1 /*
2 * Copyright 2015 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 /**
25 * @file
26 * @brief Intel Surface Layout
27 *
28 * Header Layout
29 * -------------
30 * The header is ordered as:
31 * - forward declarations
32 * - macros that may be overridden at compile-time for specific gens
33 * - enums and constants
34 * - structs and unions
35 * - functions
36 */
37
38 #ifndef ISL_H
39 #define ISL_H
40
41 #include <assert.h>
42 #include <stdbool.h>
43 #include <stdint.h>
44
45 #include "c99_compat.h"
46 #include "util/macros.h"
47
48 #ifdef __cplusplus
49 extern "C" {
50 #endif
51
52 struct gen_device_info;
53 struct brw_image_param;
54
55 #ifndef ISL_DEV_GEN
56 /**
57 * @brief Get the hardware generation of isl_device.
58 *
59 * You can define this as a compile-time constant in the CFLAGS. For example,
60 * `gcc -DISL_DEV_GEN(dev)=9 ...`.
61 */
62 #define ISL_DEV_GEN(__dev) ((__dev)->info->gen)
63 #define ISL_DEV_GEN_SANITIZE(__dev)
64 #else
65 #define ISL_DEV_GEN_SANITIZE(__dev) \
66 (assert(ISL_DEV_GEN(__dev) == (__dev)->info->gen))
67 #endif
68
69 #ifndef ISL_DEV_IS_G4X
70 #define ISL_DEV_IS_G4X(__dev) ((__dev)->info->is_g4x)
71 #endif
72
73 #ifndef ISL_DEV_IS_HASWELL
74 /**
75 * @brief Get the hardware generation of isl_device.
76 *
77 * You can define this as a compile-time constant in the CFLAGS. For example,
78 * `gcc -DISL_DEV_GEN(dev)=9 ...`.
79 */
80 #define ISL_DEV_IS_HASWELL(__dev) ((__dev)->info->is_haswell)
81 #endif
82
83 #ifndef ISL_DEV_IS_BAYTRAIL
84 #define ISL_DEV_IS_BAYTRAIL(__dev) ((__dev)->info->is_baytrail)
85 #endif
86
87 #ifndef ISL_DEV_USE_SEPARATE_STENCIL
88 /**
89 * You can define this as a compile-time constant in the CFLAGS. For example,
90 * `gcc -DISL_DEV_USE_SEPARATE_STENCIL(dev)=1 ...`.
91 */
92 #define ISL_DEV_USE_SEPARATE_STENCIL(__dev) ((__dev)->use_separate_stencil)
93 #define ISL_DEV_USE_SEPARATE_STENCIL_SANITIZE(__dev)
94 #else
95 #define ISL_DEV_USE_SEPARATE_STENCIL_SANITIZE(__dev) \
96 (assert(ISL_DEV_USE_SEPARATE_STENCIL(__dev) == (__dev)->use_separate_stencil))
97 #endif
98
99 /**
100 * Hardware enumeration SURFACE_FORMAT.
101 *
102 * For the official list, see Broadwell PRM: Volume 2b: Command Reference:
103 * Enumerations: SURFACE_FORMAT.
104 */
105 enum isl_format {
106 ISL_FORMAT_R32G32B32A32_FLOAT = 0,
107 ISL_FORMAT_R32G32B32A32_SINT = 1,
108 ISL_FORMAT_R32G32B32A32_UINT = 2,
109 ISL_FORMAT_R32G32B32A32_UNORM = 3,
110 ISL_FORMAT_R32G32B32A32_SNORM = 4,
111 ISL_FORMAT_R64G64_FLOAT = 5,
112 ISL_FORMAT_R32G32B32X32_FLOAT = 6,
113 ISL_FORMAT_R32G32B32A32_SSCALED = 7,
114 ISL_FORMAT_R32G32B32A32_USCALED = 8,
115 ISL_FORMAT_R32G32B32A32_SFIXED = 32,
116 ISL_FORMAT_R64G64_PASSTHRU = 33,
117 ISL_FORMAT_R32G32B32_FLOAT = 64,
118 ISL_FORMAT_R32G32B32_SINT = 65,
119 ISL_FORMAT_R32G32B32_UINT = 66,
120 ISL_FORMAT_R32G32B32_UNORM = 67,
121 ISL_FORMAT_R32G32B32_SNORM = 68,
122 ISL_FORMAT_R32G32B32_SSCALED = 69,
123 ISL_FORMAT_R32G32B32_USCALED = 70,
124 ISL_FORMAT_R32G32B32_SFIXED = 80,
125 ISL_FORMAT_R16G16B16A16_UNORM = 128,
126 ISL_FORMAT_R16G16B16A16_SNORM = 129,
127 ISL_FORMAT_R16G16B16A16_SINT = 130,
128 ISL_FORMAT_R16G16B16A16_UINT = 131,
129 ISL_FORMAT_R16G16B16A16_FLOAT = 132,
130 ISL_FORMAT_R32G32_FLOAT = 133,
131 ISL_FORMAT_R32G32_SINT = 134,
132 ISL_FORMAT_R32G32_UINT = 135,
133 ISL_FORMAT_R32_FLOAT_X8X24_TYPELESS = 136,
134 ISL_FORMAT_X32_TYPELESS_G8X24_UINT = 137,
135 ISL_FORMAT_L32A32_FLOAT = 138,
136 ISL_FORMAT_R32G32_UNORM = 139,
137 ISL_FORMAT_R32G32_SNORM = 140,
138 ISL_FORMAT_R64_FLOAT = 141,
139 ISL_FORMAT_R16G16B16X16_UNORM = 142,
140 ISL_FORMAT_R16G16B16X16_FLOAT = 143,
141 ISL_FORMAT_A32X32_FLOAT = 144,
142 ISL_FORMAT_L32X32_FLOAT = 145,
143 ISL_FORMAT_I32X32_FLOAT = 146,
144 ISL_FORMAT_R16G16B16A16_SSCALED = 147,
145 ISL_FORMAT_R16G16B16A16_USCALED = 148,
146 ISL_FORMAT_R32G32_SSCALED = 149,
147 ISL_FORMAT_R32G32_USCALED = 150,
148 ISL_FORMAT_R32G32_FLOAT_LD = 151,
149 ISL_FORMAT_R32G32_SFIXED = 160,
150 ISL_FORMAT_R64_PASSTHRU = 161,
151 ISL_FORMAT_B8G8R8A8_UNORM = 192,
152 ISL_FORMAT_B8G8R8A8_UNORM_SRGB = 193,
153 ISL_FORMAT_R10G10B10A2_UNORM = 194,
154 ISL_FORMAT_R10G10B10A2_UNORM_SRGB = 195,
155 ISL_FORMAT_R10G10B10A2_UINT = 196,
156 ISL_FORMAT_R10G10B10_SNORM_A2_UNORM = 197,
157 ISL_FORMAT_R8G8B8A8_UNORM = 199,
158 ISL_FORMAT_R8G8B8A8_UNORM_SRGB = 200,
159 ISL_FORMAT_R8G8B8A8_SNORM = 201,
160 ISL_FORMAT_R8G8B8A8_SINT = 202,
