intel/isl: Ignore base_array_layer and array_len for 3D storage surfaces
[mesa.git] / src / intel / isl / isl_surface_state.c
1 /*
2 * Copyright 2016 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 <stdint.h>
25
26 #define __gen_address_type uint64_t
27 #define __gen_user_data void
28
29 static inline uint64_t
30 __gen_combine_address(void *data, void *loc, uint64_t addr, uint32_t delta)
31 {
32 return addr + delta;
33 }
34
35 #include "genxml/gen_macros.h"
36 #include "genxml/genX_pack.h"
37
38 #include "isl_priv.h"
39
40 #define __PASTE2(x, y) x ## y
41 #define __PASTE(x, y) __PASTE2(x, y)
42 #define isl_genX(x) __PASTE(isl_, genX(x))
43
44 #if GEN_GEN >= 8
45 static const uint8_t isl_to_gen_halign[] = {
46 [4] = HALIGN4,
47 [8] = HALIGN8,
48 [16] = HALIGN16,
49 };
50 #elif GEN_GEN >= 7
51 static const uint8_t isl_to_gen_halign[] = {
52 [4] = HALIGN_4,
53 [8] = HALIGN_8,
54 };
55 #endif
56
57 #if GEN_GEN >= 8
58 static const uint8_t isl_to_gen_valign[] = {
59 [4] = VALIGN4,
60 [8] = VALIGN8,
61 [16] = VALIGN16,
62 };
63 #elif GEN_GEN >= 6
64 static const uint8_t isl_to_gen_valign[] = {
65 [2] = VALIGN_2,
66 [4] = VALIGN_4,
67 };
68 #endif
69
70 #if GEN_GEN >= 8
71 static const uint8_t isl_to_gen_tiling[] = {
72 [ISL_TILING_LINEAR] = LINEAR,
73 [ISL_TILING_X] = XMAJOR,
74 [ISL_TILING_Y0] = YMAJOR,
75 [ISL_TILING_Yf] = YMAJOR,
76 [ISL_TILING_Ys] = YMAJOR,
77 [ISL_TILING_W] = WMAJOR,
78 };
79 #endif
80
81 #if GEN_GEN >= 7
82 static const uint32_t isl_to_gen_multisample_layout[] = {
83 [ISL_MSAA_LAYOUT_NONE] = MSFMT_MSS,
84 [ISL_MSAA_LAYOUT_INTERLEAVED] = MSFMT_DEPTH_STENCIL,
85 [ISL_MSAA_LAYOUT_ARRAY] = MSFMT_MSS,
86 };
87 #endif
88
89 #if GEN_GEN >= 9
90 static const uint32_t isl_to_gen_aux_mode[] = {
91 [ISL_AUX_USAGE_NONE] = AUX_NONE,
92 [ISL_AUX_USAGE_HIZ] = AUX_HIZ,
93 [ISL_AUX_USAGE_MCS] = AUX_CCS_D,
94 [ISL_AUX_USAGE_CCS_D] = AUX_CCS_D,
95 [ISL_AUX_USAGE_CCS_E] = AUX_CCS_E,
96 };
97 #elif GEN_GEN >= 8
98 static const uint32_t isl_to_gen_aux_mode[] = {
99 [ISL_AUX_USAGE_NONE] = AUX_NONE,
100 [ISL_AUX_USAGE_HIZ] = AUX_HIZ,
101 [ISL_AUX_USAGE_MCS] = AUX_MCS,
102 [ISL_AUX_USAGE_CCS_D] = AUX_MCS,
103 };
104 #endif
105
106 static uint8_t
107 get_surftype(enum isl_surf_dim dim, isl_surf_usage_flags_t usage)
108 {
109 switch (dim) {
110 default:
111 unreachable("bad isl_surf_dim");
112 case ISL_SURF_DIM_1D:
113 assert(!(usage & ISL_SURF_USAGE_CUBE_BIT));
114 return SURFTYPE_1D;
115 case ISL_SURF_DIM_2D:
116 if (usage & ISL_SURF_USAGE_STORAGE_BIT) {
117 /* Storage images are always plain 2-D, not cube */
118 return SURFTYPE_2D;
119 } else if (usage & ISL_SURF_USAGE_CUBE_BIT) {
120 return SURFTYPE_CUBE;
121 } else {
122 return SURFTYPE_2D;
123 }
124 case ISL_SURF_DIM_3D:
125 assert(!(usage & ISL_SURF_USAGE_CUBE_BIT));
126 return SURFTYPE_3D;
127 }
128 }
129
130 /**
131 * Get the horizontal and vertical alignment in the units expected by the
132 * hardware. Note that this does NOT give you the actual hardware enum values
133 * but an index into the isl_to_gen_[hv]align arrays above.
