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 #ifndef ELK_FS_H
29 #define ELK_FS_H
30
31 #include "elk_shader.h"
32 #include "elk_ir_fs.h"
33 #include "elk_fs_live_variables.h"
34 #include "elk_ir_performance.h"
35 #include "compiler/nir/nir.h"
36
37 struct elk_bblock_t;
38 namespace {
39 struct acp_entry;
40 }
41
42 class elk_fs_visitor;
43
44 namespace elk {
45 /**
46 * Register pressure analysis of a shader. Estimates how many registers
47 * are live at any point of the program in GRF units.
48 */
49 struct register_pressure {
50 register_pressure(const elk_fs_visitor *v);
51 ~register_pressure();
52
53 analysis_dependency_class
dependency_classregister_pressure54 dependency_class() const
55 {
56 return (DEPENDENCY_INSTRUCTION_IDENTITY |
57 DEPENDENCY_INSTRUCTION_DATA_FLOW |
58 DEPENDENCY_VARIABLES);
59 }
60
61 bool
validateregister_pressure62 validate(const elk_fs_visitor *) const
63 {
64 /* FINISHME */
65 return true;
66 }
67
68 unsigned *regs_live_at_ip;
69 };
70 }
71
72 struct elk_gs_compile;
73
74 namespace elk {
75 class fs_builder;
76 }
77
78 struct shader_stats {
79 const char *scheduler_mode;
80 unsigned promoted_constants;
81 unsigned spill_count;
82 unsigned fill_count;
83 unsigned max_register_pressure;
84 };
85
86 /** Register numbers for thread payload fields. */
87 struct elk_elk_thread_payload {
88 /** The number of thread payload registers the hardware will supply. */
89 uint8_t num_regs;
90
91 virtual ~elk_elk_thread_payload() = default;
92
93 protected:
elk_elk_thread_payloadelk_elk_thread_payload94 elk_elk_thread_payload() : num_regs() {}
95 };
96
97 struct elk_vs_thread_payload : public elk_elk_thread_payload {
98 elk_vs_thread_payload(const elk_fs_visitor &v);
99
100 elk_fs_reg urb_handles;
101 };
102
103 struct elk_tcs_thread_payload : public elk_elk_thread_payload {
104 elk_tcs_thread_payload(const elk_fs_visitor &v);
105
106 elk_fs_reg patch_urb_output;
107 elk_fs_reg primitive_id;
108 elk_fs_reg icp_handle_start;
109 };
110
111 struct elk_tes_thread_payload : public elk_elk_thread_payload {
112 elk_tes_thread_payload(const elk_fs_visitor &v);
113
114 elk_fs_reg patch_urb_input;
115 elk_fs_reg primitive_id;
116 elk_fs_reg coords[3];
117 elk_fs_reg urb_output;
118 };
119
120 struct elk_gs_thread_payload : public elk_elk_thread_payload {
121 elk_gs_thread_payload(elk_fs_visitor &v);
122
123 elk_fs_reg urb_handles;
124 elk_fs_reg primitive_id;
125 elk_fs_reg instance_id;
126 elk_fs_reg icp_handle_start;
127 };
128
129 struct elk_fs_thread_payload : public elk_elk_thread_payload {
130 elk_fs_thread_payload(const elk_fs_visitor &v,
131 bool &source_depth_to_render_target,
132 bool &runtime_check_aads_emit);
133
134 uint8_t subspan_coord_reg[2];
135 uint8_t source_depth_reg[2];
136 uint8_t source_w_reg[2];
137 uint8_t aa_dest_stencil_reg[2];
138 uint8_t dest_depth_reg[2];
139 uint8_t sample_pos_reg[2];
140 uint8_t sample_mask_in_reg[2];
141 uint8_t depth_w_coef_reg;
142 uint8_t barycentric_coord_reg[ELK_BARYCENTRIC_MODE_COUNT][2];
143 };
144
145 struct elk_cs_thread_payload : public elk_elk_thread_payload {
146 elk_cs_thread_payload(const elk_fs_visitor &v);
147
148 void load_subgroup_id(const elk::fs_builder &bld, elk_fs_reg &dest) const;
149
150 elk_fs_reg local_invocation_id[3];
151
152 protected:
153 elk_fs_reg subgroup_id_;
154 };
155
156 class elk_fs_instruction_scheduler;
157
158 /**
159 * The fragment shader front-end.
160 *
161 * Translates either GLSL IR or Mesa IR (for ARB_fragment_program) into FS IR.
