0a637cd3b3733edcdb525564f64d1e5f2746dafc
[mesa.git] / src / mesa / drivers / dri / i965 / brw_fs.h
1 /*
2 * Copyright © 2010 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 * Eric Anholt <eric@anholt.net>
25 *
26 */
27
28 #pragma once
29
30 #include "brw_shader.h"
31 #include "brw_ir_fs.h"
32
33 extern "C" {
34
35 #include <sys/types.h>
36
37 #include "main/macros.h"
38 #include "main/shaderobj.h"
39 #include "main/uniforms.h"
40 #include "program/prog_parameter.h"
41 #include "program/prog_print.h"
42 #include "program/prog_optimize.h"
43 #include "util/register_allocate.h"
44 #include "program/hash_table.h"
45 #include "brw_context.h"
46 #include "brw_eu.h"
47 #include "brw_wm.h"
48 #include "intel_asm_annotation.h"
49 }
50 #include "glsl/glsl_types.h"
51 #include "glsl/ir.h"
52 #include "glsl/nir/nir.h"
53 #include "program/sampler.h"
54
55 struct bblock_t;
56 namespace {
57 struct acp_entry;
58 }
59
60 namespace brw {
61 class fs_live_variables;
62 }
63
64 /**
65 * The fragment shader front-end.
66 *
67 * Translates either GLSL IR or Mesa IR (for ARB_fragment_program) into FS IR.
68 */
69 class fs_visitor : public backend_visitor
70 {
71 public:
72 const fs_reg reg_null_f;
73 const fs_reg reg_null_d;
74 const fs_reg reg_null_ud;
75
76 fs_visitor(struct brw_context *brw,
77 void *mem_ctx,
78 const struct brw_wm_prog_key *key,
79 struct brw_wm_prog_data *prog_data,
80 struct gl_shader_program *shader_prog,
81 struct gl_fragment_program *fp,
82 unsigned dispatch_width);
83
84 fs_visitor(struct brw_context *brw,
85 void *mem_ctx,
86 const struct brw_vs_prog_key *key,
87 struct brw_vs_prog_data *prog_data,
88 struct gl_shader_program *shader_prog,
89 struct gl_vertex_program *cp,
90 unsigned dispatch_width);
91
92 fs_visitor(struct brw_context *brw,
93 void *mem_ctx,
94 const struct brw_cs_prog_key *key,
95 struct brw_cs_prog_data *prog_data,
96 struct gl_shader_program *shader_prog,
97 struct gl_compute_program *cp,
98 unsigned dispatch_width);
99
100 ~fs_visitor();
101 void init();
102
103 fs_reg *variable_storage(ir_variable *var);
104 fs_reg vgrf(const glsl_type *const type);
105 fs_reg vgrf(int num_components);
106 void import_uniforms(fs_visitor *v);
107 void setup_uniform_clipplane_values();
108 void compute_clip_distance();
109
110 void visit(ir_variable *ir);
111 void visit(ir_assignment *ir);
112 void visit(ir_dereference_variable *ir);
113 void visit(ir_dereference_record *ir);
114 void visit(ir_dereference_array *ir);
115 void visit(ir_expression *ir);
116 void visit(ir_texture *ir);
117 void visit(ir_if *ir);
118 void visit(ir_constant *ir);
119 void visit(ir_swizzle *ir);
120 void visit(ir_return *ir);
121 void visit(ir_loop *ir);
122 void visit(ir_loop_jump *ir);
123 void visit(ir_discard *ir);
124 void visit(ir_call *ir);
125 void visit(ir_function *ir);
126 void visit(ir_function_signature *ir);
127 void visit(ir_emit_vertex *);
128 void visit(ir_end_primitive *);
129
130 uint32_t gather_channel(int orig_chan, uint32_t sampler);
131 void swizzle_result(ir_texture_opcode op, int dest_components,
132 fs_reg orig_val, uint32_t sampler);
133
134 fs_inst *emit(fs_inst *inst);
135 void emit(exec_list list);
136
137 fs_inst *emit(enum opcode opcode);
138 fs_inst *emit(enum opcode opcode, const fs_reg &dst);
139 fs_inst *emit(enum opcode opcode, const fs_reg &dst, const fs_reg &src0);