134 */
135 static inline struct isl_extent3d
136 get_image_alignment(const struct isl_surf *surf)
137 {
138 if (GEN_GEN >= 9) {
139 if (isl_tiling_is_std_y(surf->tiling) ||
140 surf->dim_layout == ISL_DIM_LAYOUT_GEN9_1D) {
141 /* The hardware ignores the alignment values. Anyway, the surface's
142 * true alignment is likely outside the enum range of HALIGN* and
143 * VALIGN*.
144 */
145 return isl_extent3d(4, 4, 1);
146 } else {
147 /* In Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in units
148 * of surface elements (not pixels nor samples). For compressed formats,
149 * a "surface element" is defined as a compression block. For example,
150 * if SurfaceVerticalAlignment is VALIGN_4 and SurfaceFormat is an ETC2
151 * format (ETC2 has a block height of 4), then the vertical alignment is
152 * 4 compression blocks or, equivalently, 16 pixels.
153 */
154 return isl_surf_get_image_alignment_el(surf);
155 }
156 } else {
157 /* Pre-Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in
158 * units of surface samples. For example, if SurfaceVerticalAlignment
159 * is VALIGN_4 and the surface is singlesampled, then for any surface
160 * format (compressed or not) the vertical alignment is
161 * 4 pixels.
162 */
163 return isl_surf_get_image_alignment_sa(surf);
164 }
165 }
166
167 #if GEN_GEN >= 8
168 static uint32_t
169 get_qpitch(const struct isl_surf *surf)
170 {
171 switch (surf->dim_layout) {
172 default:
173 unreachable("Bad isl_surf_dim");
174 case ISL_DIM_LAYOUT_GEN4_2D:
175 case ISL_DIM_LAYOUT_GEN4_3D:
176 if (GEN_GEN >= 9) {
177 return isl_surf_get_array_pitch_el_rows(surf);
178 } else {
179 /* From the Broadwell PRM for RENDER_SURFACE_STATE.QPitch
180 *
181 * "This field must be set to an integer multiple of the Surface
182 * Vertical Alignment. For compressed textures (BC*, FXT1,
183 * ETC*, and EAC* Surface Formats), this field is in units of
184 * rows in the uncompressed surface, and must be set to an
185 * integer multiple of the vertical alignment parameter "j"
186 * defined in the Common Surface Formats section."
187 */
188 return isl_surf_get_array_pitch_sa_rows(surf);
189 }
190 case ISL_DIM_LAYOUT_GEN9_1D:
191 /* QPitch is usually expressed as rows of surface elements (where
192 * a surface element is an compression block or a single surface
193 * sample). Skylake 1D is an outlier.
194 *
195 * From the Skylake BSpec >> Memory Views >> Common Surface
196 * Formats >> Surface Layout and Tiling >> 1D Surfaces:
197 *
198 * Surface QPitch specifies the distance in pixels between array
199 * slices.
200 */
201 return isl_surf_get_array_pitch_el(surf);
202 }
203 }
204 #endif /* GEN_GEN >= 8 */
205
206 void
207 isl_genX(surf_fill_state_s)(const struct isl_device *dev, void *state,
208 const struct isl_surf_fill_state_info *restrict info)
209 {
210 struct GENX(RENDER_SURFACE_STATE) s = { 0 };
211
212 s.SurfaceType = get_surftype(info->surf->dim, info->view->usage);
213
214 if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT)
215 assert(isl_format_supports_rendering(dev->info, info->view->format));
216 else if (info->view->usage & ISL_SURF_USAGE_TEXTURE_BIT)
217 assert(isl_format_supports_sampling(dev->info, info->view->format));
218 s.SurfaceFormat = info->view->format;
219
220 #if GEN_IS_HASWELL
221 s.IntegerSurfaceFormat = isl_format_has_int_channel(s.SurfaceFormat);
222 #endif
223
224 s.Width = info->surf->logical_level0_px.width - 1;
225 s.Height = info->surf->logical_level0_px.height - 1;
226
227 /* In the gen6 PRM Volume 1 Part 1: Graphics Core, Section 7.18.3.7.1
228 * (Surface Arrays For all surfaces other than separate stencil buffer):
229 *
230 * "[DevSNB] Errata: Sampler MSAA Qpitch will be 4 greater than the value
231 * calculated in the equation above , for every other odd Surface Height
232 * starting from 1 i.e. 1,5,9,13"
233 *
234 * Since this Qpitch errata only impacts the sampler, we have to adjust the
235 * input for the rendering surface to achieve the same qpitch. For the
236 * affected heights, we increment the height by 1 for the rendering
237 * surface.