162 */
163 class elk_fs_visitor : public elk_backend_shader
164 {
165 public:
166 elk_fs_visitor(const struct elk_compiler *compiler,
167 const struct elk_compile_params *params,
168 const elk_base_prog_key *key,
169 struct elk_stage_prog_data *prog_data,
170 const nir_shader *shader,
171 unsigned dispatch_width,
172 bool needs_register_pressure,
173 bool debug_enabled);
174 elk_fs_visitor(const struct elk_compiler *compiler,
175 const struct elk_compile_params *params,
176 const elk_wm_prog_key *key,
177 struct elk_wm_prog_data *prog_data,
178 const nir_shader *shader,
179 unsigned dispatch_width,
180 unsigned num_polygons,
181 bool needs_register_pressure,
182 bool debug_enabled);
183 elk_fs_visitor(const struct elk_compiler *compiler,
184 const struct elk_compile_params *params,
185 struct elk_gs_compile *gs_compile,
186 struct elk_gs_prog_data *prog_data,
187 const nir_shader *shader,
188 bool needs_register_pressure,
189 bool debug_enabled);
190 void init();
191 ~elk_fs_visitor();
192
193 elk_fs_reg vgrf(const glsl_type *const type);
194 void import_uniforms(elk_fs_visitor *v);
195
196 void VARYING_PULL_CONSTANT_LOAD(const elk::fs_builder &bld,
197 const elk_fs_reg &dst,
198 const elk_fs_reg &surface,
199 const elk_fs_reg &surface_handle,
200 const elk_fs_reg &varying_offset,
201 uint32_t const_offset,
202 uint8_t alignment,
203 unsigned components);
204 void DEP_RESOLVE_MOV(const elk::fs_builder &bld, int grf);
205
206 bool run_fs(bool allow_spilling, bool do_rep_send);
207 bool run_vs();
208 bool run_tcs();
209 bool run_tes();
210 bool run_gs();
211 bool run_cs(bool allow_spilling);
212 void optimize();
213 void allocate_registers(bool allow_spilling);
214 uint32_t compute_max_register_pressure();
215 bool fixup_sends_duplicate_payload();
216 void fixup_3src_null_dest();
217 void emit_dummy_memory_fence_before_eot();
218 void emit_dummy_mov_instruction();
219 bool fixup_nomask_control_flow();
220 void assign_curb_setup();
221 void assign_urb_setup();
222 void convert_attr_sources_to_hw_regs(elk_fs_inst *inst);
223 void assign_vs_urb_setup();
224 void assign_tcs_urb_setup();
225 void assign_tes_urb_setup();
226 void assign_gs_urb_setup();
227 bool assign_regs(bool allow_spilling, bool spill_all);
228 void assign_regs_trivial();
229 void calculate_payload_ranges(unsigned payload_node_count,
230 int *payload_last_use_ip) const;
231 bool split_virtual_grfs();
232 bool compact_virtual_grfs();
233 void assign_constant_locations();
234 bool get_pull_locs(const elk_fs_reg &src, unsigned *out_surf_index,
235 unsigned *out_pull_index);
236 bool lower_constant_loads();
237 virtual void invalidate_analysis(elk::analysis_dependency_class c);
238
239 #ifndef NDEBUG
240 void validate();
241 #else
validate()242 void validate() {}
243 #endif
244
245 bool opt_algebraic();
246 bool opt_redundant_halt();
247 bool opt_cse();
248 bool opt_cse_local(const elk::fs_live_variables &live, elk_bblock_t *block, int &ip);
249
250 bool opt_copy_propagation();
251 bool opt_bank_conflicts();
252 bool opt_split_sends();
253 bool register_coalesce();
254 bool compute_to_mrf();
255 bool eliminate_find_live_channel();
256 bool dead_code_eliminate();
257 bool remove_duplicate_mrf_writes();
258 bool remove_extra_rounding_modes();
259
260 elk_fs_instruction_scheduler *prepare_scheduler(void *mem_ctx);
261 void schedule_instructions_pre_ra(elk_fs_instruction_scheduler *sched,
262 instruction_scheduler_mode mode);
263 void schedule_instructions_post_ra();
264
265 void insert_gfx4_send_dependency_workarounds();
266 void insert_gfx4_pre_send_dependency_workarounds(elk_bblock_t *block,
267 elk_fs_inst *inst);
268 void insert_gfx4_post_send_dependency_workarounds(elk_bblock_t *block,
