xref: /aosp_15_r20/external/mesa3d/src/intel/vulkan/anv_internal_kernels.c (revision 6104692788411f58d303aa86923a9ff6ecaded22)
1 /*
2  * Copyright © 2022 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 
24 #include "anv_private.h"
25 
26 #include "compiler/intel_nir.h"
27 #include "compiler/brw_compiler.h"
28 #include "compiler/brw_nir.h"
29 #include "compiler/nir/nir.h"
30 #include "compiler/nir/nir_builder.h"
31 #include "dev/intel_debug.h"
32 #include "intel/compiler/intel_nir.h"
33 #include "util/macros.h"
34 
35 #include "vk_nir.h"
36 
37 #include "anv_internal_kernels.h"
38 
39 static bool
lower_base_workgroup_id(nir_builder * b,nir_intrinsic_instr * intrin,UNUSED void * data)40 lower_base_workgroup_id(nir_builder *b, nir_intrinsic_instr *intrin,
41                         UNUSED void *data)
42 {
43    if (intrin->intrinsic != nir_intrinsic_load_base_workgroup_id)
44       return false;
45 
46    b->cursor = nir_instr_remove(&intrin->instr);
47    nir_def_rewrite_uses(&intrin->def, nir_imm_zero(b, 3, 32));
48    return true;
49 }
50 
51 static void
link_libanv(nir_shader * nir,const nir_shader * libanv)52 link_libanv(nir_shader *nir, const nir_shader *libanv)
53 {
54    nir_link_shader_functions(nir, libanv);
55    NIR_PASS_V(nir, nir_inline_functions);
56    NIR_PASS_V(nir, nir_remove_non_entrypoints);
57    NIR_PASS_V(nir, nir_lower_vars_to_explicit_types, nir_var_function_temp,
58               glsl_get_cl_type_size_align);
59    NIR_PASS_V(nir, nir_opt_deref);
60    NIR_PASS_V(nir, nir_lower_vars_to_ssa);
61    NIR_PASS_V(nir, nir_lower_explicit_io,
62               nir_var_shader_temp | nir_var_function_temp | nir_var_mem_shared |
63                  nir_var_mem_global,
64               nir_address_format_62bit_generic);
65 }
66 
67 static struct anv_shader_bin *
compile_shader(struct anv_device * device,const nir_shader * libanv,enum anv_internal_kernel_name shader_name,gl_shader_stage stage,const char * name,const void * hash_key,uint32_t hash_key_size,uint32_t sends_count_expectation)68 compile_shader(struct anv_device *device,
69                const nir_shader *libanv,
70                enum anv_internal_kernel_name shader_name,
71                gl_shader_stage stage,
72                const char *name,
73                const void *hash_key,
74                uint32_t hash_key_size,
75                uint32_t sends_count_expectation)
76 {
77    const nir_shader_compiler_options *nir_options =
78       device->physical->compiler->nir_options[stage];
79 
80    nir_builder b = nir_builder_init_simple_shader(stage, nir_options,
81                                                   "%s", name);
82 
83    uint32_t uniform_size =
84       anv_genX(device->info, call_internal_shader)(&b, shader_name);
85 
86    nir_shader *nir = b.shader;
87 
88    link_libanv(nir, libanv);
89 
90    if (INTEL_DEBUG(DEBUG_SHADER_PRINT)) {
91       nir_lower_printf_options printf_opts = {
92          .ptr_bit_size               = 64,
93          .use_printf_base_identifier = true,
94       };
95       NIR_PASS_V(nir, nir_lower_printf, &printf_opts);
96    }
97 
98    NIR_PASS_V(nir, nir_lower_vars_to_ssa);
99    NIR_PASS_V(nir, nir_opt_cse);
100    NIR_PASS_V(nir, nir_opt_gcm, true);
101    NIR_PASS_V(nir, nir_opt_peephole_select, 1, false, false);
102 
103    NIR_PASS_V(nir, nir_lower_variable_initializers, ~0);
104 
105    NIR_PASS_V(nir, nir_split_var_copies);
106    NIR_PASS_V(nir, nir_split_per_member_structs);
107 
108    if (stage == MESA_SHADER_COMPUTE) {
109       nir->info.workgroup_size[0] = 16;
110       nir->info.workgroup_size[1] = 1;
111       nir->info.workgroup_size[2] = 1;
112    }
113 
114    struct brw_compiler *compiler = device->physical->compiler;
115    struct brw_nir_compiler_opts opts = {};
116    brw_preprocess_nir(compiler, nir, &opts);
117 
118    NIR_PASS_V(nir, nir_propagate_invariant, false);
119 
120    if (stage == MESA_SHADER_FRAGMENT) {
121       NIR_PASS_V(nir, nir_lower_input_attachments,
122                  &(nir_input_attachment_options) {
123                     .use_fragcoord_sysval = true,
124                     .use_layer_id_sysval = true,
125                  });
126    } else {
127       nir_lower_compute_system_values_options options = {
128          .has_base_workgroup_id = true,
129          .lower_cs_local_id_to_index = true,
130          .lower_workgroup_id_to_index = gl_shader_stage_is_mesh(stage),
131       };
132       NIR_PASS_V(nir, nir_lower_compute_system_values, &options);
133       NIR_PASS_V(nir, nir_shader_intrinsics_pass, lower_base_workgroup_id,
134                  nir_metadata_control_flow, NULL);
135    }
136 
137    /* Reset sizes before gathering information */
138    nir->global_mem_size = 0;
139    nir->scratch_size = 0;
140    nir->info.shared_size = 0;
141    nir_shader_gather_info(nir, nir_shader_get_entrypoint(nir));
142 
143    NIR_PASS_V(nir, nir_copy_prop);
144    NIR_PASS_V(nir, nir_opt_constant_folding);
145    NIR_PASS_V(nir, nir_opt_dce);
146 
147    union brw_any_prog_key key;
148    memset(&key, 0, sizeof(key));
149 
150    union brw_any_prog_data prog_data;
151    memset(&prog_data, 0, sizeof(prog_data));
152 
153    if (stage == MESA_SHADER_COMPUTE) {
154       NIR_PASS_V(nir, brw_nir_lower_cs_intrinsics,
155                  device->info, &prog_data.cs);
156    }
157 
158    /* Do vectorizing here. For some reason when trying to do it in the back
159     * this just isn't working.