238 */
239 if (GEN_GEN == 6 && (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) &&
240 info->surf->samples > 1 &&
241 (info->surf->logical_level0_px.height % 4) == 1)
242 s.Height++;
243
244 switch (s.SurfaceType) {
245 case SURFTYPE_1D:
246 case SURFTYPE_2D:
247 /* From the Ivy Bridge PRM >> RENDER_SURFACE_STATE::MinimumArrayElement:
248 *
249 * "If Number of Multisamples is not MULTISAMPLECOUNT_1, this field
250 * must be set to zero if this surface is used with sampling engine
251 * messages."
252 *
253 * This restriction appears to exist only on Ivy Bridge.
254 */
255 if (GEN_GEN == 7 && !GEN_IS_HASWELL && !ISL_DEV_IS_BAYTRAIL(dev) &&
256 (info->view->usage & ISL_SURF_USAGE_TEXTURE_BIT) &&
257 info->surf->samples > 1)
258 assert(info->view->base_array_layer == 0);
259
260 s.MinimumArrayElement = info->view->base_array_layer;
261
262 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
263 *
264 * For SURFTYPE_1D, 2D, and CUBE: The range of this field is reduced
265 * by one for each increase from zero of Minimum Array Element. For
266 * example, if Minimum Array Element is set to 1024 on a 2D surface,
267 * the range of this field is reduced to [0,1023].
268 *
269 * In other words, 'Depth' is the number of array layers.
270 */
271 s.Depth = info->view->array_len - 1;
272
273 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
274 *
275 * For Render Target and Typed Dataport 1D and 2D Surfaces:
276 * This field must be set to the same value as the Depth field.
277 */
278 if (info->view->usage & (ISL_SURF_USAGE_RENDER_TARGET_BIT |
279 ISL_SURF_USAGE_STORAGE_BIT))
280 s.RenderTargetViewExtent = s.Depth;
281 break;
282 case SURFTYPE_CUBE:
283 s.MinimumArrayElement = info->view->base_array_layer;
284 /* Same as SURFTYPE_2D, but divided by 6 */
285 s.Depth = info->view->array_len / 6 - 1;
286 if (info->view->usage & (ISL_SURF_USAGE_RENDER_TARGET_BIT |
287 ISL_SURF_USAGE_STORAGE_BIT))
288 s.RenderTargetViewExtent = s.Depth;
289 break;
290 case SURFTYPE_3D:
291 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
292 *
293 * If the volume texture is MIP-mapped, this field specifies the
294 * depth of the base MIP level.
295 */
296 s.Depth = info->surf->logical_level0_px.depth - 1;
297
298 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
299 *
300 * For Render Target and Typed Dataport 3D Surfaces: This field
301 * indicates the extent of the accessible 'R' coordinates minus 1 on
302 * the LOD currently being rendered to.
303 *
304 * The docs specify that this only matters for render targets and
305 * surfaces used with typed dataport messages. Prior to Ivy Bridge, the
306 * Depth field has more bits than RenderTargetViewExtent so we can have
307 * textures with more levels than we can render to. In order to prevent
308 * assert-failures in the packing function below, we only set the field
309 * when it's actually going to be used by the hardware.
310 *
311 * Similaraly, the MinimumArrayElement field is ignored by all hardware
312 * prior to Sky Lake when texturing and we want it set to 0 anyway.
313 * Since it's already initialized to 0, we can just leave it alone for
314 * texture surfaces.
315 */
316 if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) {
317 s.MinimumArrayElement = info->view->base_array_layer;
318 s.RenderTargetViewExtent = info->view->array_len - 1;
319 } else if (info->view->usage & ISL_SURF_USAGE_STORAGE_BIT) {
320 s.MinimumArrayElement = 0;
321 s.RenderTargetViewExtent =
322 isl_minify(info->surf->logical_level0_px.depth,
323 info->view->base_level) - 1;
324 }
325 break;
326 default:
327 unreachable("bad SurfaceType");
328 }
329
330 #if GEN_GEN >= 7
331 s.SurfaceArray = info->surf->dim != ISL_SURF_DIM_3D;
332 #endif
333
334 if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) {
335 /* For render target surfaces, the hardware interprets field
336 * MIPCount/LOD as LOD. The Broadwell PRM says:
337 *
338 * MIPCountLOD defines the LOD that will be rendered into.
