summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/iris/iris_draw.c
blob: 63693b8bd34945d715d76f2c3790d0765fb858e7 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
/*
 * Copyright © 2017 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included
 * in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

/**
 * @file iris_draw.c
 *
 * The main driver hooks for drawing and launching compute shaders.
 */

#include <stdio.h>
#include <errno.h>
#include "pipe/p_defines.h"
#include "pipe/p_state.h"
#include "pipe/p_context.h"
#include "pipe/p_screen.h"
#include "util/u_inlines.h"
#include "util/u_transfer.h"
#include "util/u_upload_mgr.h"
#include "intel/compiler/brw_compiler.h"
#include "intel/compiler/brw_eu_defines.h"
#include "iris_context.h"
#include "iris_defines.h"

static bool
prim_is_points_or_lines(const struct pipe_draw_info *draw)
{
   /* We don't need to worry about adjacency - it can only be used with
    * geometry shaders, and we don't care about this info when GS is on.
    */
   return draw->mode == PIPE_PRIM_POINTS ||
          draw->mode == PIPE_PRIM_LINES ||
          draw->mode == PIPE_PRIM_LINE_LOOP ||
          draw->mode == PIPE_PRIM_LINE_STRIP;
}

/**
 * Record the current primitive mode and restart information, flagging
 * related packets as dirty if necessary.
 *
 * This must be called before updating compiled shaders, because the patch
 * information informs the TCS key.
 */
static void
iris_update_draw_info(struct iris_context *ice,
                      const struct pipe_draw_info *info)
{
   if (ice->state.prim_mode != info->mode) {
      ice->state.prim_mode = info->mode;
      ice->state.dirty |= IRIS_DIRTY_VF_TOPOLOGY;


      /* For XY Clip enables */
      bool points_or_lines = prim_is_points_or_lines(info);
      if (points_or_lines != ice->state.prim_is_points_or_lines) {
         ice->state.prim_is_points_or_lines = points_or_lines;
         ice->state.dirty |= IRIS_DIRTY_CLIP;
      }
   }

   if (info->mode == PIPE_PRIM_PATCHES &&
       ice->state.vertices_per_patch != info->vertices_per_patch) {
      ice->state.vertices_per_patch = info->vertices_per_patch;
      ice->state.dirty |= IRIS_DIRTY_VF_TOPOLOGY;

      /* Flag constants dirty for gl_PatchVerticesIn if needed. */
      const struct shader_info *tcs_info =
         iris_get_shader_info(ice, MESA_SHADER_TESS_CTRL);
      if (tcs_info &&
          tcs_info->system_values_read & (1ull << SYSTEM_VALUE_VERTICES_IN)) {
         ice->state.dirty |= IRIS_DIRTY_CONSTANTS_TCS;
         ice->state.shaders[MESA_SHADER_TESS_CTRL].sysvals_need_upload = true;
      }
   }

   if (ice->state.primitive_restart != info->primitive_restart ||
       ice->state.cut_index != info->restart_index) {
      ice->state.dirty |= IRIS_DIRTY_VF;
      ice->state.primitive_restart = info->primitive_restart;
      ice->state.cut_index = info->restart_index;
   }
}

/**
 * Update shader draw parameters, flagging VF packets as dirty if necessary.
 */
static void
iris_update_draw_parameters(struct iris_context *ice,
                            const struct pipe_draw_info *info)
{
   if (info->indirect) {
      pipe_resource_reference(&ice->draw.draw_params_res,
                              info->indirect->buffer);
      ice->draw.draw_params_offset = info->indirect->offset +
                                     (info->index_size ? 12 : 8);
      ice->draw.params.firstvertex = 0;
      ice->draw.params.baseinstance = 0;
      ice->state.dirty |= IRIS_DIRTY_VERTEX_BUFFERS |
                          IRIS_DIRTY_VERTEX_ELEMENTS |
                          IRIS_DIRTY_VF_SGVS;
   } else if (ice->draw.is_indirect ||
              ice->draw.params.firstvertex !=
              (info->index_size ? info->index_bias : info->start) ||
              (ice->draw.params.baseinstance != info->start_instance)) {
      pipe_resource_reference(&ice->draw.draw_params_res, NULL);
      ice->draw.draw_params_offset = 0;
      ice->draw.params.firstvertex =
         info->index_size ? info->index_bias : info->start;
      ice->draw.params.baseinstance = info->start_instance;
      ice->state.dirty |= IRIS_DIRTY_VERTEX_BUFFERS |
                          IRIS_DIRTY_VERTEX_ELEMENTS |
                          IRIS_DIRTY_VF_SGVS;
   }
   ice->draw.is_indirect = info->indirect;