140 fs_inst *emit(enum opcode opcode, const fs_reg &dst, const fs_reg &src0,
141 const fs_reg &src1);
142 fs_inst *emit(enum opcode opcode, const fs_reg &dst,
143 const fs_reg &src0, const fs_reg &src1, const fs_reg &src2);
144 fs_inst *emit(enum opcode opcode, const fs_reg &dst,
145 fs_reg src[], int sources);
146
147 fs_inst *MOV(const fs_reg &dst, const fs_reg &src);
148 fs_inst *NOT(const fs_reg &dst, const fs_reg &src);
149 fs_inst *RNDD(const fs_reg &dst, const fs_reg &src);
150 fs_inst *RNDE(const fs_reg &dst, const fs_reg &src);
151 fs_inst *RNDZ(const fs_reg &dst, const fs_reg &src);
152 fs_inst *FRC(const fs_reg &dst, const fs_reg &src);
153 fs_inst *ADD(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
154 fs_inst *MUL(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
155 fs_inst *MACH(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
156 fs_inst *MAC(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
157 fs_inst *SHL(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
158 fs_inst *SHR(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
159 fs_inst *ASR(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
160 fs_inst *AND(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
161 fs_inst *OR(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
162 fs_inst *XOR(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
163 fs_inst *IF(enum brw_predicate predicate);
164 fs_inst *IF(const fs_reg &src0, const fs_reg &src1,
165 enum brw_conditional_mod condition);
166 fs_inst *CMP(fs_reg dst, fs_reg src0, fs_reg src1,
167 enum brw_conditional_mod condition);
168 fs_inst *LRP(const fs_reg &dst, const fs_reg &a, const fs_reg &y,
169 const fs_reg &x);
170 fs_inst *DEP_RESOLVE_MOV(int grf);
171 fs_inst *BFREV(const fs_reg &dst, const fs_reg &value);
172 fs_inst *BFE(const fs_reg &dst, const fs_reg &bits, const fs_reg &offset,
173 const fs_reg &value);
174 fs_inst *BFI1(const fs_reg &dst, const fs_reg &bits, const fs_reg &offset);
175 fs_inst *BFI2(const fs_reg &dst, const fs_reg &bfi1_dst,
176 const fs_reg &insert, const fs_reg &base);
177 fs_inst *FBH(const fs_reg &dst, const fs_reg &value);
178 fs_inst *FBL(const fs_reg &dst, const fs_reg &value);
179 fs_inst *CBIT(const fs_reg &dst, const fs_reg &value);
180 fs_inst *MAD(const fs_reg &dst, const fs_reg &c, const fs_reg &b,
181 const fs_reg &a);
182 fs_inst *ADDC(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
183 fs_inst *SUBB(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
184 fs_inst *SEL(const fs_reg &dst, const fs_reg &src0, const fs_reg &src1);
185
186 int type_size(const struct glsl_type *type);
187 fs_inst *get_instruction_generating_reg(fs_inst *start,
188 fs_inst *end,
189 const fs_reg &reg);
190
191 fs_inst *LOAD_PAYLOAD(const fs_reg &dst, fs_reg *src, int sources);
192
193 exec_list VARYING_PULL_CONSTANT_LOAD(const fs_reg &dst,
194 const fs_reg &surf_index,
195 const fs_reg &varying_offset,
196 uint32_t const_offset);
197
198 bool run_fs();
199 bool run_vs();
200 bool run_cs();
201 void optimize();
202 void allocate_registers();
203 void assign_binding_table_offsets();
204 void setup_payload_gen4();
205 void setup_payload_gen6();
206 void setup_vs_payload();
207 void setup_cs_payload();
208 void fixup_3src_null_dest();
209 void assign_curb_setup();