339 * SurfaceMinLOD is ignored.
340 */
341 s.MIPCountLOD = info->view->base_level;
342 s.SurfaceMinLOD = 0;
343 } else {
344 /* For non render target surfaces, the hardware interprets field
345 * MIPCount/LOD as MIPCount. The range of levels accessible by the
346 * sampler engine is [SurfaceMinLOD, SurfaceMinLOD + MIPCountLOD].
347 */
348 s.SurfaceMinLOD = info->view->base_level;
349 s.MIPCountLOD = MAX(info->view->levels, 1) - 1;
350 }
351
352 #if GEN_GEN >= 9
353 /* We don't use miptails yet. The PRM recommends that you set "Mip Tail
354 * Start LOD" to 15 to prevent the hardware from trying to use them.
355 */
356 s.TiledResourceMode = NONE;
357 s.MipTailStartLOD = 15;
358 #endif
359
360 #if GEN_GEN >= 6
361 const struct isl_extent3d image_align = get_image_alignment(info->surf);
362 s.SurfaceVerticalAlignment = isl_to_gen_valign[image_align.height];
363 #if GEN_GEN >= 7
364 s.SurfaceHorizontalAlignment = isl_to_gen_halign[image_align.width];
365 #endif
366 #endif
367
368 if (info->surf->dim_layout == ISL_DIM_LAYOUT_GEN9_1D) {
369 /* For gen9 1-D textures, surface pitch is ignored */
370 s.SurfacePitch = 0;
371 } else {
372 s.SurfacePitch = info->surf->row_pitch - 1;
373 }
374
375 #if GEN_GEN >= 8
376 s.SurfaceQPitch = get_qpitch(info->surf) >> 2;
377 #elif GEN_GEN == 7
378 s.SurfaceArraySpacing = info->surf->array_pitch_span ==
379 ISL_ARRAY_PITCH_SPAN_COMPACT;
380 #endif
381
382 #if GEN_GEN >= 8
383 s.TileMode = isl_to_gen_tiling[info->surf->tiling];
384 #else
385 s.TiledSurface = info->surf->tiling != ISL_TILING_LINEAR,
386 s.TileWalk = info->surf->tiling == ISL_TILING_Y0 ? TILEWALK_YMAJOR :
387 TILEWALK_XMAJOR,
388 #endif
389
390 #if GEN_GEN >= 8
391 s.RenderCacheReadWriteMode = WriteOnlyCache;
392 #else
393 s.RenderCacheReadWriteMode = 0;
394 #endif
395
396 if (info->view->usage & ISL_SURF_USAGE_CUBE_BIT) {
397 #if GEN_GEN >= 8
398 s.CubeFaceEnablePositiveZ = 1;
399 s.CubeFaceEnableNegativeZ = 1;
400 s.CubeFaceEnablePositiveY = 1;
401 s.CubeFaceEnableNegativeY = 1;
402 s.CubeFaceEnablePositiveX = 1;
403 s.CubeFaceEnableNegativeX = 1;
404 #else
405 s.CubeFaceEnables = 0x3f;
406 #endif
407 }
408
409 #if GEN_GEN >= 6
410 s.NumberofMultisamples = ffs(info->surf->samples) - 1;
411 #if GEN_GEN >= 7
412 s.MultisampledSurfaceStorageFormat =
413 isl_to_gen_multisample_layout[info->surf->msaa_layout];
414 #endif
415 #endif
416
417 #if (GEN_GEN >= 8 || GEN_IS_HASWELL)
418 s.ShaderChannelSelectRed = info->view->swizzle.r;
419 s.ShaderChannelSelectGreen = info->view->swizzle.g;
420 s.ShaderChannelSelectBlue = info->view->swizzle.b;
421 s.ShaderChannelSelectAlpha = info->view->swizzle.a;
422 #endif
423
424 s.SurfaceBaseAddress = info->address;
425
426 #if GEN_GEN >= 6
427 s.MOCS = info->mocs;
428 #endif
429
430 #if GEN_GEN > 4 || GEN_IS_G4X
431 if (info->x_offset_sa != 0 || info->y_offset_sa != 0) {
432 /* There are fairly strict rules about when the offsets can be used.