161 ISL_FORMAT_R8G8B8A8_UINT = 203,
162 ISL_FORMAT_R16G16_UNORM = 204,
163 ISL_FORMAT_R16G16_SNORM = 205,
164 ISL_FORMAT_R16G16_SINT = 206,
165 ISL_FORMAT_R16G16_UINT = 207,
166 ISL_FORMAT_R16G16_FLOAT = 208,
167 ISL_FORMAT_B10G10R10A2_UNORM = 209,
168 ISL_FORMAT_B10G10R10A2_UNORM_SRGB = 210,
169 ISL_FORMAT_R11G11B10_FLOAT = 211,
170 ISL_FORMAT_R32_SINT = 214,
171 ISL_FORMAT_R32_UINT = 215,
172 ISL_FORMAT_R32_FLOAT = 216,
173 ISL_FORMAT_R24_UNORM_X8_TYPELESS = 217,
174 ISL_FORMAT_X24_TYPELESS_G8_UINT = 218,
175 ISL_FORMAT_L32_UNORM = 221,
176 ISL_FORMAT_A32_UNORM = 222,
177 ISL_FORMAT_L16A16_UNORM = 223,
178 ISL_FORMAT_I24X8_UNORM = 224,
179 ISL_FORMAT_L24X8_UNORM = 225,
180 ISL_FORMAT_A24X8_UNORM = 226,
181 ISL_FORMAT_I32_FLOAT = 227,
182 ISL_FORMAT_L32_FLOAT = 228,
183 ISL_FORMAT_A32_FLOAT = 229,
184 ISL_FORMAT_X8B8_UNORM_G8R8_SNORM = 230,
185 ISL_FORMAT_A8X8_UNORM_G8R8_SNORM = 231,
186 ISL_FORMAT_B8X8_UNORM_G8R8_SNORM = 232,
187 ISL_FORMAT_B8G8R8X8_UNORM = 233,
188 ISL_FORMAT_B8G8R8X8_UNORM_SRGB = 234,
189 ISL_FORMAT_R8G8B8X8_UNORM = 235,
190 ISL_FORMAT_R8G8B8X8_UNORM_SRGB = 236,
191 ISL_FORMAT_R9G9B9E5_SHAREDEXP = 237,
192 ISL_FORMAT_B10G10R10X2_UNORM = 238,
193 ISL_FORMAT_L16A16_FLOAT = 240,
194 ISL_FORMAT_R32_UNORM = 241,
195 ISL_FORMAT_R32_SNORM = 242,
196 ISL_FORMAT_R10G10B10X2_USCALED = 243,
197 ISL_FORMAT_R8G8B8A8_SSCALED = 244,
198 ISL_FORMAT_R8G8B8A8_USCALED = 245,
199 ISL_FORMAT_R16G16_SSCALED = 246,
200 ISL_FORMAT_R16G16_USCALED = 247,
201 ISL_FORMAT_R32_SSCALED = 248,
202 ISL_FORMAT_R32_USCALED = 249,
203 ISL_FORMAT_B5G6R5_UNORM = 256,
204 ISL_FORMAT_B5G6R5_UNORM_SRGB = 257,
205 ISL_FORMAT_B5G5R5A1_UNORM = 258,
206 ISL_FORMAT_B5G5R5A1_UNORM_SRGB = 259,
207 ISL_FORMAT_B4G4R4A4_UNORM = 260,
208 ISL_FORMAT_B4G4R4A4_UNORM_SRGB = 261,
209 ISL_FORMAT_R8G8_UNORM = 262,
210 ISL_FORMAT_R8G8_SNORM = 263,
211 ISL_FORMAT_R8G8_SINT = 264,
212 ISL_FORMAT_R8G8_UINT = 265,
213 ISL_FORMAT_R16_UNORM = 266,
214 ISL_FORMAT_R16_SNORM = 267,
215 ISL_FORMAT_R16_SINT = 268,
216 ISL_FORMAT_R16_UINT = 269,
217 ISL_FORMAT_R16_FLOAT = 270,
218 ISL_FORMAT_A8P8_UNORM_PALETTE0 = 271,
219 ISL_FORMAT_A8P8_UNORM_PALETTE1 = 272,
220 ISL_FORMAT_I16_UNORM = 273,
221 ISL_FORMAT_L16_UNORM = 274,
222 ISL_FORMAT_A16_UNORM = 275,
223 ISL_FORMAT_L8A8_UNORM = 276,
224 ISL_FORMAT_I16_FLOAT = 277,
225 ISL_FORMAT_L16_FLOAT = 278,
226 ISL_FORMAT_A16_FLOAT = 279,
227 ISL_FORMAT_L8A8_UNORM_SRGB = 280,
228 ISL_FORMAT_R5G5_SNORM_B6_UNORM = 281,
229 ISL_FORMAT_B5G5R5X1_UNORM = 282,
230 ISL_FORMAT_B5G5R5X1_UNORM_SRGB = 283,
231 ISL_FORMAT_R8G8_SSCALED = 284,
232 ISL_FORMAT_R8G8_USCALED = 285,
233 ISL_FORMAT_R16_SSCALED = 286,
234 ISL_FORMAT_R16_USCALED = 287,
235 ISL_FORMAT_P8A8_UNORM_PALETTE0 = 290,
236 ISL_FORMAT_P8A8_UNORM_PALETTE1 = 291,
237 ISL_FORMAT_A1B5G5R5_UNORM = 292,
238 ISL_FORMAT_A4B4G4R4_UNORM = 293,
239 ISL_FORMAT_L8A8_UINT = 294,
240 ISL_FORMAT_L8A8_SINT = 295,
241 ISL_FORMAT_R8_UNORM = 320,
242 ISL_FORMAT_R8_SNORM = 321,
243 ISL_FORMAT_R8_SINT = 322,
244 ISL_FORMAT_R8_UINT = 323,
245 ISL_FORMAT_A8_UNORM = 324,
246 ISL_FORMAT_I8_UNORM = 325,
247 ISL_FORMAT_L8_UNORM = 326,
248 ISL_FORMAT_P4A4_UNORM_PALETTE0 = 327,
249 ISL_FORMAT_A4P4_UNORM_PALETTE0 = 328,
250 ISL_FORMAT_R8_SSCALED = 329,
251 ISL_FORMAT_R8_USCALED = 330,
252 ISL_FORMAT_P8_UNORM_PALETTE0 = 331,
253 ISL_FORMAT_L8_UNORM_SRGB = 332,
254 ISL_FORMAT_P8_UNORM_PALETTE1 = 333,
255 ISL_FORMAT_P4A4_UNORM_PALETTE1 = 334,
256 ISL_FORMAT_A4P4_UNORM_PALETTE1 = 335,
257 ISL_FORMAT_Y8_UNORM = 336,
258 ISL_FORMAT_L8_UINT = 338,
259 ISL_FORMAT_L8_SINT = 339,
260 ISL_FORMAT_I8_UINT = 340,
261 ISL_FORMAT_I8_SINT = 341,
262 ISL_FORMAT_DXT1_RGB_SRGB = 384,
263 ISL_FORMAT_R1_UNORM = 385,
264 ISL_FORMAT_YCRCB_NORMAL = 386,
265 ISL_FORMAT_YCRCB_SWAPUVY = 387,
266 ISL_FORMAT_P2_UNORM_PALETTE0 = 388,
267 ISL_FORMAT_P2_UNORM_PALETTE1 = 389,
268 ISL_FORMAT_BC1_UNORM = 390,
269 ISL_FORMAT_BC2_UNORM = 391,
270 ISL_FORMAT_BC3_UNORM = 392,
271 ISL_FORMAT_BC4_UNORM = 393,
272 ISL_FORMAT_BC5_UNORM = 394,
273 ISL_FORMAT_BC1_UNORM_SRGB = 395,
274 ISL_FORMAT_BC2_UNORM_SRGB = 396,
275 ISL_FORMAT_BC3_UNORM_SRGB = 397,
276 ISL_FORMAT_MONO8 = 398,
277 ISL_FORMAT_YCRCB_SWAPUV = 399,
278 ISL_FORMAT_YCRCB_SWAPY = 400,
279 ISL_FORMAT_DXT1_RGB = 401,
280 ISL_FORMAT_FXT1 = 402,
281 ISL_FORMAT_R8G8B8_UNORM = 403,
282 ISL_FORMAT_R8G8B8_SNORM = 404,
283 ISL_FORMAT_R8G8B8_SSCALED = 405,