160     */
161    nir_load_store_vectorize_options options = {
162       .modes = nir_var_mem_ubo | nir_var_mem_ssbo | nir_var_mem_global,
163       .callback = brw_nir_should_vectorize_mem,
164       .robust_modes = (nir_variable_mode)0,
165    };
166    NIR_PASS_V(nir, nir_opt_load_store_vectorize, &options);
167 
168    nir->num_uniforms = uniform_size;
169 
170    prog_data.base.nr_params = nir->num_uniforms / 4;
171 
172    brw_nir_analyze_ubo_ranges(compiler, nir, prog_data.base.ubo_ranges);
173 
174    void *temp_ctx = ralloc_context(NULL);
175 
176    const unsigned *program;
177    if (stage == MESA_SHADER_FRAGMENT) {
178       struct brw_compile_stats stats[3];
179       struct brw_compile_fs_params params = {
180          .base = {
181             .nir = nir,
182             .log_data = device,
183             .debug_flag = DEBUG_WM,
184             .stats = stats,
185             .mem_ctx = temp_ctx,
186          },
187          .key = &key.wm,
188          .prog_data = &prog_data.wm,
189       };
190       program = brw_compile_fs(compiler, &params);
191 
192       if (!INTEL_DEBUG(DEBUG_SHADER_PRINT)) {
193          unsigned stat_idx = 0;
194          if (prog_data.wm.dispatch_8) {
195             assert(stats[stat_idx].spills == 0);
196             assert(stats[stat_idx].fills == 0);
197             assert(stats[stat_idx].sends == sends_count_expectation);
198             stat_idx++;
199          }
200          if (prog_data.wm.dispatch_16) {
201             assert(stats[stat_idx].spills == 0);
202             assert(stats[stat_idx].fills == 0);
203             assert(stats[stat_idx].sends == sends_count_expectation);
204             stat_idx++;
205          }
206          if (prog_data.wm.dispatch_32) {
207             assert(stats[stat_idx].spills == 0);
208             assert(stats[stat_idx].fills == 0);
209             assert(stats[stat_idx].sends ==
210                    sends_count_expectation *
211                    (device->info->ver < 20 ? 2 : 1));
212             stat_idx++;
213          }
214       }
215    } else {
216       struct brw_compile_stats stats;
217       struct brw_compile_cs_params params = {
218          .base = {
219             .nir = nir,
220             .stats = &stats,
221             .log_data = device,
222             .debug_flag = DEBUG_CS,
223             .mem_ctx = temp_ctx,
224          },
225          .key = &key.cs,
226          .prog_data = &prog_data.cs,
227       };
228       program = brw_compile_cs(compiler, &params);
229 
230       if (!INTEL_DEBUG(DEBUG_SHADER_PRINT)) {
231          assert(stats.spills == 0);
232          assert(stats.fills == 0);
233          assert(stats.sends == sends_count_expectation);
234       }
235    }
236 
237    assert(prog_data.base.total_scratch == 0);
238    assert(program != NULL);
239    struct anv_shader_bin *kernel = NULL;
240    if (program == NULL)
241       goto exit;
242 
243    struct anv_pipeline_bind_map empty_bind_map = {};
244    struct anv_push_descriptor_info empty_push_desc_info = {};
245    struct anv_shader_upload_params upload_params = {
246       .stage               = nir->info.stage,
247       .key_data            = hash_key,
248       .key_size            = hash_key_size,
249       .kernel_data         = program,
250       .kernel_size         = prog_data.base.program_size,
251       .prog_data           = &prog_data.base,
252       .prog_data_size      = sizeof(prog_data),
253       .bind_map            = &empty_bind_map,
254       .push_desc_info      = &empty_push_desc_info,
255    };
256 
257    kernel = anv_device_upload_kernel(device, device->internal_cache, &upload_params);
258 
259 exit:
260    ralloc_free(temp_ctx);
261    ralloc_free(nir);
262 
263    return kernel;
264 }
265 
266 VkResult
anv_device_get_internal_shader(struct anv_device * device,enum anv_internal_kernel_name name,struct anv_shader_bin ** out_bin)267 anv_device_get_internal_shader(struct anv_device *device,
268                                enum anv_internal_kernel_name name,
269                                struct anv_shader_bin **out_bin)
270 {
271    const struct {
272       struct {
273          char name[40];
274       } key;
275 
276       gl_shader_stage stage;
277 
278       uint32_t        send_count;
279    } internal_kernels[] = {
280       [ANV_INTERNAL_KERNEL_GENERATED_DRAWS] = {
281          .key        = {
282             .name    = "anv-generated-indirect-draws",
283          },
284          .stage      = MESA_SHADER_FRAGMENT,
285          .send_count =  (device->info->ver == 9 ?
