_components; i++) {
2342 uint32_t index = indices[i];
2343 if (index == 0xffffffff) {
2344 vec->src[i].src =
2345 nir_src_for_ssa(nir_ssa_undef(&b->nb, 1, src0->bit_size));
2346 } else if (index < src0->num_components) {
2347 vec->src[i].src = nir_src_for_ssa(src0);
2348 vec->src[i].swizzle[0] = index;
2349 } else {
2350 vec->src[i].src = nir_src_for_ssa(src1);
2351 vec->src[i].swizzle[0] = index - src0->num_components;
2352 }
2353 }
2354
2355 nir_builder_instr_insert(&b->nb, &vec->instr);
2356
2357 return &vec->dest.dest.ssa;
2358 }
2359
2360 /*
2361 * Concatentates a number of vectors/scalars together to produce a vector
2362 */
2363 static nir_ssa_def *
2364 vtn_vector_construct(struct vtn_builder *b, unsigned num_components,
2365 unsigned num_srcs, nir_ssa_def **srcs)
2366 {
2367 nir_alu_instr *vec = create_vec(b->shader, num_components,
2368 srcs[0]->bit_size);
2369
2370 /* From the SPIR-V 1.1 spec for OpCompositeConstruct:
2371 *
2372 * "When constructing a vector, there must be at least two Constituent
2373 * operands."
2374 */
2375 assert(num_srcs >= 2);
2376
2377 unsigned dest_idx = 0;
2378 for (unsigned i = 0; i < num_srcs; i++) {
2379 nir_ssa_def *src = srcs[i];
2380 assert(dest_idx + src->num_components <= num_components);
2381 for (unsigned j = 0; j < src->num_components; j++) {
2382 vec->src[dest_idx].src = nir_src_for_ssa(src);
2383 vec->src[dest_idx].swizzle[0] = j;
2384 dest_idx++;
2385 }
2386 }
2387
2388 /* From the SPIR-V 1.1 spec for OpCompositeConstruct:
2389 *
2390 * "When constructing a vector, the total number of components in all
2391 * the operands must equal the number of components in Result Type."
2392 */
2393 assert(dest_idx == num_components);
2394
2395 nir_builder_instr_insert(&b->nb, &vec->instr);
2396
2397 return &vec->dest.dest.ssa;
2398 }
2399
2400 static struct vtn_ssa_value *
2401 vtn_composite_copy(void *mem_ctx, struct vtn_ssa_value *src)
2402 {
2403 struct vtn_ssa_value *dest = rzalloc(mem_ctx, struct vtn_ssa_value);
2404 dest->type = src->type;
2405
2406 if (glsl_type_is_vector_or_scalar(src->type)) {
2407 dest->def = src->def;
2408 } else {
2409 unsigned elems = glsl_get_length(src->type);
2410
2411 dest->elems = ralloc_array(mem_ctx, struct vtn_ssa_value *, elems);
2412 for (unsigned i = 0; i < elems; i++)
2413 dest->elems[i] = vtn_composite_copy(mem_ctx, src->elems[i]);
2414 }
2415
2416 return dest;
2417 }
2418
2419 static struct vtn_ssa_value *
2420 vtn_composite_insert(struct vtn_builder *b, struct vtn_ssa_value *src,
2421 struct vtn_ssa_value *insert, const uint32_t *indices,
2422 unsigned num_indices)
2423 {
2424 struct vtn_ssa_value *dest = vtn_composite_copy(b, src);
2425
2426 struct vtn_ssa_value *cur = dest;
2427 unsigned i;
2428 for (i = 0; i < num_indices - 1; i++) {
2429 cur = cur->elems[indices[i]];
2430 }
2431
2432 if (glsl_type_is_vector_or_scalar(cur->type)) {
2433 /* According to the SPIR-V spec, OpCompositeInsert may work down to
2434 * the component granularity. In that case, the last index will be
2435 * the index to insert the scalar into the vector.
2436 */
2437
2438 cur->def = vtn_vector_insert(b, cur->def, insert->def, indices[i]);
2439 } else {
2440 cur->elems[indices[i]] = insert;
2441 }
2442
2443 return dest;
2444 }
2445
2446 static struct vtn_ssa_value *
2447 vtn_composite_extract(struct vtn_builder *b, struct vtn_ssa_value *src,
2448 const uint32_t *indices, unsigned num_indices)
2449 {
2450 struct vtn_ssa_value *cur = src;
2451 for (unsigned i = 0; i < num_indices; i++) {
2452 if (glsl_type_is_vector_or_scalar(cur->type)) {
2453 assert(i == num_indices - 1);
2454 /* According to the SPIR-V spec, OpCompositeExtract may work down to
2455 * the component granularity. The last index will be the index of the
2456 * vector to extract.