284 ISL_FORMAT_R8G8B8_USCALED = 406,
285 ISL_FORMAT_R64G64B64A64_FLOAT = 407,
286 ISL_FORMAT_R64G64B64_FLOAT = 408,
287 ISL_FORMAT_BC4_SNORM = 409,
288 ISL_FORMAT_BC5_SNORM = 410,
289 ISL_FORMAT_R16G16B16_FLOAT = 411,
290 ISL_FORMAT_R16G16B16_UNORM = 412,
291 ISL_FORMAT_R16G16B16_SNORM = 413,
292 ISL_FORMAT_R16G16B16_SSCALED = 414,
293 ISL_FORMAT_R16G16B16_USCALED = 415,
294 ISL_FORMAT_BC6H_SF16 = 417,
295 ISL_FORMAT_BC7_UNORM = 418,
296 ISL_FORMAT_BC7_UNORM_SRGB = 419,
297 ISL_FORMAT_BC6H_UF16 = 420,
298 ISL_FORMAT_PLANAR_420_8 = 421,
299 ISL_FORMAT_R8G8B8_UNORM_SRGB = 424,
300 ISL_FORMAT_ETC1_RGB8 = 425,
301 ISL_FORMAT_ETC2_RGB8 = 426,
302 ISL_FORMAT_EAC_R11 = 427,
303 ISL_FORMAT_EAC_RG11 = 428,
304 ISL_FORMAT_EAC_SIGNED_R11 = 429,
305 ISL_FORMAT_EAC_SIGNED_RG11 = 430,
306 ISL_FORMAT_ETC2_SRGB8 = 431,
307 ISL_FORMAT_R16G16B16_UINT = 432,
308 ISL_FORMAT_R16G16B16_SINT = 433,
309 ISL_FORMAT_R32_SFIXED = 434,
310 ISL_FORMAT_R10G10B10A2_SNORM = 435,
311 ISL_FORMAT_R10G10B10A2_USCALED = 436,
312 ISL_FORMAT_R10G10B10A2_SSCALED = 437,
313 ISL_FORMAT_R10G10B10A2_SINT = 438,
314 ISL_FORMAT_B10G10R10A2_SNORM = 439,
315 ISL_FORMAT_B10G10R10A2_USCALED = 440,
316 ISL_FORMAT_B10G10R10A2_SSCALED = 441,
317 ISL_FORMAT_B10G10R10A2_UINT = 442,
318 ISL_FORMAT_B10G10R10A2_SINT = 443,
319 ISL_FORMAT_R64G64B64A64_PASSTHRU = 444,
320 ISL_FORMAT_R64G64B64_PASSTHRU = 445,
321 ISL_FORMAT_ETC2_RGB8_PTA = 448,
322 ISL_FORMAT_ETC2_SRGB8_PTA = 449,
323 ISL_FORMAT_ETC2_EAC_RGBA8 = 450,
324 ISL_FORMAT_ETC2_EAC_SRGB8_A8 = 451,
325 ISL_FORMAT_R8G8B8_UINT = 456,
326 ISL_FORMAT_R8G8B8_SINT = 457,
327 ISL_FORMAT_RAW = 511,
328 ISL_FORMAT_ASTC_LDR_2D_4X4_U8SRGB = 512,
329 ISL_FORMAT_ASTC_LDR_2D_5X4_U8SRGB = 520,
330 ISL_FORMAT_ASTC_LDR_2D_5X5_U8SRGB = 521,
331 ISL_FORMAT_ASTC_LDR_2D_6X5_U8SRGB = 529,
332 ISL_FORMAT_ASTC_LDR_2D_6X6_U8SRGB = 530,
333 ISL_FORMAT_ASTC_LDR_2D_8X5_U8SRGB = 545,
334 ISL_FORMAT_ASTC_LDR_2D_8X6_U8SRGB = 546,
335 ISL_FORMAT_ASTC_LDR_2D_8X8_U8SRGB = 548,
336 ISL_FORMAT_ASTC_LDR_2D_10X5_U8SRGB = 561,
337 ISL_FORMAT_ASTC_LDR_2D_10X6_U8SRGB = 562,
338 ISL_FORMAT_ASTC_LDR_2D_10X8_U8SRGB = 564,
339 ISL_FORMAT_ASTC_LDR_2D_10X10_U8SRGB = 566,
340 ISL_FORMAT_ASTC_LDR_2D_12X10_U8SRGB = 574,
341 ISL_FORMAT_ASTC_LDR_2D_12X12_U8SRGB = 575,
342 ISL_FORMAT_ASTC_LDR_2D_4X4_FLT16 = 576,
343 ISL_FORMAT_ASTC_LDR_2D_5X4_FLT16 = 584,
344 ISL_FORMAT_ASTC_LDR_2D_5X5_FLT16 = 585,
345 ISL_FORMAT_ASTC_LDR_2D_6X5_FLT16 = 593,
346 ISL_FORMAT_ASTC_LDR_2D_6X6_FLT16 = 594,
347 ISL_FORMAT_ASTC_LDR_2D_8X5_FLT16 = 609,
348 ISL_FORMAT_ASTC_LDR_2D_8X6_FLT16 = 610,
349 ISL_FORMAT_ASTC_LDR_2D_8X8_FLT16 = 612,
350 ISL_FORMAT_ASTC_LDR_2D_10X5_FLT16 = 625,
351 ISL_FORMAT_ASTC_LDR_2D_10X6_FLT16 = 626,
352 ISL_FORMAT_ASTC_LDR_2D_10X8_FLT16 = 628,
353 ISL_FORMAT_ASTC_LDR_2D_10X10_FLT16 = 630,
354 ISL_FORMAT_ASTC_LDR_2D_12X10_FLT16 = 638,
355 ISL_FORMAT_ASTC_LDR_2D_12X12_FLT16 = 639,
356 ISL_FORMAT_ASTC_HDR_2D_4X4_FLT16 = 832,
357 ISL_FORMAT_ASTC_HDR_2D_5X4_FLT16 = 840,
358 ISL_FORMAT_ASTC_HDR_2D_5X5_FLT16 = 841,
359 ISL_FORMAT_ASTC_HDR_2D_6X5_FLT16 = 849,
360 ISL_FORMAT_ASTC_HDR_2D_6X6_FLT16 = 850,
361 ISL_FORMAT_ASTC_HDR_2D_8X5_FLT16 = 865,
362 ISL_FORMAT_ASTC_HDR_2D_8X6_FLT16 = 866,
363 ISL_FORMAT_ASTC_HDR_2D_8X8_FLT16 = 868,
364 ISL_FORMAT_ASTC_HDR_2D_10X5_FLT16 = 881,
365 ISL_FORMAT_ASTC_HDR_2D_10X6_FLT16 = 882,
366 ISL_FORMAT_ASTC_HDR_2D_10X8_FLT16 = 884,
367 ISL_FORMAT_ASTC_HDR_2D_10X10_FLT16 = 886,
368 ISL_FORMAT_ASTC_HDR_2D_12X10_FLT16 = 894,
369 ISL_FORMAT_ASTC_HDR_2D_12X12_FLT16 = 895,
370
371 /* The formats that follow are internal to ISL and as such don't have an
372 * explicit number. We'll just let the C compiler assign it for us. Any
373 * actual hardware formats *must* come before these in the list.
374 */
375
376 /* Formats for auxiliary surfaces */
377 ISL_FORMAT_HIZ,
378 ISL_FORMAT_MCS_2X,
379 ISL_FORMAT_MCS_4X,
380 ISL_FORMAT_MCS_8X,
381 ISL_FORMAT_MCS_16X,
382 ISL_FORMAT_GEN7_CCS_32BPP_X,
383 ISL_FORMAT_GEN7_CCS_64BPP_X,
384 ISL_FORMAT_GEN7_CCS_128BPP_X,
385 ISL_FORMAT_GEN7_CCS_32BPP_Y,
386 ISL_FORMAT_GEN7_CCS_64BPP_Y,
387 ISL_FORMAT_GEN7_CCS_128BPP_Y,
388 ISL_FORMAT_GEN9_CCS_32BPP,
389 ISL_FORMAT_GEN9_CCS_64BPP,
390 ISL_FORMAT_GEN9_CCS_128BPP,
391
392 /* Hardware doesn't understand this out-of-band value */
393 ISL_FORMAT_UNSUPPORTED = UINT16_MAX,
394 };
395
396 /**
397 * Numerical base type for channels of isl_format.