286                          /* 1 load +
287                           * 4 stores +
288                           * 2 * (2 loads + 2 stores) +
289                           * 3 stores
290                           */
291                          16 :
292                          /* 1 load +
293                           * 2 * (2 loads + 3 stores) +
294                           * 3 stores
295                           */
296                          14),
297       },
298       [ANV_INTERNAL_KERNEL_COPY_QUERY_RESULTS_COMPUTE] = {
299          .key        = {
300             .name    = "anv-copy-query-compute",
301          },
302          .stage      = MESA_SHADER_COMPUTE,
303          .send_count = device->info->verx10 >= 125 ?
304                        9 /* 4 loads + 4 stores + 1 EOT */ :
305                        8 /* 3 loads + 4 stores + 1 EOT */,
306       },
307       [ANV_INTERNAL_KERNEL_COPY_QUERY_RESULTS_FRAGMENT] = {
308          .key        = {
309             .name    = "anv-copy-query-fragment",
310          },
311          .stage      = MESA_SHADER_FRAGMENT,
312          .send_count = 8 /* 3 loads + 4 stores + 1 EOT */,
313       },
314       [ANV_INTERNAL_KERNEL_MEMCPY_COMPUTE] = {
315          .key        = {
316             .name    = "anv-memcpy-compute",
317          },
318          .stage      = MESA_SHADER_COMPUTE,
319          .send_count = device->info->verx10 >= 125 ?
320                        10 /* 5 loads (1 pull constants) + 4 stores + 1 EOT */ :
321                        9 /* 4 loads + 4 stores + 1 EOT */,
322       },
323    };
324 
325    struct anv_shader_bin *bin =
326       p_atomic_read(&device->internal_kernels[name]);
327    if (bin != NULL) {
328       *out_bin = bin;
329       return VK_SUCCESS;
330    }
331 
332    bin =
333       anv_device_search_for_kernel(device,
334                                    device->internal_cache,
335                                    &internal_kernels[name].key,
336                                    sizeof(internal_kernels[name].key),
337                                    NULL);
338    if (bin != NULL) {
339       p_atomic_set(&device->internal_kernels[name], bin);
340       *out_bin = bin;
341       return VK_SUCCESS;
342    }
343 
344    void *mem_ctx = ralloc_context(NULL);
345 
346    nir_shader *libanv_shaders =
347       anv_genX(device->info, load_libanv_shader)(device, mem_ctx);
348 
349    bin = compile_shader(device,
350                         libanv_shaders,
351                         name,
352                         internal_kernels[name].stage,
353                         internal_kernels[name].key.name,
354                         &internal_kernels[name].key,
355                         sizeof(internal_kernels[name].key),
356                         internal_kernels[name].send_count);
357    if (bin == NULL)
358       return vk_errorf(device, VK_ERROR_OUT_OF_HOST_MEMORY,
359                        "Unable to compiler internal kernel");
360 
361    /* The cache already has a reference and it's not going anywhere so
362     * there is no need to hold a second reference.
363     */
364    anv_shader_bin_unref(device, bin);
365 
366    p_atomic_set(&device->internal_kernels[name], bin);
367 
368    *out_bin = bin;
369    return VK_SUCCESS;
370 }
371 
372 VkResult
anv_device_init_internal_kernels(struct anv_device * device)373 anv_device_init_internal_kernels(struct anv_device *device)
374 {
375    const struct intel_l3_weights w =
376       intel_get_default_l3_weights(device->info,
377                                    true /* wants_dc_cache */,
378                                    false /* needs_slm */);
379    device->internal_kernels_l3_config = intel_get_l3_config(device->info, w);
380 
381    return VK_SUCCESS;
382 }
383 
384 void
anv_device_finish_internal_kernels(struct anv_device * device)385 anv_device_finish_internal_kernels(struct anv_device *device)
386 {
387 }
388