   if (ice->draw.derived_params.drawid != info->drawid ||
       ice->draw.derived_params.is_indexed_draw != (info->index_size ? ~0 : 0)) {
      ice->draw.derived_params.drawid = info->drawid;
      ice->draw.derived_params.is_indexed_draw = info->index_size ? ~0 : 0;
      ice->state.dirty |= IRIS_DIRTY_VERTEX_BUFFERS |
                          IRIS_DIRTY_VERTEX_ELEMENTS |
                          IRIS_DIRTY_VF_SGVS;
   }
}

static void
iris_indirect_draw_vbo(struct iris_context *ice,
                       const struct pipe_draw_info *dinfo)
{
   struct iris_batch *batch = &ice->batches[IRIS_BATCH_RENDER];
   struct pipe_draw_info info = *dinfo;

   if (info.indirect->indirect_draw_count &&
       ice->state.predicate == IRIS_PREDICATE_STATE_USE_BIT) {
      /* Upload MI_PREDICATE_RESULT to GPR2.*/
      ice->vtbl.load_register_reg64(batch, CS_GPR(2), MI_PREDICATE_RESULT);
   }

   uint64_t orig_dirty = ice->state.dirty;

   for (int i = 0; i < info.indirect->draw_count; i++) {
      info.drawid = i;

      iris_batch_maybe_flush(batch, 1500);

      iris_update_draw_parameters(ice, &info);

      ice->vtbl.upload_render_state(ice, batch, &info);

      ice->state.dirty &= ~IRIS_ALL_DIRTY_FOR_RENDER;

      info.indirect->offset += info.indirect->stride;
   }

   if (info.indirect->indirect_draw_count &&
       ice->state.predicate == IRIS_PREDICATE_STATE_USE_BIT) {
      /* Restore MI_PREDICATE_RESULT. */
      ice->vtbl.load_register_reg64(batch, MI_PREDICATE_RESULT, CS_GPR(2));
   }

   /* Put this back for post-draw resolves, we'll clear it again after. */
   ice->state.dirty = orig_dirty;
}

static void
iris_simple_draw_vbo(struct iris_context *ice,
                     const struct pipe_draw_info *draw)
{
   struct iris_batch *batch = &ice->batches[IRIS_BATCH_RENDER];

   iris_batch_maybe_flush(batch, 1500);

   iris_update_draw_parameters(ice, draw);

   ice->vtbl.upload_render_state(ice, batch, draw);
}

/**
 * The pipe->draw_vbo() driver hook.  Performs a draw on the GPU.
 */
void
iris_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen*)ice->ctx.screen;
   const struct gen_device_info *devinfo = &screen->devinfo;
   struct iris_batch *batch = &ice->batches[IRIS_BATCH_RENDER];

   if (ice->state.predicate == IRIS_PREDICATE_STATE_DONT_RENDER)
      return;

   /* We can't safely re-emit 3DSTATE_SO_BUFFERS because it may zero the
    * write offsets, changing the behavior.
    */
   if (unlikely(INTEL_DEBUG & DEBUG_REEMIT))
      ice->state.dirty |= IRIS_ALL_DIRTY_FOR_RENDER & ~IRIS_DIRTY_SO_BUFFERS;

   iris_update_draw_info(ice, info);

   if (devinfo->gen == 9)
      gen9_toggle_preemption(ice, batch, info);

   iris_update_compiled_shaders(ice);

   if (ice->state.dirty & IRIS_DIRTY_RENDER_RESOLVES_AND_FLUSHES) {
      bool draw_aux_buffer_disabled[BRW_MAX_DRAW_BUFFERS] = { };
      for (gl_shader_stage stage = 0; stage < MESA_SHADER_COMPUTE; stage++) {
         if (ice->shaders.prog[stage])
            iris_predraw_resolve_inputs(ice, batch, draw_aux_buffer_disabled,
                                        stage, true);
      }
      iris_predraw_resolve_framebuffer(ice, batch, draw_aux_buffer_disabled);
   }

   iris_binder_reserve_3d(ice);

   ice->vtbl.update_surface_base_address(batch, &ice->state.binder);

   if (info->indirect)
      iris_indirect_draw_vbo(ice, info);
   else
      iris_simple_draw_vbo(ice, info);