398 */
399 enum isl_base_type {
400 ISL_VOID,
401 ISL_RAW,
402 ISL_UNORM,
403 ISL_SNORM,
404 ISL_UFLOAT,
405 ISL_SFLOAT,
406 ISL_UFIXED,
407 ISL_SFIXED,
408 ISL_UINT,
409 ISL_SINT,
410 ISL_USCALED,
411 ISL_SSCALED,
412 };
413
414 /**
415 * Colorspace of isl_format.
416 */
417 enum isl_colorspace {
418 ISL_COLORSPACE_NONE = 0,
419 ISL_COLORSPACE_LINEAR,
420 ISL_COLORSPACE_SRGB,
421 ISL_COLORSPACE_YUV,
422 };
423
424 /**
425 * Texture compression mode of isl_format.
426 */
427 enum isl_txc {
428 ISL_TXC_NONE = 0,
429 ISL_TXC_DXT1,
430 ISL_TXC_DXT3,
431 ISL_TXC_DXT5,
432 ISL_TXC_FXT1,
433 ISL_TXC_RGTC1,
434 ISL_TXC_RGTC2,
435 ISL_TXC_BPTC,
436 ISL_TXC_ETC1,
437 ISL_TXC_ETC2,
438 ISL_TXC_ASTC,
439
440 /* Used for auxiliary surface formats */
441 ISL_TXC_HIZ,
442 ISL_TXC_MCS,
443 ISL_TXC_CCS,
444 };
445
446 /**
447 * @brief Hardware tile mode
448 *
449 * WARNING: These values differ from the hardware enum values, which are
450 * unstable across hardware generations.
451 *
452 * Note that legacy Y tiling is ISL_TILING_Y0 instead of ISL_TILING_Y, to
453 * clearly distinguish it from Yf and Ys.
454 */
455 enum isl_tiling {
456 ISL_TILING_LINEAR = 0,
457 ISL_TILING_W,
458 ISL_TILING_X,
459 ISL_TILING_Y0, /**< Legacy Y tiling */
460 ISL_TILING_Yf, /**< Standard 4K tiling. The 'f' means "four". */
461 ISL_TILING_Ys, /**< Standard 64K tiling. The 's' means "sixty-four". */
462 ISL_TILING_HIZ, /**< Tiling format for HiZ surfaces */
463 ISL_TILING_CCS, /**< Tiling format for CCS surfaces */
464 };
465
466 /**
467 * @defgroup Tiling Flags
468 * @{
469 */
470 typedef uint32_t isl_tiling_flags_t;
471 #define ISL_TILING_LINEAR_BIT (1u << ISL_TILING_LINEAR)
472 #define ISL_TILING_W_BIT (1u << ISL_TILING_W)
473 #define ISL_TILING_X_BIT (1u << ISL_TILING_X)
474 #define ISL_TILING_Y0_BIT (1u << ISL_TILING_Y0)
475 #define ISL_TILING_Yf_BIT (1u << ISL_TILING_Yf)
476 #define ISL_TILING_Ys_BIT (1u << ISL_TILING_Ys)
477 #define ISL_TILING_HIZ_BIT (1u << ISL_TILING_HIZ)
478 #define ISL_TILING_CCS_BIT (1u << ISL_TILING_CCS)
479 #define ISL_TILING_ANY_MASK (~0u)
480 #define ISL_TILING_NON_LINEAR_MASK (~ISL_TILING_LINEAR_BIT)
481
482 /** Any Y tiling, including legacy Y tiling. */
483 #define ISL_TILING_ANY_Y_MASK (ISL_TILING_Y0_BIT | \
484 ISL_TILING_Yf_BIT | \
485 ISL_TILING_Ys_BIT)
486
487 /** The Skylake BSpec refers to Yf and Ys as "standard tiling formats". */
488 #define ISL_TILING_STD_Y_MASK (ISL_TILING_Yf_BIT | \
489 ISL_TILING_Ys_BIT)
490 /** @} */
491
492 /**
493 * @brief Logical dimension of surface.
494 *
495 * Note: There is no dimension for cube map surfaces. ISL interprets cube maps
496 * as 2D array surfaces.
497 */
498 enum isl_surf_dim {
499 ISL_SURF_DIM_1D,
500 ISL_SURF_DIM_2D,
501 ISL_SURF_DIM_3D,
502 };
503
504 /**
505 * @brief Physical layout of the surface's dimensions.
506 */
507 enum isl_dim_layout {
508 /**
509 * For details, see the G35 PRM >> Volume 1: Graphics Core >> Section
510 * 6.17.3: 2D Surfaces.
511 *
512 * On many gens, 1D surfaces share the same layout as 2D surfaces. From
513 * the G35 PRM >> Volume 1: Graphics Core >> Section 6.17.2: 1D Surfaces:
514 *
515 * One-dimensional surfaces are identical to 2D surfaces with height of
516 * one.
517 *
518 * @invariant isl_surf::phys_level0_sa::depth == 1
519 */
520 ISL_DIM_LAYOUT_GEN4_2D,
521
522 /**
523 * For details, see the G35 PRM >> Volume 1: Graphics Core >> Section
524 * 6.17.5: 3D Surfaces.
525 *
526 * @invariant isl_surf::phys_level0_sa::array_len == 1
527 */
528 ISL_DIM_LAYOUT_GEN4_3D,
529
530 /**
531 * Special layout used for HiZ and stencil on Sandy Bridge to work around
532 * the hardware's lack of mipmap support. On gen6, HiZ and stencil buffers
533 * work the same as on gen7+ except that they don't technically support
534 * mipmapping. That does not, however, stop us from doing it. As far as
535 * Sandy Bridge hardware is concerned, HiZ and stencil always operates on a
536 * single miplevel 2D (possibly array) image. The dimensions of that image
537 * are NOT minified.
538 *
539 * In order to implement HiZ and stencil on Sandy Bridge, we create one
540 * full-sized 2D (possibly array) image for every LOD with every image
541 * aligned to a page boundary. When the surface is used with the stencil
542 * or HiZ hardware, we manually offset to the image for the given LOD.
543 *
544 * As a memory saving measure, we pretend that the width of each miplevel
545 * is minified and we place LOD1 and above below LOD0 but horizontally
546 * adjacent to each other. When considered as full-sized images, LOD1 and
547 * above technically overlap. However, since we only write to part of that
548 * image, the hardware will never notice the overlap.
549 *
550 * This layout looks something like this:
551 *
552 * +---------+
553 * | |
554 * | |
555 * +---------+
556 * | |
557 * | |
558 * +---------+
559 *
560 * +----+ +-+ .
561 * | | +-+
562 * +----+
563 *
564 * +----+ +-+ .
565 * | | +-+
566 * +----+
567 */
568 ISL_DIM_LAYOUT_GEN6_STENCIL_HIZ,
569
570 /**
571 * For details, see the Skylake BSpec >> Memory Views >> Common Surface
572 * Formats >> Surface Layout and Tiling >> » 1D Surfaces.
573 */
574 ISL_DIM_LAYOUT_GEN9_1D,
575 };
576
577 enum isl_aux_usage {
578 /** No Auxiliary surface is used */
579 ISL_AUX_USAGE_NONE,
580
581 /** The primary surface is a depth surface and the auxiliary surface is HiZ */
582 ISL_AUX_USAGE_HIZ,
583
584 /** The auxiliary surface is an MCS
585 *
586 * @invariant isl_surf::samples > 1
587 */
588 ISL_AUX_USAGE_MCS,
589
590 /** The auxiliary surface is a fast-clear-only compression surface
591 *
592 * @invariant isl_surf::samples == 1
593 */
594 ISL_AUX_USAGE_CCS_D,
595
596 /** The auxiliary surface provides full lossless color compression
597 *
598 * @invariant isl_surf::samples == 1
599 */
600 ISL_AUX_USAGE_CCS_E,
601 };
602
603 /**
604 * Enum for keeping track of the state an auxiliary compressed surface.