210 void calculate_urb_setup();
211 void assign_urb_setup();
212 void assign_vs_urb_setup();
213 bool assign_regs(bool allow_spilling);
214 void assign_regs_trivial();
215 void get_used_mrfs(bool *mrf_used);
216 void setup_payload_interference(struct ra_graph *g, int payload_reg_count,
217 int first_payload_node);
218 void setup_mrf_hack_interference(struct ra_graph *g,
219 int first_mrf_hack_node);
220 int choose_spill_reg(struct ra_graph *g);
221 void spill_reg(int spill_reg);
222 void split_virtual_grfs();
223 bool compact_virtual_grfs();
224 void move_uniform_array_access_to_pull_constants();
225 void assign_constant_locations();
226 void demote_pull_constants();
227 void invalidate_live_intervals();
228 void calculate_live_intervals();
229 void calculate_register_pressure();
230 bool opt_algebraic();
231 bool opt_redundant_discard_jumps();
232 bool opt_cse();
233 bool opt_cse_local(bblock_t *block);
234 bool opt_copy_propagate();
235 bool try_copy_propagate(fs_inst *inst, int arg, acp_entry *entry);
236 bool try_constant_propagate(fs_inst *inst, acp_entry *entry);
237 bool opt_copy_propagate_local(void *mem_ctx, bblock_t *block,
238 exec_list *acp);
239 bool opt_register_renaming();
240 bool register_coalesce();
241 bool compute_to_mrf();
242 bool dead_code_eliminate();
243 bool remove_duplicate_mrf_writes();
244
245 bool opt_sampler_eot();
246 bool virtual_grf_interferes(int a, int b);
247 void schedule_instructions(instruction_scheduler_mode mode);
248 void insert_gen4_send_dependency_workarounds();
249 void insert_gen4_pre_send_dependency_workarounds(bblock_t *block,
250 fs_inst *inst);
251 void insert_gen4_post_send_dependency_workarounds(bblock_t *block,
252 fs_inst *inst);
253 void vfail(const char *msg, va_list args);
254 void fail(const char *msg, ...);
255 void no16(const char *msg, ...);
256 void lower_uniform_pull_constant_loads();
257 bool lower_load_payload();
258 bool opt_combine_constants();
259
260 void emit_dummy_fs();
261 void emit_repclear_shader();
262 fs_reg *emit_fragcoord_interpolation(bool pixel_center_integer,
263 bool origin_upper_left);
264 fs_inst *emit_linterp(const fs_reg &attr, const fs_reg &interp,
265 glsl_interp_qualifier interpolation_mode,
266 bool is_centroid, bool is_sample);
267 fs_reg *emit_frontfacing_interpolation();
268 fs_reg *emit_samplepos_setup();
269 fs_reg *emit_sampleid_setup();
270 void emit_general_interpolation(fs_reg attr, const char *name,
271 const glsl_type *type,
272 glsl_interp_qualifier interpolation_mode,
273 int location, bool mod_centroid,
274 bool mod_sample);
275 fs_reg *emit_vs_system_value(int location);
276 void emit_interpolation_setup_gen4();
277 void emit_interpolation_setup_gen6();
278 void compute_sample_position(fs_reg dst, fs_reg int_sample_pos);
279 fs_reg rescale_texcoord(fs_reg coordinate, int coord_components,
280 bool is_rect, uint32_t sampler, int texunit);
281 fs_inst *emit_texture_gen4(ir_texture_opcode op, fs_reg dst,
282 fs_reg coordinate, int coord_components,
283 fs_reg shadow_comp,
284 fs_reg lod, fs_reg lod2, int grad_components,
285 uint32_t sampler);
286 fs_inst *emit_texture_gen4_simd16(ir_texture_opcode op, fs_reg dst,
287 fs_reg coordinate, int vector_elements,
288 fs_reg shadow_c, fs_reg lod,
289 uint32_t sampler);
290 fs_inst *emit_texture_gen5(ir_texture_opcode op, fs_reg dst,