   iris_postdraw_update_resolve_tracking(ice, batch);

   ice->state.dirty &= ~IRIS_ALL_DIRTY_FOR_RENDER;
}

static void
iris_update_grid_size_resource(struct iris_context *ice,
                               const struct pipe_grid_info *grid)
{
   const struct iris_screen *screen = (void *) ice->ctx.screen;
   const struct isl_device *isl_dev = &screen->isl_dev;
   struct iris_state_ref *grid_ref = &ice->state.grid_size;
   struct iris_state_ref *state_ref = &ice->state.grid_surf_state;

   const struct iris_compiled_shader *shader = ice->shaders.prog[MESA_SHADER_COMPUTE];
   bool grid_needs_surface = shader->bt.used_mask[IRIS_SURFACE_GROUP_CS_WORK_GROUPS];
   bool grid_updated = false;

   if (grid->indirect) {
      pipe_resource_reference(&grid_ref->res, grid->indirect);
      grid_ref->offset = grid->indirect_offset;

      /* Zero out the grid size so that the next non-indirect grid launch will
       * re-upload it properly.
       */
      memset(ice->state.last_grid, 0, sizeof(ice->state.last_grid));
      grid_updated = true;
   } else if (memcmp(ice->state.last_grid, grid->grid, sizeof(grid->grid)) != 0) {
      memcpy(ice->state.last_grid, grid->grid, sizeof(grid->grid));
      u_upload_data(ice->state.dynamic_uploader, 0, sizeof(grid->grid), 4,
                    grid->grid, &grid_ref->offset, &grid_ref->res);
      grid_updated = true;
   }

   /* If we changed the grid, the old surface state is invalid. */
   if (grid_updated)
      pipe_resource_reference(&state_ref->res, NULL);

   /* Skip surface upload if we don't need it or we already have one */
   if (!grid_needs_surface || state_ref->res)
      return;

   void *surf_map = NULL;
   u_upload_alloc(ice->state.surface_uploader, 0, isl_dev->ss.size,
                  isl_dev->ss.align, &state_ref->offset, &state_ref->res,
                  &surf_map);
   state_ref->offset +=
      iris_bo_offset_from_base_address(iris_resource_bo(state_ref->res));
   isl_buffer_fill_state(&screen->isl_dev, surf_map,
                         .address = grid_ref->offset +
                            iris_resource_bo(grid_ref->res)->gtt_offset,
                         .size_B = sizeof(grid->grid),
                         .format = ISL_FORMAT_RAW,
                         .stride_B = 1,
                         .mocs = 4); // XXX: MOCS

   ice->state.dirty |= IRIS_DIRTY_BINDINGS_CS;
}

void
iris_launch_grid(struct pipe_context *ctx, const struct pipe_grid_info *grid)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_batch *batch = &ice->batches[IRIS_BATCH_COMPUTE];

   if (ice->state.predicate == IRIS_PREDICATE_STATE_DONT_RENDER)
      return;

   if (unlikely(INTEL_DEBUG & DEBUG_REEMIT))
      ice->state.dirty |= IRIS_ALL_DIRTY_FOR_COMPUTE;

   /* We can't do resolves on the compute engine, so awkwardly, we have to
    * do them on the render batch...
    */
   if (ice->state.dirty & IRIS_DIRTY_COMPUTE_RESOLVES_AND_FLUSHES) {
      iris_predraw_resolve_inputs(ice, &ice->batches[IRIS_BATCH_RENDER], NULL,
                                  MESA_SHADER_COMPUTE, false);
   }

   iris_batch_maybe_flush(batch, 1500);

   if (ice->state.dirty & IRIS_DIRTY_UNCOMPILED_CS)
      iris_update_compiled_compute_shader(ice);

   iris_update_grid_size_resource(ice, grid);

   iris_binder_reserve_compute(ice);
   ice->vtbl.update_surface_base_address(batch, &ice->state.binder);

   if (ice->state.compute_predicate) {
      ice->vtbl.load_register_mem64(batch, MI_PREDICATE_RESULT,
                                    ice->state.compute_predicate, 0);
      ice->state.compute_predicate = NULL;
   }

   ice->vtbl.upload_compute_state(ice, batch, grid);

   ice->state.dirty &= ~IRIS_ALL_DIRTY_FOR_COMPUTE;

   /* Note: since compute shaders can't access the framebuffer, there's
    * no need to call iris_postdraw_update_resolve_tracking.
    */
}