605 *
606 * For any given auxiliary surface compression format (HiZ, CCS, or MCS), any
607 * given slice (lod + array layer) can be in one of the six states described
608 * by this enum. Draw and resolve operations may cause the slice to change
609 * from one state to another. The six valid states are:
610 *
611 * 1) Clear: In this state, each block in the auxiliary surface contains a
612 * magic value that indicates that the block is in the clear state. If
613 * a block is in the clear state, it's values in the primary surface are
614 * ignored and the color of the samples in the block is taken either the
615 * RENDER_SURFACE_STATE packet for color or 3DSTATE_CLEAR_PARAMS for
616 * depth. Since neither the primary surface nor the auxiliary surface
617 * contains the clear value, the surface can be cleared to a different
618 * color by simply changing the clear color without modifying either
619 * surface.
620 *
621 * 2) Partial Clear: In this state, each block in the auxiliary surface
622 * contains either the magic clear or pass-through value. See Clear and
623 * Pass-through for more details.
624 *
625 * 3) Compressed w/ Clear: In this state, neither the auxiliary surface
626 * nor the primary surface has a complete representation of the data.
627 * Instead, both surfaces must be used together or else rendering
628 * corruption may occur. Depending on the auxiliary compression format
629 * and the data, any given block in the primary surface may contain all,
630 * some, or none of the data required to reconstruct the actual sample
631 * values. Blocks may also be in the clear state (see Clear) and have
632 * their value taken from outside the surface.
633 *
634 * 4) Compressed w/o Clear: This state is identical to the state above
635 * except that no blocks are in the clear state. In this state, all of
636 * the data required to reconstruct the final sample values is contained
637 * in the auxiliary and primary surface and the clear value is not
638 * considered.
639 *
640 * 5) Resolved: In this state, the primary surface contains 100% of the
641 * data. The auxiliary surface is also valid so the surface can be
642 * validly used with or without aux enabled. The auxiliary surface may,
643 * however, contain non-trivial data and any update to the primary
644 * surface with aux disabled will cause the two to get out of sync.
645 *
646 * 6) Pass-through: In this state, the primary surface contains 100% of the
647 * data and every block in the auxiliary surface contains a magic value
648 * which indicates that the auxiliary surface should be ignored and the
649 * only the primary surface should be considered. Updating the primary
650 * surface without aux works fine and can be done repeatedly in this
651 * mode. Writing to a surface in pass-through mode with aux enabled may
652 * cause the auxiliary buffer to contain non-trivial data and no longer
653 * be in the pass-through state.
654 *
655 * 7) Aux Invalid: In this state, the primary surface contains 100% of the
656 * data and the auxiliary surface is completely bogus. Any attempt to
657 * use the auxiliary surface is liable to result in rendering
658 * corruption. The only thing that one can do to re-enable aux once
659 * this state is reached is to use an ambiguate pass to transition into
660 * the pass-through state.
661 *
662 * Drawing with or without aux enabled may implicitly cause the surface to
663 * transition between these states. There are also four types of auxiliary
664 * compression operations which cause an explicit transition:
665 *
666 * 1) Fast Clear: This operation writes the magic "clear" value to the
667 * auxiliary surface. This operation will safely transition any slice
668 * of a surface from any state to the clear state so long as the entire
669 * slice is fast cleared at once. A fast clear that only covers part of
670 * a slice of a surface is called a partial fast clear.
671 *
672 * 2) Full Resolve: This operation combines the auxiliary surface data
673 * with the primary surface data and writes the result to the primary.
674 * For HiZ, the docs call this a depth resolve. For CCS, the hardware
675 * full resolve operation does both a full resolve and an ambiguate so
676 * it actually takes you all the way to the pass-through state.
677 *
678 * 3) Partial Resolve: This operation considers blocks which are in the
679 * "clear" state and writes the clear value directly into the primary or
680 * auxiliary surface. Once this operation completes, the surface is
681 * still compressed but no longer references the clear color. This
682 * operation is only available for CCS.
683 *
684 * 4) Ambiguate: This operation throws away the current auxiliary data and
685 * replaces it with the magic pass-through value. If an ambiguate
686 * operation is performed when the primary surface does not contain 100%
687 * of the data, data will be lost. This operation is only implemented
688 * in hardware for depth where it is called a HiZ resolve.
689 *
690 * Not all operations are valid or useful in all states. The diagram below
691 * contains a complete description of the states and all valid and useful
692 * transitions except clear.
693 *
694 * Draw w/ Aux
695 * +----------+
696 * | |
697 * | +-------------+ Draw w/ Aux +-------------+
698 * +------>| Compressed |<-------------------| Clear |
699 * | w/ Clear |----->----+ | |
700 * +-------------+ | +-------------+
701 * | /|\ | | |
702 * | | | | |
703 * | | +------<-----+ | Draw w/
704 * | | | | Clear Only
705 * | | Full | | +----------+
706 * Partial | | Resolve | \|/ | |
707 * Resolve | | | +-------------+ |
708 * | | | | Partial |<------+
709 * | | | | Clear |<----------+
710 * | | | +-------------+ |
711 * | | | | |
712 * | | +------>---------+ Full |
713 * | | | Resolve |
714 * Draw w/ aux | | Partial Fast Clear | |
715 * +----------+ | +--------------------------+ | |
716 * | | \|/ | \|/ |
717 * | +-------------+ Full Resolve +-------------+ |
718 * +------>| Compressed |------------------->| Resolved | |
719 * | w/o Clear |<-------------------| | |
720 * +-------------+ Draw w/ Aux +-------------+ |
721 * /|\ | | |
722 * | Draw | | Draw |
723 * | w/ Aux | | w/o Aux |
724 * | Ambiguate | | |
725 * | +--------------------------+ | |
726 * Draw w/o Aux | | | Draw w/o Aux |
727 * +----------+ | | | +----------+ |
728 * | | | \|/ \|/ | | |
729 * | +-------------+ Ambiguate +-------------+ | |
730 * +------>| Pass- |<-------------------| Aux |<------+ |
731 * +------>| through | | Invalid | |
732 * | +-------------+ +-------------+ |
733 * | | | |
734 * +----------+ +-----------------------------------------------------+
735 * Draw w/ Partial Fast Clear
736 * Clear Only
737 *
738 *
739 * While the above general theory applies to all forms of auxiliary
740 * compression on Intel hardware, not all states and operations are available
741 * on all compression types. However, each of the auxiliary states and
742 * operations can be fairly easily mapped onto the above diagram:
743 *
744 * HiZ: Hierarchical depth compression is capable of being in any of the
745 * states above. Hardware provides three HiZ operations: "Depth
746 * Clear", "Depth Resolve", and "HiZ Resolve" which map to "Fast
747 * Clear", "Full Resolve", and "Ambiguate" respectively. The
748 * hardware provides no HiZ partial resolve operation so the only way
749 * to get into the "Compressed w/o Clear" state is to render with HiZ
750 * when the surface is in the resolved or pass-through states.
751 *
752 * MCS: Multisample compression is technically capable of being in any of
753 * the states above except that most of them aren't useful. Both the
754 * render engine and the sampler support MCS compression and, apart
755 * from clear color, MCS is format-unaware so we leave the surface
756 * compressed 100% of the time. The hardware provides no MCS
757 * operations.
758 *
759 * CCS_D: Single-sample fast-clears (also called CCS_D in ISL) are one of
760 * the simplest forms of compression since they don't do anything
761 * beyond clear color tracking. They really only support three of
762 * the six states: Clear, Partial Clear, and Pass-through. The
763 * only CCS_D operation is "Resolve" which maps to a full resolve
764 * followed by an ambiguate.