291 fs_reg coordinate, int coord_components,
292 fs_reg shadow_comp,
293 fs_reg lod, fs_reg lod2, int grad_components,
294 fs_reg sample_index, uint32_t sampler,
295 bool has_offset);
296 fs_inst *emit_texture_gen7(ir_texture_opcode op, fs_reg dst,
297 fs_reg coordinate, int coord_components,
298 fs_reg shadow_comp,
299 fs_reg lod, fs_reg lod2, int grad_components,
300 fs_reg sample_index, fs_reg mcs, fs_reg sampler,
301 fs_reg offset_value);
302 void emit_texture(ir_texture_opcode op,
303 const glsl_type *dest_type,
304 fs_reg coordinate, int components,
305 fs_reg shadow_c,
306 fs_reg lod, fs_reg dpdy, int grad_components,
307 fs_reg sample_index,
308 fs_reg offset,
309 fs_reg mcs,
310 int gather_component,
311 bool is_cube_array,
312 bool is_rect,
313 uint32_t sampler,
314 fs_reg sampler_reg,
315 int texunit);
316 fs_reg emit_mcs_fetch(fs_reg coordinate, int components, fs_reg sampler);
317 void emit_gen6_gather_wa(uint8_t wa, fs_reg dst);
318 void resolve_source_modifiers(fs_reg *src);
319 fs_reg fix_math_operand(fs_reg src);
320 fs_inst *emit_math(enum opcode op, fs_reg dst, fs_reg src0);
321 fs_inst *emit_math(enum opcode op, fs_reg dst, fs_reg src0, fs_reg src1);
322 fs_inst *emit_lrp(const fs_reg &dst, const fs_reg &x, const fs_reg &y,
323 const fs_reg &a);
324 void emit_minmax(enum brw_conditional_mod conditionalmod, const fs_reg &dst,
325 const fs_reg &src0, const fs_reg &src1);
326 void emit_discard_jump();
327 bool try_emit_b2f_of_comparison(ir_expression *ir);
328 bool try_emit_saturate(ir_expression *ir);
329 bool try_emit_line(ir_expression *ir);
330 bool try_emit_mad(ir_expression *ir);
331 bool try_replace_with_sel();
332 bool try_opt_frontfacing_ternary(ir_if *ir);
333 bool opt_peephole_sel();
334 bool opt_peephole_predicated_break();
335 bool opt_saturate_propagation();
336 bool opt_cmod_propagation();
337 bool opt_zero_samples();
338 void emit_bool_to_cond_code(ir_rvalue *condition);
339 void emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]);
340 void emit_if_gen6(ir_if *ir);
341 void emit_unspill(bblock_t *block, fs_inst *inst, fs_reg reg,
342 uint32_t spill_offset, int count);
343 void emit_spill(bblock_t *block, fs_inst *inst, fs_reg reg,
344 uint32_t spill_offset, int count);
345
346 void emit_fragment_program_code();
347 void setup_fp_regs();
348 fs_reg get_fp_src_reg(const prog_src_register *src);
349 fs_reg get_fp_dst_reg(const prog_dst_register *dst);
350 void emit_fp_alu1(enum opcode opcode,
351 const struct prog_instruction *fpi,
352 fs_reg dst, fs_reg src);
353 void emit_fp_alu2(enum opcode opcode,
354 const struct prog_instruction *fpi,
355 fs_reg dst, fs_reg src0, fs_reg src1);
356 void emit_fp_scalar_write(const struct prog_instruction *fpi,
357 fs_reg dst, fs_reg src);
358 void emit_fp_scalar_math(enum opcode opcode,
359 const struct prog_instruction *fpi,
360 fs_reg dst, fs_reg src);
361
362 void emit_fp_minmax(const struct prog_instruction *fpi,
363 fs_reg dst, fs_reg src0, fs_reg src1);
364
365 void emit_fp_sop(enum brw_conditional_mod conditional_mod,
366 const struct prog_instruction *fpi,
367 fs_reg dst, fs_reg src0, fs_reg src1, fs_reg one);
368
369 void emit_nir_code();
370 void nir_setup_inputs(nir_shader *shader);
371 void nir_setup_outputs(nir_shader *shader);