2457 */
2458
2459 struct vtn_ssa_value *ret = rzalloc(b, struct vtn_ssa_value);
2460 ret->type = glsl_scalar_type(glsl_get_base_type(cur->type));
2461 ret->def = vtn_vector_extract(b, cur->def, indices[i]);
2462 return ret;
2463 } else {
2464 cur = cur->elems[indices[i]];
2465 }
2466 }
2467
2468 return cur;
2469 }
2470
2471 static void
2472 vtn_handle_composite(struct vtn_builder *b, SpvOp opcode,
2473 const uint32_t *w, unsigned count)
2474 {
2475 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_ssa);
2476 const struct glsl_type *type =
2477 vtn_value(b, w[1], vtn_value_type_type)->type->type;
2478 val->ssa = vtn_create_ssa_value(b, type);
2479
2480 switch (opcode) {
2481 case SpvOpVectorExtractDynamic:
2482 val->ssa->def = vtn_vector_extract_dynamic(b, vtn_ssa_value(b, w[3])->def,
2483 vtn_ssa_value(b, w[4])->def);
2484 break;
2485
2486 case SpvOpVectorInsertDynamic:
2487 val->ssa->def = vtn_vector_insert_dynamic(b, vtn_ssa_value(b, w[3])->def,
2488 vtn_ssa_value(b, w[4])->def,
2489 vtn_ssa_value(b, w[5])->def);
2490 break;
2491
2492 case SpvOpVectorShuffle:
2493 val->ssa->def = vtn_vector_shuffle(b, glsl_get_vector_elements(type),
2494 vtn_ssa_value(b, w[3])->def,
2495 vtn_ssa_value(b, w[4])->def,
2496 w + 5);
2497 break;
2498
2499 case SpvOpCompositeConstruct: {
2500 unsigned elems = count - 3;
2501 if (glsl_type_is_vector_or_scalar(type)) {
2502 nir_ssa_def *srcs[4];
2503 for (unsigned i = 0; i < elems; i++)
2504 srcs[i] = vtn_ssa_value(b, w[3 + i])->def;
2505 val->ssa->def =
2506 vtn_vector_construct(b, glsl_get_vector_elements(type),
2507 elems, srcs);
2508 } else {
2509 val->ssa->elems = ralloc_array(b, struct vtn_ssa_value *, elems);
2510 for (unsigned i = 0; i < elems; i++)
2511 val->ssa->elems[i] = vtn_ssa_value(b, w[3 + i]);
2512 }
2513 break;
2514 }
2515 case SpvOpCompositeExtract:
2516 val->ssa = vtn_composite_extract(b, vtn_ssa_value(b, w[3]),
2517 w + 4, count - 4);
2518 break;
2519
2520 case SpvOpCompositeInsert:
2521 val->ssa = vtn_composite_insert(b, vtn_ssa_value(b, w[4]),
2522 vtn_ssa_value(b, w[3]),
2523 w + 5, count - 5);
2524 break;
2525
2526 case SpvOpCopyObject:
2527 val->ssa = vtn_composite_copy(b, vtn_ssa_value(b, w[3]));
2528 break;
2529
2530 default:
2531 unreachable("unknown composite operation");
2532 }
2533 }
2534
2535 static void
2536 vtn_handle_barrier(struct vtn_builder *b, SpvOp opcode,
2537 const uint32_t *w, unsigned count)
2538 {
2539 nir_intrinsic_op intrinsic_op;
2540 switch (opcode) {
2541 case SpvOpEmitVertex:
2542 case SpvOpEmitStreamVertex:
2543 intrinsic_op = nir_intrinsic_emit_vertex;
2544 break;
2545 case SpvOpEndPrimitive:
2546 case SpvOpEndStreamPrimitive:
2547 intrinsic_op = nir_intrinsic_end_primitive;
2548 break;
2549 case SpvOpMemoryBarrier:
2550 intrinsic_op = nir_intrinsic_memory_barrier;
2551 break;
2552 case SpvOpControlBarrier:
2553 intrinsic_op = nir_intrinsic_barrier;
2554 break;
2555 default:
2556 unreachable("unknown barrier instruction");
2557 }
2558
2559 nir_intrinsic_instr *intrin =
2560 nir_intrinsic_instr_create(b->shader, intrinsic_op);
2561
2562 if (opcode == SpvOpEmitStreamVertex || opcode == SpvOpEndStreamPrimitive)
2563 nir_intrinsic_set_stream_id(intrin, w[1]);
2564
2565 nir_builder_instr_insert(&b->nb, &intrin->instr);
2566 }
2567
2568 static unsigned
2569 gl_primitive_from_spv_execution_mode(SpvExecutionMode mode)
2570 {
2571 switch (mode) {
2572 case SpvExecutionModeInputPoints:
2573 case SpvExecutionModeOutputPoints:
2574 return 0; /* GL_POINTS */
2575 case SpvExecutionModeInputLines:
2576 return 1; /* GL_LINES */
2577 case SpvExecutionModeInputLinesAdjacency:
2578 return 0x000A; /* GL_LINE_STRIP_ADJACENCY_ARB */