765 *
766 * CCS_E: Single-sample render target compression (also called CCS_E in ISL)
767 * is capable of being in almost all of the above states. THe only
768 * exception is that it does not have separate resolved and pass-
769 * through states. Instead, the CCS_E full resolve operation does
770 * both a resolve and an ambiguate so it goes directly into the
771 * pass-through state. CCS_E also provides fast clear and partial
772 * resolve operations which work as described above.
773 *
774 * While it is technically possible to perform a CCS_E ambiguate, it
775 * is not provided by Sky Lake hardware so we choose to avoid the aux
776 * invalid state. If the aux invalid state were determined to be
777 * useful, a CCS ambiguate could be done by carefully rendering to
778 * the CCS and filling it with zeros.
779 */
780 enum isl_aux_state {
781 ISL_AUX_STATE_CLEAR = 0,
782 ISL_AUX_STATE_PARTIAL_CLEAR,
783 ISL_AUX_STATE_COMPRESSED_CLEAR,
784 ISL_AUX_STATE_COMPRESSED_NO_CLEAR,
785 ISL_AUX_STATE_RESOLVED,
786 ISL_AUX_STATE_PASS_THROUGH,
787 ISL_AUX_STATE_AUX_INVALID,
788 };
789
790 /* TODO(chadv): Explain */
791 enum isl_array_pitch_span {
792 ISL_ARRAY_PITCH_SPAN_FULL,
793 ISL_ARRAY_PITCH_SPAN_COMPACT,
794 };
795
796 /**
797 * @defgroup Surface Usage
798 * @{
799 */
800 typedef uint64_t isl_surf_usage_flags_t;
801 #define ISL_SURF_USAGE_RENDER_TARGET_BIT (1u << 0)
802 #define ISL_SURF_USAGE_DEPTH_BIT (1u << 1)
803 #define ISL_SURF_USAGE_STENCIL_BIT (1u << 2)
804 #define ISL_SURF_USAGE_TEXTURE_BIT (1u << 3)
805 #define ISL_SURF_USAGE_CUBE_BIT (1u << 4)
806 #define ISL_SURF_USAGE_DISABLE_AUX_BIT (1u << 5)
807 #define ISL_SURF_USAGE_DISPLAY_BIT (1u << 6)
808 #define ISL_SURF_USAGE_DISPLAY_ROTATE_90_BIT (1u << 7)
809 #define ISL_SURF_USAGE_DISPLAY_ROTATE_180_BIT (1u << 8)
810 #define ISL_SURF_USAGE_DISPLAY_ROTATE_270_BIT (1u << 9)
811 #define ISL_SURF_USAGE_DISPLAY_FLIP_X_BIT (1u << 10)
812 #define ISL_SURF_USAGE_DISPLAY_FLIP_Y_BIT (1u << 11)
813 #define ISL_SURF_USAGE_STORAGE_BIT (1u << 12)
814 #define ISL_SURF_USAGE_HIZ_BIT (1u << 13)
815 #define ISL_SURF_USAGE_MCS_BIT (1u << 14)
816 #define ISL_SURF_USAGE_CCS_BIT (1u << 15)
817 /** @} */
818
819 /**
820 * @defgroup Channel Mask
821 *
822 * These #define values are chosen to match the values of
823 * RENDER_SURFACE_STATE::Color Buffer Component Write Disables
824 *
825 * @{
826 */
827 typedef uint8_t isl_channel_mask_t;
828 #define ISL_CHANNEL_BLUE_BIT (1 << 0)
829 #define ISL_CHANNEL_GREEN_BIT (1 << 1)
830 #define ISL_CHANNEL_RED_BIT (1 << 2)
831 #define ISL_CHANNEL_ALPHA_BIT (1 << 3)
832 /** @} */
833
834 /**
835 * @brief A channel select (also known as texture swizzle) value
836 */
837 enum isl_channel_select {
838 ISL_CHANNEL_SELECT_ZERO = 0,
839 ISL_CHANNEL_SELECT_ONE = 1,
840 ISL_CHANNEL_SELECT_RED = 4,
841 ISL_CHANNEL_SELECT_GREEN = 5,
842 ISL_CHANNEL_SELECT_BLUE = 6,
843 ISL_CHANNEL_SELECT_ALPHA = 7,
844 };
845
846 /**
847 * Identical to VkSampleCountFlagBits.
848 */
849 enum isl_sample_count {
850 ISL_SAMPLE_COUNT_1_BIT = 1u,
851 ISL_SAMPLE_COUNT_2_BIT = 2u,
852 ISL_SAMPLE_COUNT_4_BIT = 4u,
853 ISL_SAMPLE_COUNT_8_BIT = 8u,
854 ISL_SAMPLE_COUNT_16_BIT = 16u,
855 };
856 typedef uint32_t isl_sample_count_mask_t;
857
858 /**
859 * @brief Multisample Format
860 */
861 enum isl_msaa_layout {
862 /**
863 * @brief Suface is single-sampled.
864 */
865 ISL_MSAA_LAYOUT_NONE,
866
867 /**
868 * @brief [SNB+] Interleaved Multisample Format
869 *
870 * In this format, multiple samples are interleaved into each cacheline.
871 * In other words, the sample index is swizzled into the low 6 bits of the
872 * surface's virtual address space.
873 *
874 * For example, suppose the surface is legacy Y tiled, is 4x multisampled,
875 * and its pixel format is 32bpp. Then the first cacheline is arranged
876 * thus:
877 *
878 * (0,0,0) (0,1,0) (0,0,1) (1,0,1)
879 * (1,0,0) (1,1,0) (0,1,1) (1,1,1)
880 *
881 * (0,0,2) (1,0,2) (0,0,3) (1,0,3)
882 * (0,1,2) (1,1,2) (0,1,3) (1,1,3)
883 *
884 * The hardware docs refer to this format with multiple terms. In
885 * Sandybridge, this is the only multisample format; so no term is used.
886 * The Ivybridge docs refer to surfaces in this format as IMS (Interleaved
887 * Multisample Surface). Later hardware docs additionally refer to this
888 * format as MSFMT_DEPTH_STENCIL (because the format is deprecated for
889 * color surfaces).
890 *
891 * See the Sandybridge PRM, Volume 4, Part 1, Section 2.7 "Multisampled
892 * Surface Behavior".
893 *
894 * See the Ivybridge PRM, Volume 1, Part 1, Section 6.18.4.1 "Interleaved
895 * Multisampled Surfaces".
896 */
897 ISL_MSAA_LAYOUT_INTERLEAVED,
898
899 /**
900 * @brief [IVB+] Array Multisample Format
901 *
902 * In this format, the surface's physical layout resembles that of a
903 * 2D array surface.
904 *
905 * Suppose the multisample surface's logical extent is (w, h) and its
906 * sample count is N. Then surface's physical extent is the same as
907 * a singlesample 2D surface whose logical extent is (w, h) and array
908 * length is N. Array slice `i` contains the pixel values for sample
909 * index `i`.
910 *
911 * The Ivybridge docs refer to surfaces in this format as UMS
912 * (Uncompressed Multsample Layout) and CMS (Compressed Multisample
913 * Surface). The Broadwell docs additionally refer to this format as
914 * MSFMT_MSS (MSS=Multisample Surface Storage).
915 *
916 * See the Broadwell PRM, Volume 5 "Memory Views", Section "Uncompressed
917 * Multisample Surfaces".
918 *
919 * See the Broadwell PRM, Volume 5 "Memory Views", Section "Compressed
920 * Multisample Surfaces".
921 */
922 ISL_MSAA_LAYOUT_ARRAY,
923 };
924
925
926 struct isl_device {
927 const struct gen_device_info *info;
928 bool use_separate_stencil;
929 bool has_bit6_swizzling;
930
931 /**
932 * Describes the layout of a RENDER_SURFACE_STATE structure for the
933 * current gen.
934 */
935 struct {
936 uint8_t size;
937 uint8_t align;
938 uint8_t addr_offset;
939 uint8_t aux_addr_offset;
940
941 /* Rounded up to the nearest dword to simplify GPU memcpy operations. */
942 uint8_t clear_value_size;
943 uint8_t clear_value_offset;
944 } ss;
945
946 /**
947 * Describes the layout of the depth/stencil/hiz commands as emitted by
948 * isl_emit_depth_stencil_hiz.