372 void nir_setup_uniforms(nir_shader *shader);
373 void nir_setup_uniform(nir_variable *var);
374 void nir_setup_builtin_uniform(nir_variable *var);
375 void nir_emit_system_values(nir_shader *shader);
376 void nir_emit_impl(nir_function_impl *impl);
377 void nir_emit_cf_list(exec_list *list);
378 void nir_emit_if(nir_if *if_stmt);
379 void nir_emit_loop(nir_loop *loop);
380 void nir_emit_block(nir_block *block);
381 void nir_emit_instr(nir_instr *instr);
382 void nir_emit_alu(nir_alu_instr *instr);
383 void nir_emit_intrinsic(nir_intrinsic_instr *instr);
384 void nir_emit_texture(nir_tex_instr *instr);
385 void nir_emit_jump(nir_jump_instr *instr);
386 fs_reg get_nir_src(nir_src src);
387 fs_reg get_nir_dest(nir_dest dest);
388 void emit_percomp(fs_inst *inst, unsigned wr_mask);
389
390 bool optimize_frontfacing_ternary(nir_alu_instr *instr,
391 const fs_reg &result);
392
393 int setup_color_payload(fs_reg *dst, fs_reg color, unsigned components,
394 bool use_2nd_half);
395 void emit_alpha_test();
396 fs_inst *emit_single_fb_write(fs_reg color1, fs_reg color2,
397 fs_reg src0_alpha, unsigned components,
398 bool use_2nd_half = false);
399 void emit_fb_writes();
400 void emit_urb_writes();
401 void emit_cs_terminate();
402
403 void emit_shader_time_begin();
404 void emit_shader_time_end();
405 fs_inst *SHADER_TIME_ADD(enum shader_time_shader_type type, fs_reg value);
406
407 void emit_untyped_atomic(unsigned atomic_op, unsigned surf_index,
408 fs_reg dst, fs_reg offset, fs_reg src0,
409 fs_reg src1);
410
411 void emit_untyped_surface_read(unsigned surf_index, fs_reg dst,
412 fs_reg offset);
413
414 void emit_interpolate_expression(ir_expression *ir);
415
416 bool try_rewrite_rhs_to_dst(ir_assignment *ir,
417 fs_reg dst,
418 fs_reg src,
419 fs_inst *pre_rhs_inst,
420 fs_inst *last_rhs_inst);
421 void emit_assignment_writes(fs_reg &l, fs_reg &r,
422 const glsl_type *type, bool predicated);
423 void resolve_ud_negate(fs_reg *reg);
424 void resolve_bool_comparison(ir_rvalue *rvalue, fs_reg *reg);
425
426 fs_reg get_timestamp(fs_inst **out_mov);
427
428 struct brw_reg interp_reg(int location, int channel);
429 void setup_uniform_values(ir_variable *ir);
430 void setup_builtin_uniform_values(ir_variable *ir);
431 int implied_mrf_writes(fs_inst *inst);
432
433 virtual void dump_instructions();
434 virtual void dump_instructions(const char *name);
435 void dump_instruction(backend_instruction *inst);
436 void dump_instruction(backend_instruction *inst, FILE *file);
437
438 void visit_atomic_counter_intrinsic(ir_call *ir);
439
440 const void *const key;
441 const struct brw_sampler_prog_key_data *key_tex;
442
443 struct brw_stage_prog_data *prog_data;
444 unsigned int sanity_param_count;
445
446 int *param_size;
447
448 int *virtual_grf_start;
449 int *virtual_grf_end;
450 brw::fs_live_variables *live_intervals;
451
452 int *regs_live_at_ip;
453
454 /** Number of uniform variable components visited. */
455 unsigned uniforms;
456
457 /** Total number of direct uniforms we can get from NIR */
458 unsigned num_direct_uniforms;
459
460 /** Byte-offset for the next available spot in the scratch space buffer. */
461 unsigned last_scratch;
462
463 /**
464 * Array mapping UNIFORM register numbers to the pull parameter index,
465 * or -1 if this uniform register isn't being uploaded as a pull constant.