2579 case SpvExecutionModeTriangles:
2580 return 4; /* GL_TRIANGLES */
2581 case SpvExecutionModeInputTrianglesAdjacency:
2582 return 0x000C; /* GL_TRIANGLES_ADJACENCY_ARB */
2583 case SpvExecutionModeQuads:
2584 return 7; /* GL_QUADS */
2585 case SpvExecutionModeIsolines:
2586 return 0x8E7A; /* GL_ISOLINES */
2587 case SpvExecutionModeOutputLineStrip:
2588 return 3; /* GL_LINE_STRIP */
2589 case SpvExecutionModeOutputTriangleStrip:
2590 return 5; /* GL_TRIANGLE_STRIP */
2591 default:
2592 assert(!"Invalid primitive type");
2593 return 4;
2594 }
2595 }
2596
2597 static unsigned
2598 vertices_in_from_spv_execution_mode(SpvExecutionMode mode)
2599 {
2600 switch (mode) {
2601 case SpvExecutionModeInputPoints:
2602 return 1;
2603 case SpvExecutionModeInputLines:
2604 return 2;
2605 case SpvExecutionModeInputLinesAdjacency:
2606 return 4;
2607 case SpvExecutionModeTriangles:
2608 return 3;
2609 case SpvExecutionModeInputTrianglesAdjacency:
2610 return 6;
2611 default:
2612 assert(!"Invalid GS input mode");
2613 return 0;
2614 }
2615 }
2616
2617 static gl_shader_stage
2618 stage_for_execution_model(SpvExecutionModel model)
2619 {
2620 switch (model) {
2621 case SpvExecutionModelVertex:
2622 return MESA_SHADER_VERTEX;
2623 case SpvExecutionModelTessellationControl:
2624 return MESA_SHADER_TESS_CTRL;
2625 case SpvExecutionModelTessellationEvaluation:
2626 return MESA_SHADER_TESS_EVAL;
2627 case SpvExecutionModelGeometry:
2628 return MESA_SHADER_GEOMETRY;
2629 case SpvExecutionModelFragment:
2630 return MESA_SHADER_FRAGMENT;
2631 case SpvExecutionModelGLCompute:
2632 return MESA_SHADER_COMPUTE;
2633 default:
2634 unreachable("Unsupported execution model");
2635 }
2636 }
2637
2638 #define spv_check_supported(name, cap) do { \
2639 if (!(b->ext && b->ext->name)) \
2640 vtn_warn("Unsupported SPIR-V capability: %s", \
2641 spirv_capability_to_string(cap)); \
2642 } while(0)
2643
2644 static bool
2645 vtn_handle_preamble_instruction(struct vtn_builder *b, SpvOp opcode,
2646 const uint32_t *w, unsigned count)
2647 {
2648 switch (opcode) {
2649 case SpvOpSource:
2650 case SpvOpSourceExtension:
2651 case SpvOpSourceContinued:
2652 case SpvOpExtension:
2653 /* Unhandled, but these are for debug so that's ok. */
2654 break;
2655
2656 case SpvOpCapability: {
2657 SpvCapability cap = w[1];
2658 switch (cap) {
2659 case SpvCapabilityMatrix:
2660 case SpvCapabilityShader:
2661 case SpvCapabilityGeometry:
2662 case SpvCapabilityGeometryPointSize:
2663 case SpvCapabilityUniformBufferArrayDynamicIndexing:
2664 case SpvCapabilitySampledImageArrayDynamicIndexing:
2665 case SpvCapabilityStorageBufferArrayDynamicIndexing:
2666 case SpvCapabilityStorageImageArrayDynamicIndexing:
2667 case SpvCapabilityImageRect:
2668 case SpvCapabilitySampledRect:
2669 case SpvCapabilitySampled1D:
2670 case SpvCapabilityImage1D:
2671 case SpvCapabilitySampledCubeArray:
2672 case SpvCapabilitySampledBuffer:
2673 case SpvCapabilityImageBuffer:
2674 case SpvCapabilityImageQuery:
2675 case SpvCapabilityDerivativeControl:
2676 case SpvCapabilityInterpolationFunction:
2677 case SpvCapabilityMultiViewport:
2678 case SpvCapabilitySampleRateShading:
2679 case SpvCapabilityClipDistance:
2680 case SpvCapabilityCullDistance:
2681 case SpvCapabilityInputAttachment:
2682 case SpvCapabilityImageGatherExtended:
2683 case SpvCapabilityStorageImageExtendedFormats:
2684 break;
2685
2686 case SpvCapabilityGeometryStreams:
2687 case SpvCapabilityLinkage:
2688 case SpvCapabilityVector16:
2689 case SpvCapabilityFloat16Buffer:
2690 case SpvCapabilityFloat16:
2691 case SpvCapabilityInt64Atomics:
2692 case SpvCapabilityAtomicStorage:
2693 case SpvCapabilityInt16:
2694 case SpvCapabilityStorageImageMultisample:
2695 case SpvCapabilityImageCubeArray:
2696 case SpvCapabilityInt8:
2697 case SpvCapabilitySparseResidency:
2698 case SpvCapabilityMinLod:
2699 case SpvCapabilityTransformFeedback:
2700 vtn_warn("Unsupported SPIR-V capability: %s",
2701 spirv_capability_to_string(cap));
2702 break;
2703
2704 case SpvCapabilityFloat64:
2705 spv_check_supported(float64, cap);
2706 break;
2707 case SpvCapabilityInt64:
2708 spv_check_supported(int64, cap);
2709 break;
2710
2711 case SpvCapabilityAddresses:
2712 case SpvCapabilityKernel:
2713 case SpvCapabilityImageBasic:
2714 case SpvCapabilityImageReadWrite:
2715 case SpvCapabilityImageMipmap:
2716 case SpvCapabilityPipes:
2717 case SpvCapabilityGroups:
2718 case SpvCapabilityDeviceEnqueue:
2719 case SpvCapabilityLiteralSampler:
2720 case SpvCapabilityGenericPointer:
2721 vtn_warn("Unsupported OpenCL-style SPIR-V capability: %s",
2722 spirv_capability_to_string(cap));
2723 break;
2724
2725 case SpvCapabilityImageMSArray:
2726 spv_check_supported(image_ms_array, cap);
2727 break;
2728
2729 case SpvCapabilityTessellation:
2730 case SpvCapabilityTessellationPointSize:
2731 spv_check_supported(tessellation, cap);
2732 break;
2733
2734 case SpvCapabilityDrawParameters:
2735 spv_check_supported(draw_parameters, cap);
2736 break;
2737
2738 case SpvCapabilityStorageImageReadWithoutFormat:
2739 spv_check_supported(image_read_without_format, cap);
2740 break;
2741
2742 case SpvCapabilityStorageImageWriteWithoutFormat:
2743 spv_check_supported(image_write_without_format, cap);
2744 break;
2745
2746 case SpvCapabilityMultiView:
2747 spv_check_supported(multiview, cap);
2748 break;
2749
2750 default:
2751 unreachable("Unhandled capability");
2752 }
2753 break;
2754 }
2755
2756 case SpvOpExtInstImport:
2757 vtn_handle_extension(b, opcode, w, count);
2758 break;
2759
2760 case SpvOpMemoryModel:
2761 assert(w[1] == SpvAddressingModelLogical);
2762 assert(w[2] == SpvMemoryModelGLSL450);
2763 break;
2764
2765 case SpvOpEntryPoint: {
2766 struct vtn_value *entry_point = &b->values[w[2]];
2767 /* Let this be a name label regardless */
2768 unsigned name_words;
2769 entry_point->name = vtn_string_literal(b, &w[3], count - 3, &name_words);
2770
2771 if (strcmp(entry_point->name, b->entry_point_name) != 0 ||
2772 stage_for_execution_model(w[1]) != b->entry_point_stage)
2773 break;
2774
2775 assert(b->entry_point == NULL);
2776 b->entry_point = entry_point;
2777 break;
2778 }
2779
2780 case SpvOpString:
2781 vtn_push_value(b, w[1], vtn_value_type_string)->str =
2782 vtn_string_literal(b, &w[2], count - 2, NULL);
2783 break;
2784
2785 case SpvOpName:
2786 b->values[w[1]].name = vtn_string_literal(b, &w[2], count - 2, NULL);
2787 break;
2788
2789 case SpvOpMemberName:
2790 /* TODO */
2791 break;
2792
2793 case SpvOpExecutionMode:
2794 case SpvOpDecorationGroup:
2795 case SpvOpDecorate:
2796 case SpvOpMemberDecorate:
2797 case SpvOpGroupDecorate:
2798 case SpvOpGroupMemberDecorate:
2799 vtn_handle_decoration(b, opcode, w, count);
2800 break;
2801
2802 default:
2803 return false; /* End of preamble */
2804 }
2805
2806 return true;
2807 }
2808
2809 static void
2810 vtn_handle_execution_mode(struct vtn_builder *b, struct vtn_value *entry_point,
2811 const struct vtn_decoration *mode, void *data)
2812 {
2813 assert(b->entry_point == entry_point);
2814
2815 switch(mode->exec_mode) {
2816 case SpvExecutionModeOriginUpperLeft:
2817 case SpvExecutionModeOriginLowerLeft:
2818 b->origin_upper_left =
2819 (mode->exec_mode == SpvExecutionModeOriginUpperLeft);
2820 break;
2821
2822 case SpvExecutionModeEarlyFragmentTests:
2823 assert(b->shader->stage == MESA_SHADER_FRAGMENT);