269 elk_fs_inst *inst);
270 void vfail(const char *msg, va_list args);
271 void fail(const char *msg, ...);
272 void limit_dispatch_width(unsigned n, const char *msg);
273 bool lower_uniform_pull_constant_loads();
274 bool lower_load_payload();
275 bool lower_pack();
276 bool lower_regioning();
277 bool lower_logical_sends();
278 bool lower_integer_multiplication();
279 bool lower_minmax();
280 bool lower_simd_width();
281 bool lower_barycentrics();
282 bool lower_derivatives();
283 bool lower_find_live_channel();
284 bool lower_scoreboard();
285 bool lower_sub_sat();
286 bool opt_combine_constants();
287
288 void emit_repclear_shader();
289 void emit_interpolation_setup_gfx4();
290 void emit_interpolation_setup_gfx6();
291 bool opt_peephole_sel();
292 bool opt_saturate_propagation();
293 bool opt_cmod_propagation();
294 bool opt_zero_samples();
295
296 void set_tcs_invocation_id();
297
298 void emit_alpha_test();
299 elk_fs_inst *emit_single_fb_write(const elk::fs_builder &bld,
300 elk_fs_reg color1, elk_fs_reg color2,
301 elk_fs_reg src0_alpha, unsigned components);
302 void do_emit_fb_writes(int nr_color_regions, bool replicate_alpha);
303 void emit_fb_writes();
304 void emit_urb_writes(const elk_fs_reg &gs_vertex_count = elk_fs_reg());
305 void emit_gs_control_data_bits(const elk_fs_reg &vertex_count);
306 void emit_gs_thread_end();
307 bool mark_last_urb_write_with_eot();
308 void emit_tcs_thread_end();
309 void emit_urb_fence();
310 void emit_cs_terminate();
311
312 elk_fs_reg interp_reg(const elk::fs_builder &bld, unsigned location,
313 unsigned channel, unsigned comp);
314 elk_fs_reg per_primitive_reg(const elk::fs_builder &bld,
315 int location, unsigned comp);
316
317 virtual void dump_instruction_to_file(const elk_backend_instruction *inst, FILE *file) const;
318 virtual void dump_instructions_to_file(FILE *file) const;
319
320 const elk_base_prog_key *const key;
321 const struct elk_sampler_prog_key_data *key_tex;
322
323 struct elk_gs_compile *gs_compile;
324
325 struct elk_stage_prog_data *prog_data;
326
327 elk_analysis<elk::fs_live_variables, elk_backend_shader> live_analysis;
328 elk_analysis<elk::register_pressure, elk_fs_visitor> regpressure_analysis;
329 elk_analysis<elk::performance, elk_fs_visitor> performance_analysis;
330
331 /** Number of uniform variable components visited. */
332 unsigned uniforms;
333
334 /** Byte-offset for the next available spot in the scratch space buffer. */
335 unsigned last_scratch;
336
337 /**
338 * Array mapping UNIFORM register numbers to the push parameter index,
339 * or -1 if this uniform register isn't being uploaded as a push constant.
340 */
341 int *push_constant_loc;
342
343 elk_fs_reg frag_depth;
344 elk_fs_reg frag_stencil;
345 elk_fs_reg sample_mask;
346 elk_fs_reg outputs[VARYING_SLOT_MAX];
347 elk_fs_reg dual_src_output;
348 int first_non_payload_grf;
349 /** Either ELK_MAX_GRF or GFX7_MRF_HACK_START */
350 unsigned max_grf;
351
352 bool failed;
353 char *fail_msg;
354
355 elk_elk_thread_payload *payload_;
356
payload()357 elk_elk_thread_payload &payload() {
358 return *this->payload_;
359 }
360
vs_payload()361 elk_vs_thread_payload &vs_payload() {
362 assert(stage == MESA_SHADER_VERTEX);
363 return *static_cast<elk_vs_thread_payload *>(this->payload_);
364 }
365
tcs_payload()366 elk_tcs_thread_payload &tcs_payload() {
367 assert(stage == MESA_SHADER_TESS_CTRL);
368 return *static_cast<elk_tcs_thread_payload *>(this->payload_);
369 }
370
tes_payload()371 elk_tes_thread_payload &tes_payload() {
372 assert(stage == MESA_SHADER_TESS_EVAL);
373 return *static_cast<elk_tes_thread_payload *>(this->payload_);
374 }
375
gs_payload()376 elk_gs_thread_payload &gs_payload() {