433 * These are mostly taken from the Sky Lake PRM documentation for
434 * RENDER_SURFACE_STATE.
435 */
436 assert(info->surf->tiling != ISL_TILING_LINEAR);
437 assert(info->surf->dim == ISL_SURF_DIM_2D);
438 assert(isl_is_pow2(isl_format_get_layout(info->view->format)->bpb));
439 assert(info->surf->levels == 1);
440 assert(info->surf->logical_level0_px.array_len == 1);
441 assert(info->aux_usage == ISL_AUX_USAGE_NONE);
442
443 if (GEN_GEN >= 8) {
444 /* Broadwell added more rules. */
445 assert(info->surf->samples == 1);
446 if (isl_format_get_layout(info->view->format)->bpb == 8)
447 assert(info->x_offset_sa % 16 == 0);
448 if (isl_format_get_layout(info->view->format)->bpb == 16)
449 assert(info->x_offset_sa % 8 == 0);
450 }
451
452 #if GEN_GEN >= 7
453 s.SurfaceArray = false;
454 #endif
455 }
456
457 const unsigned x_div = 4;
458 const unsigned y_div = GEN_GEN >= 8 ? 4 : 2;
459 assert(info->x_offset_sa % x_div == 0);
460 assert(info->y_offset_sa % y_div == 0);
461 s.XOffset = info->x_offset_sa / x_div;
462 s.YOffset = info->y_offset_sa / y_div;
463 #else
464 assert(info->x_offset_sa == 0);
465 assert(info->y_offset_sa == 0);
466 #endif
467
468 #if GEN_GEN >= 7
469 if (info->aux_surf && info->aux_usage != ISL_AUX_USAGE_NONE) {
470 struct isl_tile_info tile_info;
471 isl_surf_get_tile_info(dev, info->aux_surf, &tile_info);
472 uint32_t pitch_in_tiles =
473 info->aux_surf->row_pitch / tile_info.phys_extent_B.width;
474
475 #if GEN_GEN >= 8
476 assert(GEN_GEN >= 9 || info->aux_usage != ISL_AUX_USAGE_CCS_E);
477 s.AuxiliarySurfacePitch = pitch_in_tiles - 1;
478 /* Auxiliary surfaces in ISL have compressed formats but the hardware
479 * doesn't expect our definition of the compression, it expects qpitch
480 * in units of samples on the main surface.
481 */
482 s.AuxiliarySurfaceQPitch =
483 isl_surf_get_array_pitch_sa_rows(info->aux_surf) >> 2;
484 s.AuxiliarySurfaceBaseAddress = info->aux_address;
485 s.AuxiliarySurfaceMode = isl_to_gen_aux_mode[info->aux_usage];
486 #else
487 assert(info->aux_usage == ISL_AUX_USAGE_MCS ||
488 info->aux_usage == ISL_AUX_USAGE_CCS_D);
489 s.MCSBaseAddress = info->aux_address,
490 s.MCSSurfacePitch = pitch_in_tiles - 1;
491 s.MCSEnable = true;
492 #endif
493 }
494 #endif
495
496 #if GEN_GEN >= 8
497 /* From the CHV PRM, Volume 2d, page 321 (RENDER_SURFACE_STATE dword 0
498 * bit 9 "Sampler L2 Bypass Mode Disable" Programming Notes):
499 *
500 * This bit must be set for the following surface types: BC2_UNORM
501 * BC3_UNORM BC5_UNORM BC5_SNORM BC7_UNORM
502 */
503 if (GEN_GEN >= 9 || dev->info->is_cherryview) {
504 switch (info->view->format) {
505 case ISL_FORMAT_BC2_UNORM:
506 case ISL_FORMAT_BC3_UNORM:
507 case ISL_FORMAT_BC5_UNORM:
508 case ISL_FORMAT_BC5_SNORM:
509 case ISL_FORMAT_BC7_UNORM:
510 s.SamplerL2BypassModeDisable = true;
511 break;
512 default:
513 break;
514 }
515 }
516 #endif
517
518 if (info->aux_usage != ISL_AUX_USAGE_NONE) {
519 #if GEN_GEN >= 9
520 s.RedClearColor = info->clear_color.u32[0];
521 s.GreenClearColor = info->clear_color.u32[1];
522 s.BlueClearColor = info->clear_color.u32[2];
523 s.AlphaClearColor = info->clear_color.u32[3];
524 #elif GEN_GEN >= 7
525 /* Prior to Sky Lake, we only have one bit for the clear color which
526 * gives us 0 or 1 in whatever the surface's format happens to be.