466 */
467 int *pull_constant_loc;
468
469 /**
470 * Array mapping UNIFORM register numbers to the push parameter index,
471 * or -1 if this uniform register isn't being uploaded as a push constant.
472 */
473 int *push_constant_loc;
474
475 struct hash_table *variable_ht;
476 fs_reg frag_depth;
477 fs_reg sample_mask;
478 fs_reg outputs[VARYING_SLOT_MAX];
479 unsigned output_components[VARYING_SLOT_MAX];
480 fs_reg dual_src_output;
481 bool do_dual_src;
482 int first_non_payload_grf;
483 /** Either BRW_MAX_GRF or GEN7_MRF_HACK_START */
484 unsigned max_grf;
485
486 fs_reg *fp_temp_regs;
487 fs_reg *fp_input_regs;
488
489 fs_reg *nir_locals;
490 fs_reg *nir_globals;
491 fs_reg nir_inputs;
492 fs_reg nir_outputs;
493 fs_reg *nir_system_values;
494
495 /** @{ debug annotation info */
496 const char *current_annotation;
497 const void *base_ir;
498 /** @} */
499
500 bool failed;
501 char *fail_msg;
502 bool simd16_unsupported;
503 char *no16_msg;
504
505 /* Result of last visit() method. */
506 fs_reg result;
507
508 /** Register numbers for thread payload fields. */
509 struct {
510 uint8_t source_depth_reg;
511 uint8_t source_w_reg;
512 uint8_t aa_dest_stencil_reg;
513 uint8_t dest_depth_reg;
514 uint8_t sample_pos_reg;
515 uint8_t sample_mask_in_reg;
516 uint8_t barycentric_coord_reg[BRW_WM_BARYCENTRIC_INTERP_MODE_COUNT];
517
518 /** The number of thread payload registers the hardware will supply. */
519 uint8_t num_regs;
520 } payload;
521
522 bool source_depth_to_render_target;
523 bool runtime_check_aads_emit;
524
525 fs_reg pixel_x;
526 fs_reg pixel_y;
527 fs_reg wpos_w;
528 fs_reg pixel_w;
529 fs_reg delta_xy[BRW_WM_BARYCENTRIC_INTERP_MODE_COUNT];
530 fs_reg shader_start_time;
531 fs_reg userplane[MAX_CLIP_PLANES];
532
533 unsigned grf_used;
534 bool spilled_any_registers;
535
536 const unsigned dispatch_width; /**< 8 or 16 */
537
538 unsigned promoted_constants;
539 };
540
541 /**
542 * The fragment shader code generator.
543 *
544 * Translates FS IR to actual i965 assembly code.
545 */
546 class fs_generator
547 {
548 public:
549 fs_generator(struct brw_context *brw,
550 void *mem_ctx,
551 const void *key,
552 struct brw_stage_prog_data *prog_data,
553 struct gl_program *fp,
554 unsigned promoted_constants,
555 bool runtime_check_aads_emit,
556 const char *stage_abbrev);
557 ~fs_generator();
558
559 void enable_debug(const char *shader_name);
560 int generate_code(const cfg_t *cfg, int dispatch_width);
561 const unsigned *get_assembly(unsigned int *assembly_size);
562
563 private:
564 void fire_fb_write(fs_inst *inst,
565 struct brw_reg payload,
566 struct brw_reg implied_header,
567 GLuint nr);
568 void generate_fb_write(fs_inst *inst, struct brw_reg payload);
569 void generate_urb_write(fs_inst *inst, struct brw_reg payload);
570 void generate_cs_terminate(fs_inst *inst, struct brw_reg payload);
571 void generate_blorp_fb_write(fs_inst *inst);
572 void generate_linterp(fs_inst *inst, struct brw_reg dst,
573 struct brw_reg *src);
574 void generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src,
575 struct brw_reg sampler_index);
576 void generate_math_gen6(fs_inst *inst,
577 struct brw_reg dst,
578 struct brw_reg src0,
579 struct brw_reg src1);