377 assert(stage == MESA_SHADER_GEOMETRY);
378 return *static_cast<elk_gs_thread_payload *>(this->payload_);
379 }
380
fs_payload()381 elk_fs_thread_payload &fs_payload() {
382 assert(stage == MESA_SHADER_FRAGMENT);
383 return *static_cast<elk_fs_thread_payload *>(this->payload_);
384 };
385
cs_payload()386 elk_cs_thread_payload &cs_payload() {
387 assert(gl_shader_stage_uses_workgroup(stage));
388 return *static_cast<elk_cs_thread_payload *>(this->payload_);
389 }
390
391 bool source_depth_to_render_target;
392 bool runtime_check_aads_emit;
393
394 elk_fs_reg pixel_x;
395 elk_fs_reg pixel_y;
396 elk_fs_reg pixel_z;
397 elk_fs_reg wpos_w;
398 elk_fs_reg pixel_w;
399 elk_fs_reg delta_xy[ELK_BARYCENTRIC_MODE_COUNT];
400 elk_fs_reg final_gs_vertex_count;
401 elk_fs_reg control_data_bits;
402 elk_fs_reg invocation_id;
403
404 unsigned grf_used;
405 bool spilled_any_registers;
406 bool needs_register_pressure;
407
408 const unsigned dispatch_width; /**< 8, 16 or 32 */
409 const unsigned max_polygons;
410 unsigned max_dispatch_width;
411
412 /* The API selected subgroup size */
413 unsigned api_subgroup_size; /**< 0, 8, 16, 32 */
414
415 struct shader_stats shader_stats;
416
417 void lower_mul_dword_inst(elk_fs_inst *inst, elk_bblock_t *block);
418 void lower_mul_qword_inst(elk_fs_inst *inst, elk_bblock_t *block);
419 void lower_mulh_inst(elk_fs_inst *inst, elk_bblock_t *block);
420
421 unsigned workgroup_size() const;
422
423 void debug_optimizer(const nir_shader *nir,
424 const char *pass_name,
425 int iteration, int pass_num) const;
426 };
427
428 /**
429 * Return the flag register used in fragment shaders to keep track of live
430 * samples. On Gfx7+ we use f1.0-f1.1 to allow discard jumps in SIMD32
431 * dispatch mode, while earlier generations are constrained to f0.1, which
432 * limits the dispatch width to SIMD16 for fragment shaders that use discard.
433 */
434 static inline unsigned
sample_mask_flag_subreg(const elk_fs_visitor & s)435 sample_mask_flag_subreg(const elk_fs_visitor &s)
436 {
437 assert(s.stage == MESA_SHADER_FRAGMENT);
438 return s.devinfo->ver >= 7 ? 2 : 1;
439 }
440
441 /**
442 * The fragment shader code generator.
443 *
444 * Translates FS IR to actual i965 assembly code.
445 */
446 class elk_fs_generator
447 {
448 public:
449 elk_fs_generator(const struct elk_compiler *compiler,
450 const struct elk_compile_params *params,
451 struct elk_stage_prog_data *prog_data,
452 bool runtime_check_aads_emit,
453 gl_shader_stage stage);
454 ~elk_fs_generator();
455
456 void enable_debug(const char *shader_name);
457 int generate_code(const elk_cfg_t *cfg, int dispatch_width,
458 struct shader_stats shader_stats,
459 const elk::performance &perf,
460 struct elk_compile_stats *stats,
461 unsigned max_polygons = 0);
462 void add_const_data(void *data, unsigned size);
463 const unsigned *get_assembly();
464
465 private:
466 void fire_fb_write(elk_fs_inst *inst,
467 struct elk_reg payload,
468 struct elk_reg implied_header,
469 GLuint nr);
470 void generate_send(elk_fs_inst *inst,
471 struct elk_reg dst,
472 struct elk_reg desc,
473 struct elk_reg ex_desc,
474 struct elk_reg payload,
475 struct elk_reg payload2);
476 void generate_fb_write(elk_fs_inst *inst, struct elk_reg payload);
477 void generate_fb_read(elk_fs_inst *inst, struct elk_reg dst,
478 struct elk_reg payload);
479 void generate_cs_terminate(elk_fs_inst *inst, struct elk_reg payload);
480 void generate_barrier(elk_fs_inst *inst, struct elk_reg src);
481 bool generate_linterp(elk_fs_inst *inst, struct elk_reg dst,
482 struct elk_reg *src);
483 void generate_tex(elk_fs_inst *inst, struct elk_reg dst,
484 struct elk_reg surface_index,
485 struct elk_reg sampler_index);
486 void generate_ddx(const elk_fs_inst *inst,