949 */
950 struct {
951 uint8_t size;
952 uint8_t depth_offset;
953 uint8_t stencil_offset;
954 uint8_t hiz_offset;
955 } ds;
956 };
957
958 struct isl_extent2d {
959 union { uint32_t w, width; };
960 union { uint32_t h, height; };
961 };
962
963 struct isl_extent3d {
964 union { uint32_t w, width; };
965 union { uint32_t h, height; };
966 union { uint32_t d, depth; };
967 };
968
969 struct isl_extent4d {
970 union { uint32_t w, width; };
971 union { uint32_t h, height; };
972 union { uint32_t d, depth; };
973 union { uint32_t a, array_len; };
974 };
975
976 struct isl_channel_layout {
977 enum isl_base_type type;
978 uint8_t bits; /**< Size in bits */
979 };
980
981 /**
982 * Each format has 3D block extent (width, height, depth). The block extent of
983 * compressed formats is that of the format's compression block. For example,
984 * the block extent of ISL_FORMAT_ETC2_RGB8 is (w=4, h=4, d=1). The block
985 * extent of uncompressed pixel formats, such as ISL_FORMAT_R8G8B8A8_UNORM, is
986 * is (w=1, h=1, d=1).
987 */
988 struct isl_format_layout {
989 enum isl_format format;
990 const char *name;
991
992 uint16_t bpb; /**< Bits per block */
993 uint8_t bw; /**< Block width, in pixels */
994 uint8_t bh; /**< Block height, in pixels */
995 uint8_t bd; /**< Block depth, in pixels */
996
997 union {
998 struct {
999 struct isl_channel_layout r; /**< Red channel */
1000 struct isl_channel_layout g; /**< Green channel */
1001 struct isl_channel_layout b; /**< Blue channel */
1002 struct isl_channel_layout a; /**< Alpha channel */
1003 struct isl_channel_layout l; /**< Luminance channel */
1004 struct isl_channel_layout i; /**< Intensity channel */
1005 struct isl_channel_layout p; /**< Palette channel */
1006 } channels;
1007 struct isl_channel_layout channels_array[7];
1008 };
1009
1010 enum isl_colorspace colorspace;
1011 enum isl_txc txc;
1012 };
1013
1014 struct isl_tile_info {
1015 enum isl_tiling tiling;
1016
1017 /* The size (in bits per block) of a single surface element
1018 *
1019 * For surfaces with power-of-two formats, this is the same as
1020 * isl_format_layout::bpb. For non-power-of-two formats it may be smaller.
1021 * The logical_extent_el field is in terms of elements of this size.
1022 *
1023 * For example, consider ISL_FORMAT_R32G32B32_FLOAT for which
1024 * isl_format_layout::bpb is 96 (a non-power-of-two). In this case, none
1025 * of the tiling formats can actually hold an integer number of 96-bit
1026 * surface elements so isl_tiling_get_info returns an isl_tile_info for a
1027 * 32-bit element size. It is the responsibility of the caller to
1028 * recognize that 32 != 96 ad adjust accordingly. For instance, to compute
1029 * the width of a surface in tiles, you would do:
1030 *
1031 * width_tl = DIV_ROUND_UP(width_el * (format_bpb / tile_info.format_bpb),
1032 * tile_info.logical_extent_el.width);
1033 */
1034 uint32_t format_bpb;
1035
1036 /** The logical size of the tile in units of format_bpb size elements
1037 *
1038 * This field determines how a given surface is cut up into tiles. It is
1039 * used to compute the size of a surface in tiles and can be used to
1040 * determine the location of the tile containing any given surface element.
1041 * The exact value of this field depends heavily on the bits-per-block of
1042 * the format being used.
1043 */
1044 struct isl_extent2d logical_extent_el;
1045
1046 /** The physical size of the tile in bytes and rows of bytes
1047 *
1048 * This field determines how the tiles of a surface are physically layed
1049 * out in memory. The logical and physical tile extent are frequently the
1050 * same but this is not always the case. For instance, a W-tile (which is
1051 * always used with ISL_FORMAT_R8) has a logical size of 64el x 64el but
1052 * its physical size is 128B x 32rows, the same as a Y-tile.
1053 *
1054 * @see isl_surf::row_pitch
1055 */
1056 struct isl_extent2d phys_extent_B;
1057 };
1058
1059 /**
1060 * Metadata about a DRM format modifier.
1061 */
1062 struct isl_drm_modifier_info {
1063 uint64_t modifier;
1064
1065 /** Text name of the modifier */
1066 const char *name;
1067
1068 /** ISL tiling implied by this modifier */
1069 enum isl_tiling tiling;
1070
1071 /** ISL aux usage implied by this modifier */
1072 enum isl_aux_usage aux_usage;
1073
1074 /** Whether or not this modifier supports clear color */
1075 bool supports_clear_color;
1076 };
1077
1078 /**
1079 * @brief Input to surface initialization
1080 *
1081 * @invariant width >= 1
1082 * @invariant height >= 1
1083 * @invariant depth >= 1
1084 * @invariant levels >= 1
1085 * @invariant samples >= 1
1086 * @invariant array_len >= 1
1087 *
1088 * @invariant if 1D then height == 1 and depth == 1 and samples == 1
1089 * @invariant if 2D then depth == 1
1090 * @invariant if 3D then array_len == 1 and samples == 1
1091 */
1092 struct isl_surf_init_info {
1093 enum isl_surf_dim dim;
1094 enum isl_format format;
1095
1096 uint32_t width;
1097 uint32_t height;
1098 uint32_t depth;
1099 uint32_t levels;
1100 uint32_t array_len;
1101 uint32_t samples;
1102
1103 /** Lower bound for isl_surf::alignment, in bytes. */
1104 uint32_t min_alignment;
1105
1106 /**
1107 * Exact value for isl_surf::row_pitch. Ignored if zero. isl_surf_init()
1108 * will fail if this is misaligned or out of bounds.
1109 */
1110 uint32_t row_pitch;
1111
1112 isl_surf_usage_flags_t usage;
1113
1114 /** Flags that alter how ISL selects isl_surf::tiling. */
1115 isl_tiling_flags_t tiling_flags;
1116 };
1117
1118 struct isl_surf {
1119 enum isl_surf_dim dim;
1120 enum isl_dim_layout dim_layout;
1121 enum isl_msaa_layout msaa_layout;
1122 enum isl_tiling tiling;
1123 enum isl_format format;
1124
1125 /**
1126 * Alignment of the upper-left sample of each subimage, in units of surface
1127 * elements.
1128 */
1129 struct isl_extent3d image_alignment_el;
1130
1131 /**
1132 * Logical extent of the surface's base level, in units of pixels. This is
1133 * identical to the extent defined in isl_surf_init_info.
1134 */
1135 struct isl_extent4d logical_level0_px;
1136
1137 /**
1138 * Physical extent of the surface's base level, in units of physical
1139 * surface samples and aligned to the format's compression block.
1140 *
1141 * Consider isl_dim_layout as an operator that transforms a logical surface
1142 * layout to a physical surface layout. Then
1143 *
1144 * logical_layout := (isl_surf::dim, isl_surf::logical_level0_px)
1145 * isl_surf::phys_level0_sa := isl_surf::dim_layout * logical_layout
1146 */
1147 struct isl_extent4d phys_level0_sa;
1148
1149 uint32_t levels;
1150 uint32_t samples;
1151
1152 /** Total size of the surface, in bytes. */
1153 uint64_t size;
1154
1155 /** Required alignment for the surface's base address. */
1156 uint32_t alignment;
1157
1158 /**
1159 * The interpretation of this field depends on the value of
1160 * isl_tile_info::physical_extent_B. In particular, the width of the
1161 * surface in tiles is row_pitch / isl_tile_info::physical_extent_B.width
1162 * and the distance in bytes between vertically adjacent tiles in the image
1163 * is given by row_pitch * isl_tile_info::physical_extent_B.height.