2824 b->shader->info.fs.early_fragment_tests = true;
2825 break;
2826
2827 case SpvExecutionModeInvocations:
2828 assert(b->shader->stage == MESA_SHADER_GEOMETRY);
2829 b->shader->info.gs.invocations = MAX2(1, mode->literals[0]);
2830 break;
2831
2832 case SpvExecutionModeDepthReplacing:
2833 assert(b->shader->stage == MESA_SHADER_FRAGMENT);
2834 b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_ANY;
2835 break;
2836 case SpvExecutionModeDepthGreater:
2837 assert(b->shader->stage == MESA_SHADER_FRAGMENT);
2838 b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_GREATER;
2839 break;
2840 case SpvExecutionModeDepthLess:
2841 assert(b->shader->stage == MESA_SHADER_FRAGMENT);
2842 b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_LESS;
2843 break;
2844 case SpvExecutionModeDepthUnchanged:
2845 assert(b->shader->stage == MESA_SHADER_FRAGMENT);
2846 b->shader->info.fs.depth_layout = FRAG_DEPTH_LAYOUT_UNCHANGED;
2847 break;
2848
2849 case SpvExecutionModeLocalSize:
2850 assert(b->shader->stage == MESA_SHADER_COMPUTE);
2851 b->shader->info.cs.local_size[0] = mode->literals[0];
2852 b->shader->info.cs.local_size[1] = mode->literals[1];
2853 b->shader->info.cs.local_size[2] = mode->literals[2];
2854 break;
2855 case SpvExecutionModeLocalSizeHint:
2856 break; /* Nothing to do with this */
2857
2858 case SpvExecutionModeOutputVertices:
2859 if (b->shader->stage == MESA_SHADER_TESS_CTRL ||
2860 b->shader->stage == MESA_SHADER_TESS_EVAL) {
2861 b->shader->info.tess.tcs_vertices_out = mode->literals[0];
2862 } else {
2863 assert(b->shader->stage == MESA_SHADER_GEOMETRY);
2864 b->shader->info.gs.vertices_out = mode->literals[0];
2865 }
2866 break;
2867
2868 case SpvExecutionModeInputPoints:
2869 case SpvExecutionModeInputLines:
2870 case SpvExecutionModeInputLinesAdjacency:
2871 case SpvExecutionModeTriangles:
2872 case SpvExecutionModeInputTrianglesAdjacency:
2873 case SpvExecutionModeQuads:
2874 case SpvExecutionModeIsolines:
2875 if (b->shader->stage == MESA_SHADER_TESS_CTRL ||
2876 b->shader->stage == MESA_SHADER_TESS_EVAL) {
2877 b->shader->info.tess.primitive_mode =
2878 gl_primitive_from_spv_execution_mode(mode->exec_mode);
2879 } else {
2880 assert(b->shader->stage == MESA_SHADER_GEOMETRY);
2881 b->shader->info.gs.vertices_in =
2882 vertices_in_from_spv_execution_mode(mode->exec_mode);
2883 }
2884 break;
2885
2886 case SpvExecutionModeOutputPoints:
2887 case SpvExecutionModeOutputLineStrip:
2888 case SpvExecutionModeOutputTriangleStrip:
2889 assert(b->shader->stage == MESA_SHADER_GEOMETRY);
2890 b->shader->info.gs.output_primitive =
2891 gl_primitive_from_spv_execution_mode(mode->exec_mode);
2892 break;
2893
2894 case SpvExecutionModeSpacingEqual:
2895 assert(b->shader->stage == MESA_SHADER_TESS_CTRL ||
2896 b->shader->stage == MESA_SHADER_TESS_EVAL);
2897 b->shader->info.tess.spacing = TESS_SPACING_EQUAL;
2898 break;
2899 case SpvExecutionModeSpacingFractionalEven:
2900 assert(b->shader->stage == MESA_SHADER_TESS_CTRL ||
2901 b->shader->stage == MESA_SHADER_TESS_EVAL);
2902 b->shader->info.tess.spacing = TESS_SPACING_FRACTIONAL_EVEN;
2903 break;
2904 case SpvExecutionModeSpacingFractionalOdd:
2905 assert(b->shader->stage == MESA_SHADER_TESS_CTRL ||
2906 b->shader->stage == MESA_SHADER_TESS_EVAL);
2907 b->shader->info.tess.spacing = TESS_SPACING_FRACTIONAL_ODD;
2908 break;
2909 case SpvExecutionModeVertexOrderCw:
2910 assert(b->shader->stage == MESA_SHADER_TESS_CTRL ||
2911 b->shader->stage == MESA_SHADER_TESS_EVAL);
2912 /* Vulkan's notion of CCW seems to match the hardware backends,
2913 * but be the opposite of OpenGL. Currently NIR follows GL semantics,
2914 * so we set it backwards here.