487 struct elk_reg dst, struct elk_reg src);
488 void generate_ddy(const elk_fs_inst *inst,
489 struct elk_reg dst, struct elk_reg src);
490 void generate_scratch_write(elk_fs_inst *inst, struct elk_reg src);
491 void generate_scratch_read(elk_fs_inst *inst, struct elk_reg dst);
492 void generate_scratch_read_gfx7(elk_fs_inst *inst, struct elk_reg dst);
493 void generate_scratch_header(elk_fs_inst *inst, struct elk_reg dst);
494 void generate_uniform_pull_constant_load(elk_fs_inst *inst, struct elk_reg dst,
495 struct elk_reg index,
496 struct elk_reg offset);
497 void generate_varying_pull_constant_load_gfx4(elk_fs_inst *inst,
498 struct elk_reg dst,
499 struct elk_reg index);
500
501 void generate_set_sample_id(elk_fs_inst *inst,
502 struct elk_reg dst,
503 struct elk_reg src0,
504 struct elk_reg src1);
505
506 void generate_halt(elk_fs_inst *inst);
507
508 void generate_mov_indirect(elk_fs_inst *inst,
509 struct elk_reg dst,
510 struct elk_reg reg,
511 struct elk_reg indirect_byte_offset);
512
513 void generate_shuffle(elk_fs_inst *inst,
514 struct elk_reg dst,
515 struct elk_reg src,
516 struct elk_reg idx);
517
518 void generate_quad_swizzle(const elk_fs_inst *inst,
519 struct elk_reg dst, struct elk_reg src,
520 unsigned swiz);
521
522 bool patch_halt_jumps();
523
524 const struct elk_compiler *compiler;
525 const struct elk_compile_params *params;
526
527 const struct intel_device_info *devinfo;
528
529 struct elk_codegen *p;
530 struct elk_stage_prog_data * const prog_data;
531
532 unsigned dispatch_width; /**< 8, 16 or 32 */
533
534 exec_list discard_halt_patches;
535 bool runtime_check_aads_emit;
536 bool debug_flag;
537 const char *shader_name;
538 gl_shader_stage stage;
539 void *mem_ctx;
540 };
541
542 namespace elk {
543 elk_fs_reg
544 fetch_payload_reg(const elk::fs_builder &bld, uint8_t regs[2],
545 elk_reg_type type = ELK_REGISTER_TYPE_F,
546 unsigned n = 1);
547
548 elk_fs_reg
549 fetch_barycentric_reg(const elk::fs_builder &bld, uint8_t regs[2]);
550
551 inline elk_fs_reg
dynamic_msaa_flags(const struct elk_wm_prog_data * wm_prog_data)552 dynamic_msaa_flags(const struct elk_wm_prog_data *wm_prog_data)
553 {
554 return elk_fs_reg(UNIFORM, wm_prog_data->msaa_flags_param,
555 ELK_REGISTER_TYPE_UD);
556 }
557
558 void
559 check_dynamic_msaa_flag(const fs_builder &bld,
560 const struct elk_wm_prog_data *wm_prog_data,
561 enum intel_msaa_flags flag);
562
563 bool
564 lower_src_modifiers(elk_fs_visitor *v, elk_bblock_t *block, elk_fs_inst *inst, unsigned i);
565 }
566
567 void elk_shuffle_from_32bit_read(const elk::fs_builder &bld,
568 const elk_fs_reg &dst,
569 const elk_fs_reg &src,
570 uint32_t first_component,
571 uint32_t components);
572
573 elk_fs_reg elk_setup_imm_df(const elk::fs_builder &bld,
574 double v);
575
576 elk_fs_reg elk_setup_imm_b(const elk::fs_builder &bld,
577 int8_t v);
578
579 elk_fs_reg elk_setup_imm_ub(const elk::fs_builder &bld,
580 uint8_t v);
581
582 enum elk_barycentric_mode elk_barycentric_mode(nir_intrinsic_instr *intr);
583
584 uint32_t elk_fb_write_msg_control(const elk_fs_inst *inst,
585 const struct elk_wm_prog_data *prog_data);
586
587 void elk_compute_urb_setup_index(struct elk_wm_prog_data *wm_prog_data);
588
589 bool elk_nir_lower_simd(nir_shader *nir, unsigned dispatch_width);
590
591 elk_fs_reg elk_sample_mask_reg(const elk::fs_builder &bld);
592 void elk_emit_predicate_on_sample_mask(const elk::fs_builder &bld, elk_fs_inst *inst);
593
594 int elk_get_subgroup_id_param_index(const intel_device_info *devinfo,
595 const elk_stage_prog_data *prog_data);
596
597 bool elk_lower_dpas(elk_fs_visitor &v);
598
599 void nir_to_elk(elk_fs_visitor *s);
600
601 #endif /* ELK_FS_H */
602