527 */
528 if (isl_format_has_int_channel(info->view->format)) {
529 for (unsigned i = 0; i < 4; i++) {
530 assert(info->clear_color.u32[i] == 0 ||
531 info->clear_color.u32[i] == 1);
532 }
533 s.RedClearColor = info->clear_color.u32[0] != 0;
534 s.GreenClearColor = info->clear_color.u32[1] != 0;
535 s.BlueClearColor = info->clear_color.u32[2] != 0;
536 s.AlphaClearColor = info->clear_color.u32[3] != 0;
537 } else {
538 for (unsigned i = 0; i < 4; i++) {
539 assert(info->clear_color.f32[i] == 0.0f ||
540 info->clear_color.f32[i] == 1.0f);
541 }
542 s.RedClearColor = info->clear_color.f32[0] != 0.0f;
543 s.GreenClearColor = info->clear_color.f32[1] != 0.0f;
544 s.BlueClearColor = info->clear_color.f32[2] != 0.0f;
545 s.AlphaClearColor = info->clear_color.f32[3] != 0.0f;
546 }
547 #endif
548 }
549
550 GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &s);
551 }
552
553 void
554 isl_genX(buffer_fill_state_s)(void *state,
555 const struct isl_buffer_fill_state_info *restrict info)
556 {
557 uint32_t num_elements = info->size / info->stride;
558
559 if (GEN_GEN >= 7) {
560 /* From the IVB PRM, SURFACE_STATE::Height,
561 *
562 * For typed buffer and structured buffer surfaces, the number
563 * of entries in the buffer ranges from 1 to 2^27. For raw buffer
564 * surfaces, the number of entries in the buffer is the number of bytes
565 * which can range from 1 to 2^30.
566 */
567 if (info->format == ISL_FORMAT_RAW) {
568 assert(num_elements <= (1ull << 30));
569 assert((num_elements & 3) == 0);
570 } else {
571 assert(num_elements <= (1ull << 27));
572 }
573 } else {
574 assert(num_elements <= (1ull << 27));
575 }
576
577 struct GENX(RENDER_SURFACE_STATE) s = { 0, };
578
579 s.SurfaceType = SURFTYPE_BUFFER;
580 s.SurfaceFormat = info->format;
581
582 #if GEN_GEN >= 6
583 s.SurfaceVerticalAlignment = isl_to_gen_valign[4];
584 #if GEN_GEN >= 7
585 s.SurfaceHorizontalAlignment = isl_to_gen_halign[4];
586 s.SurfaceArray = false;
587 #endif
588 #endif
589
590 #if GEN_GEN >= 7
591 s.Height = ((num_elements - 1) >> 7) & 0x3fff;
592 s.Width = (num_elements - 1) & 0x7f;
593 s.Depth = ((num_elements - 1) >> 21) & 0x3ff;
594 #else
595 s.Height = ((num_elements - 1) >> 7) & 0x1fff;
596 s.Width = (num_elements - 1) & 0x7f;
597 s.Depth = ((num_elements - 1) >> 20) & 0x7f;
598 #endif
599
600 s.SurfacePitch = info->stride - 1;
601
602 #if GEN_GEN >= 6
603 s.NumberofMultisamples = MULTISAMPLECOUNT_1;
604 #endif
605
606 #if (GEN_GEN >= 8)
607 s.TileMode = LINEAR;
608 #else
609 s.TiledSurface = false;
610 #endif
611
612 #if (GEN_GEN >= 8)
613 s.RenderCacheReadWriteMode = WriteOnlyCache;
614 #else
615 s.RenderCacheReadWriteMode = 0;
616 #endif
617
618 s.SurfaceBaseAddress = info->address;
619 #if GEN_GEN >= 6
620 s.MOCS = info->mocs;
621 #endif
622
623 #if (GEN_GEN >= 8 || GEN_IS_HASWELL)
624 s.ShaderChannelSelectRed = SCS_RED;
625 s.ShaderChannelSelectGreen = SCS_GREEN;
626 s.ShaderChannelSelectBlue = SCS_BLUE;
627 s.ShaderChannelSelectAlpha = SCS_ALPHA;
628 #endif
629
630 GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &s);
631 }