580 void generate_math_gen4(fs_inst *inst,
581 struct brw_reg dst,
582 struct brw_reg src);
583 void generate_math_g45(fs_inst *inst,
584 struct brw_reg dst,
585 struct brw_reg src);
586 void generate_ddx(enum opcode op, struct brw_reg dst, struct brw_reg src);
587 void generate_ddy(enum opcode op, struct brw_reg dst, struct brw_reg src,
588 bool negate_value);
589 void generate_scratch_write(fs_inst *inst, struct brw_reg src);
590 void generate_scratch_read(fs_inst *inst, struct brw_reg dst);
591 void generate_scratch_read_gen7(fs_inst *inst, struct brw_reg dst);
592 void generate_uniform_pull_constant_load(fs_inst *inst, struct brw_reg dst,
593 struct brw_reg index,
594 struct brw_reg offset);
595 void generate_uniform_pull_constant_load_gen7(fs_inst *inst,
596 struct brw_reg dst,
597 struct brw_reg surf_index,
598 struct brw_reg offset);
599 void generate_varying_pull_constant_load(fs_inst *inst, struct brw_reg dst,
600 struct brw_reg index,
601 struct brw_reg offset);
602 void generate_varying_pull_constant_load_gen7(fs_inst *inst,
603 struct brw_reg dst,
604 struct brw_reg index,
605 struct brw_reg offset);
606 void generate_mov_dispatch_to_flags(fs_inst *inst);
607
608 void generate_pixel_interpolator_query(fs_inst *inst,
609 struct brw_reg dst,
610 struct brw_reg src,
611 struct brw_reg msg_data,
612 unsigned msg_type);
613
614 void generate_set_omask(fs_inst *inst,
615 struct brw_reg dst,
616 struct brw_reg sample_mask);
617
618 void generate_set_sample_id(fs_inst *inst,
619 struct brw_reg dst,
620 struct brw_reg src0,
621 struct brw_reg src1);
622
623 void generate_set_simd4x2_offset(fs_inst *inst,
624 struct brw_reg dst,
625 struct brw_reg offset);
626 void generate_discard_jump(fs_inst *inst);
627
628 void generate_pack_half_2x16_split(fs_inst *inst,
629 struct brw_reg dst,
630 struct brw_reg x,
631 struct brw_reg y);
632 void generate_unpack_half_2x16_split(fs_inst *inst,
633 struct brw_reg dst,
634 struct brw_reg src);
635
636 void generate_shader_time_add(fs_inst *inst,
637 struct brw_reg payload,
638 struct brw_reg offset,
639 struct brw_reg value);
640
641 void generate_untyped_atomic(fs_inst *inst,
642 struct brw_reg dst,
643 struct brw_reg payload,
644 struct brw_reg atomic_op,
645 struct brw_reg surf_index);
646
647 void generate_untyped_surface_read(fs_inst *inst,
648 struct brw_reg dst,
649 struct brw_reg payload,
650 struct brw_reg surf_index);
651
652 bool patch_discard_jumps_to_fb_writes();
653
654 struct brw_context *brw;
655 const struct brw_device_info *devinfo;
656
657 struct brw_codegen *p;
658 const void * const key;
659 struct brw_stage_prog_data * const prog_data;
660
661 const struct gl_program *prog;
662
663 unsigned dispatch_width; /**< 8 or 16 */
664
665 exec_list discard_halt_patches;
666 unsigned promoted_constants;
667 bool runtime_check_aads_emit;
668 bool debug_flag;
669 const char *shader_name;
670 const char *stage_abbrev;
671 void *mem_ctx;
672 };
673
674 bool brw_do_channel_expressions(struct exec_list *instructions);
675 bool brw_do_vector_splitting(struct exec_list *instructions);
676 void brw_setup_tex_for_precompile(struct brw_context *brw,
677 struct brw_sampler_prog_key_data *tex,
678 struct gl_program *prog);