2915 */
2916 b->shader->info.tess.ccw = true;
2917 break;
2918 case SpvExecutionModeVertexOrderCcw:
2919 assert(b->shader->stage == MESA_SHADER_TESS_CTRL ||
2920 b->shader->stage == MESA_SHADER_TESS_EVAL);
2921 /* Backwards; see above */
2922 b->shader->info.tess.ccw = false;
2923 break;
2924 case SpvExecutionModePointMode:
2925 assert(b->shader->stage == MESA_SHADER_TESS_CTRL ||
2926 b->shader->stage == MESA_SHADER_TESS_EVAL);
2927 b->shader->info.tess.point_mode = true;
2928 break;
2929
2930 case SpvExecutionModePixelCenterInteger:
2931 b->pixel_center_integer = true;
2932 break;
2933
2934 case SpvExecutionModeXfb:
2935 assert(!"Unhandled execution mode");
2936 break;
2937
2938 case SpvExecutionModeVecTypeHint:
2939 case SpvExecutionModeContractionOff:
2940 break; /* OpenCL */
2941
2942 default:
2943 unreachable("Unhandled execution mode");
2944 }
2945 }
2946
2947 static bool
2948 vtn_handle_variable_or_type_instruction(struct vtn_builder *b, SpvOp opcode,
2949 const uint32_t *w, unsigned count)
2950 {
2951 switch (opcode) {
2952 case SpvOpSource:
2953 case SpvOpSourceContinued:
2954 case SpvOpSourceExtension:
2955 case SpvOpExtension:
2956 case SpvOpCapability:
2957 case SpvOpExtInstImport:
2958 case SpvOpMemoryModel:
2959 case SpvOpEntryPoint:
2960 case SpvOpExecutionMode:
2961 case SpvOpString:
2962 case SpvOpName:
2963 case SpvOpMemberName:
2964 case SpvOpDecorationGroup:
2965 case SpvOpDecorate:
2966 case SpvOpMemberDecorate:
2967 case SpvOpGroupDecorate:
2968 case SpvOpGroupMemberDecorate:
2969 assert(!"Invalid opcode types and variables section");
2970 break;
2971
2972 case SpvOpTypeVoid:
2973 case SpvOpTypeBool:
2974 case SpvOpTypeInt:
2975 case SpvOpTypeFloat:
2976 case SpvOpTypeVector:
2977 case SpvOpTypeMatrix:
2978 case SpvOpTypeImage:
2979 case SpvOpTypeSampler:
2980 case SpvOpTypeSampledImage:
2981 case SpvOpTypeArray:
2982 case SpvOpTypeRuntimeArray:
2983 case SpvOpTypeStruct:
2984 case SpvOpTypeOpaque:
2985 case SpvOpTypePointer:
2986 case SpvOpTypeFunction:
2987 case SpvOpTypeEvent:
2988 case SpvOpTypeDeviceEvent:
2989 case SpvOpTypeReserveId:
2990 case SpvOpTypeQueue:
2991 case SpvOpTypePipe:
2992 vtn_handle_type(b, opcode, w, count);
2993 break;
2994
2995 case SpvOpConstantTrue:
2996 case SpvOpConstantFalse:
2997 case SpvOpConstant:
2998 case SpvOpConstantComposite:
2999 case SpvOpConstantSampler:
3000 case SpvOpConstantNull:
3001 case SpvOpSpecConstantTrue:
3002 case SpvOpSpecConstantFalse:
3003 case SpvOpSpecConstant:
3004 case SpvOpSpecConstantComposite:
3005 case SpvOpSpecConstantOp:
3006 vtn_handle_constant(b, opcode, w, count);
3007 break;
3008
3009 case SpvOpUndef:
3010 case SpvOpVariable:
3011 vtn_handle_variables(b, opcode, w, count);
3012 break;
3013
3014 default:
3015 return false; /* End of preamble */
3016 }
3017
3018 return true;
3019 }
3020
3021 static bool
3022 vtn_handle_body_instruction(struct vtn_builder *b, SpvOp opcode,
3023 const uint32_t *w, unsigned count)
3024 {
3025 switch (opcode) {
3026 case SpvOpLabel:
3027 break;
3028
3029 case SpvOpLoopMerge:
3030 case SpvOpSelectionMerge:
3031 /* This is handled by cfg pre-pass and walk_blocks */
3032 break;
3033
3034 case SpvOpUndef: {
3035 struct vtn_value *val = vtn_push_value(b, w[2], vtn_value_type_undef);
3036 val->type = vtn_value(b, w[1], vtn_value_type_type)->type;
3037 break;
3038 }
3039
3040 case SpvOpExtInst:
3041 vtn_handle_extension(b, opcode, w, count);
3042 break;
3043
3044 case SpvOpVariable:
3045 case SpvOpLoad:
3046 case SpvOpStore:
3047 case SpvOpCopyMemory:
3048 case SpvOpCopyMemorySized:
3049 case SpvOpAccessChain:
3050 case SpvOpInBoundsAccessChain:
3051 case SpvOpArrayLength:
3052 vtn_handle_variables(b, opcode, w, count);
3053 break;
3054
3055 case SpvOpFunctionCall:
3056 vtn_handle_function_call(b, opcode, w, count);
3057 break;
3058
3059 case SpvOpSampledImage:
3060 case SpvOpImage:
3061 case SpvOpImageSampleImplicitLod:
3062 case SpvOpImageSampleExplicitLod:
3063 case SpvOpImageSampleDrefImplicitLod:
3064 case SpvOpImageSampleDrefExplicitLod:
3065 case SpvOpImageSampleProjImplicitLod:
3066 case SpvOpImageSampleProjExplicitLod:
3067 case SpvOpImageSampleProjDrefImplicitLod:
3068 case SpvOpImageSampleProjDrefExplicitLod:
3069 case SpvOpImageFetch:
3070 case SpvOpImageGather:
3071 case SpvOpImageDrefGather:
3072 case SpvOpImageQuerySizeLod:
3073 case SpvOpImageQueryLod:
3074 case SpvOpImageQueryLevels:
3075 case SpvOpImageQuerySamples:
3076 vtn_handle_texture(b, opcode, w, count);
3077 break;
3078
3079 case SpvOpImageRead:
3080 case SpvOpImageWrite:
3081 case SpvOpImageTexelPointer:
3082 vtn_handle_image(b, opcode, w, count);
3083 break;
3084
3085 case SpvOpImageQuerySize: {
3086 struct vtn_pointer *image =
3087 vtn_value(b, w[3], vtn_value_type_pointer)->pointer;
3088 if (image->mode == vtn_variable_mode_image) {
3089 vtn_handle_image(b, opcode, w, count);
3090 } else {
3091 assert(image->mode == vtn_variable_mode_sampler);
3092 vtn_handle_texture(b, opcode, w, count);
3093 }
3094 break;
3095 }
3096
3097 case SpvOpAtomicLoad:
3098 case SpvOpAtomicExchange:
3099 case SpvOpAtomicCompareExchange:
3100 case SpvOpAtomicCompareExchangeWeak:
3101 case SpvOpAtomicIIncrement:
3102 case SpvOpAtomicIDecrement:
3103 case SpvOpAtomicIAdd:
3104 case SpvOpAtomicISub:
3105 case SpvOpAtomicSMin:
3106 case SpvOpAtomicUMin:
3107 case SpvOpAtomicSMax:
3108 case SpvOpAtomicUMax:
3109 case SpvOpAtomicAnd:
3110 case SpvOpAtomicOr:
3111 case SpvOpAtomicXor: {
3112 struct vtn_value *pointer = vtn_untyped_value(b, w[3]);
3113 if (pointer->value_type == vtn_value_type_image_pointer) {
3114 vtn_handle_image(b, opcode, w, count);
3115 } else {
3116 assert(pointer->value_type == vtn_value_type_pointer);
3117 vtn_handle_ssbo_or_shared_atomic(b, opcode, w, count);
3118 }
3119 break;
3120 }
3121
3122 case SpvOpAtomicStore: {
3123 struct vtn_value *pointer = vtn_untyped_value(b, w[1]);
3124 if (pointer->value_type == vtn_value_type_image_pointer) {
3125 vtn_handle_image(b, opcode, w, count);
3126 } else {
3127 assert(pointer->value_type == vtn_value_type_pointer);
3128 vtn_handle_ssbo_or_shared_atomic(b, opcode, w, count);
3129 }
3130 break;
3131 }
3132
3133 case SpvOpSNegate:
3134 case SpvOpFNegate:
3135 case SpvOpNot:
3136 case SpvOpAny:
3137 case SpvOpAll:
3138 case SpvOpConvertFToU:
3139 case SpvOpConvertFToS:
3140 case SpvOpConvertSToF:
3141 case SpvOpConvertUToF:
3142 case SpvOpUConvert:
3143 case SpvOpSConvert:
3144 case SpvOpFConvert:
3145 case SpvOpQuantizeToF16:
3146 case SpvOpConvertPtrToU:
3147 case SpvOpConvertUToPtr:
3148 case SpvOpPtrCastToGeneric:
3149 case SpvOpGenericCastToPtr:
3150 case SpvOpBitcast:
3151 case SpvOpIsNan:
3152 case SpvOpIsInf:
3153 case SpvOpIsFinite:
3154 case SpvOpIsNormal:
3155 case SpvOpSignBitSet:
3156 case SpvOpLessOrGreater:
3157 case SpvOpOrdered:
3158 case SpvOpUnordered:
3159 case SpvOpIAdd:
3160 case SpvOpFAdd:
3161 case SpvOpISub:
3162 case SpvOpFSub:
3163 case SpvOpIMul:
3164 case SpvOpFMul:
3165 case SpvOpUDiv:
3166 case SpvOpSDiv:
3167 case SpvOpFDiv:
3168 case SpvOpUMod:
3169 case SpvOpSRem:
3170 case SpvOpSMod:
3171 case SpvOpFRem:
3172 case SpvOpFMod:
3173 case SpvOpVectorTimesScalar:
3174 case SpvOpDot:
3175 case SpvOpIAddCarry:
3176 case SpvOpISubBorrow:
3177 case SpvOpUMulExtended:
3178 case SpvOpSMulExtended:
3179 case SpvOpShiftRightLogical:
3180 case SpvOpShiftRightArithmetic:
3181 case SpvOpShiftLeftLogical:
3182 case SpvOpLogicalEqual:
3183 case SpvOpLogicalNotEqual:
3184 case SpvOpLogicalOr:
3185 case SpvOpLogicalAnd:
3186 case SpvOpLogicalNot:
3187 case SpvOpBitwiseOr:
3188 case SpvOpBitwiseXor:
3189 case SpvOpBitwiseAnd:
3190 case SpvOpSelect:
3191 case SpvOpIEqual:
3192 case SpvOpFOrdEqual:
3193 case SpvOpFUnordEqual:
3194 case SpvOpINotEqual:
3195 case SpvOpFOrdNotEqual:
3196 case SpvOpFUnordNotEqual:
3197 case SpvOpULessThan:
3198 case SpvOpSLessThan:
3199 case SpvOpFOrdLessThan:
3200 case SpvOpFUnordLessThan:
3201 case SpvOpUGreaterThan:
3202 case SpvOpSGreaterThan:
3203 case SpvOpFOrdGreaterThan:
3204 case SpvOpFUnordGreaterThan:
3205 case SpvOpULessThanEqual:
3206 case SpvOpSLessThanEqual:
3207 case SpvOpFOrdLessThanEqual:
3208 case SpvOpFUnordLessThanEqual:
3209 case SpvOpUGreaterThanEqual:
3210 case SpvOpSGreaterThanEqual:
3211 case SpvOpFOrdGreaterThanEqual:
3212 case SpvOpFUnordGreaterThanEqual:
3213 case SpvOpDPdx:
3214 case SpvOpDPdy:
3215 case SpvOpFwidth:
3216 case SpvOpDPdxFine:
3217 case SpvOpDPdyFine:
3218 case SpvOpFwidthFine:
3219 case SpvOpDPdxCoarse:
3220 case SpvOpDPdyCoarse:
3221 case SpvOpFwidthCoarse:
3222 case SpvOpBitFieldInsert:
3223 case SpvOpBitFieldSExtract:
3224 case SpvOpBitFieldUExtract:
3225 case SpvOpBitReverse:
3226 case SpvOpBitCount:
3227 case SpvOpTranspose:
3228 case SpvOpOuterProduct:
3229 case SpvOpMatrixTimesScalar:
3230 case SpvOpVectorTimesMatrix:
3231 case SpvOpMatrixTimesVector:
3232 case SpvOpMatrixTimesMatrix:
3233 vtn_handle_alu(b, opcode, w, count);
3234 break;
3235
3236 case SpvOpVectorExtractDynamic:
3237 case SpvOpVectorInsertDynamic:
3238 case SpvOpVectorShuffle:
3239 case SpvOpCompositeConstruct:
3240 case SpvOpCompositeExtract:
3241 case SpvOpCompositeInsert:
3242 case SpvOpCopyObject:
3243 vtn_handle_composite(b, opcode, w, count);
3244 break;
3245
3246 case SpvOpEmitVertex:
3247 case SpvOpEndPrimitive:
3248 case SpvOpEmitStreamVertex:
3249 case SpvOpEndStreamPrimitive:
3250 case SpvOpControlBarrier:
3251 case SpvOpMemoryBarrier:
3252 vtn_handle_barrier(b, opcode, w, count);
3253 break;
3254
3255 default:
3256 unreachable("Unhandled opcode");
3257 }
3258
3259 return true;
3260 }
3261
3262 nir_function *
3263 spirv_to_nir(const uint32_t *words, size_t word_count,
3264 struct nir_spirv_specialization *spec, unsigned num_spec,
3265 gl_shader_stage stage, const char *entry_point_name,
3266 const struct nir_spirv_supported_extensions *ext,
3267 const nir_shader_compiler_options *options)
3268 {
3269 const uint32_t *word_end = words + word_count;
3270
3271 /* Handle the SPIR-V header (first 4 dwords) */
3272 assert(word_count > 5);
3273
3274 assert(words[0] == SpvMagicNumber);
3275 assert(words[1] >= 0x10000);
3276 /* words[2] == generator magic */
3277 unsigned value_id_bound = words[3];
3278 assert(words[4] == 0);
3279
3280 words+= 5;
3281
3282 /* Initialize the stn_builder object */
3283 struct vtn_builder *b = rzalloc(NULL, struct vtn_builder);
3284 b->value_id_bound = value_id_bound;
3285 b->values = rzalloc_array(b, struct vtn_value, value_id_bound);
3286 exec_list_make_empty(&b->functions);
3287 b->entry_point_stage = stage;
3288 b->entry_point_name = entry_point_name;
3289 b->ext = ext;
3290
3291 /* Handle all the preamble instructions */
3292 words = vtn_foreach_instruction(b, words, word_end,
3293 vtn_handle_preamble_instruction);
3294
3295 if (b->entry_point == NULL) {
3296 assert(!"Entry point not found");
3297 ralloc_free(b);
3298 return NULL;
3299 }
3300
3301 b->shader = nir_shader_create(NULL, stage, options, NULL);
3302
3303 /* Set shader info defaults */
3304 b->shader->info.gs.invocations = 1;
3305
3306 /* Parse execution modes */
3307 vtn_foreach_execution_mode(b, b->entry_point,
3308 vtn_handle_execution_mode, NULL);
3309
3310 b->specializations = spec;
3311 b->num_specializations = num_spec;
3312
3313 /* Handle all variable, type, and constant instructions */
3314 words = vtn_foreach_instruction(b, words, word_end,
3315 vtn_handle_variable_or_type_instruction);
3316
3317 vtn_build_cfg(b, words, word_end);
3318
3319 foreach_list_typed(struct vtn_function, func, node, &b->functions) {
3320 b->impl = func->impl;
3321 b->const_table = _mesa_hash_table_create(b, _mesa_hash_pointer,
3322 _mesa_key_pointer_equal);
3323
3324 vtn_function_emit(b, func, vtn_handle_body_instruction);
3325 }
3326
3327 assert(b->entry_point->value_type == vtn_value_type_function);
3328 nir_function *entry_point = b->entry_point->func->impl->function;
3329 assert(entry_point);
3330
3331 ralloc_free(b);
3332
3333 return entry_point;
3334 }