1164 *
1165 * For linear images where isl_tile_info::physical_extent_B.height == 1,
1166 * this cleanly reduces to being the distance, in bytes, between vertically
1167 * adjacent surface elements.
1168 *
1169 * @see isl_tile_info::phys_extent_B;
1170 */
1171 uint32_t row_pitch;
1172
1173 /**
1174 * Pitch between physical array slices, in rows of surface elements.
1175 */
1176 uint32_t array_pitch_el_rows;
1177
1178 enum isl_array_pitch_span array_pitch_span;
1179
1180 /** Copy of isl_surf_init_info::usage. */
1181 isl_surf_usage_flags_t usage;
1182 };
1183
1184 struct isl_swizzle {
1185 enum isl_channel_select r:4;
1186 enum isl_channel_select g:4;
1187 enum isl_channel_select b:4;
1188 enum isl_channel_select a:4;
1189 };
1190
1191 #define ISL_SWIZZLE(R, G, B, A) ((struct isl_swizzle) { \
1192 .r = ISL_CHANNEL_SELECT_##R, \
1193 .g = ISL_CHANNEL_SELECT_##G, \
1194 .b = ISL_CHANNEL_SELECT_##B, \
1195 .a = ISL_CHANNEL_SELECT_##A, \
1196 })
1197
1198 #define ISL_SWIZZLE_IDENTITY ISL_SWIZZLE(RED, GREEN, BLUE, ALPHA)
1199
1200 struct isl_view {
1201 /**
1202 * Indicates the usage of the particular view
1203 *
1204 * Normally, this is one bit. However, for a cube map texture, it
1205 * should be ISL_SURF_USAGE_TEXTURE_BIT | ISL_SURF_USAGE_CUBE_BIT.
1206 */
1207 isl_surf_usage_flags_t usage;
1208
1209 /**
1210 * The format to use in the view
1211 *
1212 * This may differ from the format of the actual isl_surf but must have
1213 * the same block size.
1214 */
1215 enum isl_format format;
1216
1217 uint32_t base_level;
1218 uint32_t levels;
1219
1220 /**
1221 * Base array layer
1222 *
1223 * For cube maps, both base_array_layer and array_len should be
1224 * specified in terms of 2-D layers and must be a multiple of 6.
1225 *
1226 * 3-D textures are effectively treated as 2-D arrays when used as a
1227 * storage image or render target. If `usage` contains
1228 * ISL_SURF_USAGE_RENDER_TARGET_BIT or ISL_SURF_USAGE_STORAGE_BIT then
1229 * base_array_layer and array_len are applied. If the surface is only used
1230 * for texturing, they are ignored.
1231 */
1232 uint32_t base_array_layer;
1233
1234 /**
1235 * Array Length
1236 *
1237 * Indicates the number of array elements starting at Base Array Layer.
1238 */
1239 uint32_t array_len;
1240
1241 struct isl_swizzle swizzle;
1242 };
1243
1244 union isl_color_value {
1245 float f32[4];
1246 uint32_t u32[4];
1247 int32_t i32[4];
1248 };
1249
1250 struct isl_surf_fill_state_info {
1251 const struct isl_surf *surf;
1252 const struct isl_view *view;
1253
1254 /**
1255 * The address of the surface in GPU memory.
1256 */
1257 uint64_t address;
1258
1259 /**
1260 * The Memory Object Control state for the filled surface state.
1261 *
1262 * The exact format of this value depends on hardware generation.
1263 */
1264 uint32_t mocs;
1265
1266 /**
1267 * The auxilary surface or NULL if no auxilary surface is to be used.
1268 */
1269 const struct isl_surf *aux_surf;
1270 enum isl_aux_usage aux_usage;
1271 uint64_t aux_address;
1272
1273 /**
1274 * The clear color for this surface
1275 *
1276 * Valid values depend on hardware generation.
1277 */
1278 union isl_color_value clear_color;
1279
1280 /**
1281 * Surface write disables for gen4-5
1282 */
1283 isl_channel_mask_t write_disables;
1284
1285 /* Intra-tile offset */
1286 uint16_t x_offset_sa, y_offset_sa;
1287 };
1288
1289 struct isl_buffer_fill_state_info {
1290 /**
1291 * The address of the surface in GPU memory.
1292 */
1293 uint64_t address;
1294
1295 /**
1296 * The size of the buffer
1297 */
1298 uint64_t size;
1299
1300 /**
1301 * The Memory Object Control state for the filled surface state.
1302 *
1303 * The exact format of this value depends on hardware generation.
1304 */
1305 uint32_t mocs;
1306
1307 /**
1308 * The format to use in the surface state
1309 *
1310 * This may differ from the format of the actual isl_surf but have the
1311 * same block size.
1312 */
1313 enum isl_format format;
1314
1315 uint32_t stride;
1316 };
1317
1318 struct isl_depth_stencil_hiz_emit_info {
1319 /**
1320 * The depth surface
1321 */
1322 const struct isl_surf *depth_surf;
1323
1324 /**
1325 * The stencil surface
1326 *
1327 * If separate stencil is not available, this must point to the same
1328 * isl_surf as depth_surf.
1329 */
1330 const struct isl_surf *stencil_surf;
1331
1332 /**
1333 * The view into the depth and stencil surfaces.
1334 *
1335 * This view applies to both surfaces simultaneously.
1336 */
1337 const struct isl_view *view;
1338
1339 /**
1340 * The address of the depth surface in GPU memory
1341 */
1342 uint64_t depth_address;
1343
1344 /**
1345 * The address of the stencil surface in GPU memory
1346 *
1347 * If separate stencil is not available, this must have the same value as
1348 * depth_address.
1349 */
1350 uint64_t stencil_address;
1351
1352 /**
1353 * The Memory Object Control state for depth and stencil buffers
1354 *
1355 * Both depth and stencil will get the same MOCS value. The exact format
1356 * of this value depends on hardware generation.
1357 */
1358 uint32_t mocs;
1359
1360 /**
1361 * The HiZ surface or NULL if HiZ is disabled.
1362 */
1363 const struct isl_surf *hiz_surf;
1364 enum isl_aux_usage hiz_usage;
1365 uint64_t hiz_address;
1366
1367 /**
1368 * The depth clear value
1369 */
1370 float depth_clear_value;
1371 };
1372
1373 extern const struct isl_format_layout isl_format_layouts[];
1374
1375 void
1376 isl_device_init(struct isl_device *dev,
1377 const struct gen_device_info *info,
1378 bool has_bit6_swizzling);
1379
1380 isl_sample_count_mask_t ATTRIBUTE_CONST
1381 isl_device_get_sample_counts(struct isl_device *dev);
1382
1383 static inline const struct isl_format_layout * ATTRIBUTE_CONST
isl_format_get_layout(enum isl_format fmt)1384 isl_format_get_layout(enum isl_format fmt)
1385 {
1386 return &isl_format_layouts[fmt];
1387 }
1388
1389 static inline const char * ATTRIBUTE_CONST
isl_format_get_name(enum isl_format fmt)1390 isl_format_get_name(enum isl_format fmt)
1391 {
1392 return isl_format_layouts[fmt].name;
1393 }
1394
1395 bool isl_format_supports_rendering(const struct gen_device_info *devinfo,
1396 enum isl_format format);
1397 bool isl_format_supports_alpha_blending(const struct gen_device_info *devinfo,
1398 enum isl_format format);
1399 bool isl_format_supports_sampling(const struct gen_device_info *devinfo,
1400 enum isl_format format);
1401 bool isl_format_supports_filtering(const struct gen_device_info *devinfo,
1402 enum isl_format format);
1403 bool isl_format_supports_vertex_fetch(const struct gen_device_info *devinfo,
1404 enum isl_format format);
1405 bool isl_format_supports_typed_writes(const struct gen_device_info *devinfo,
1406 enum isl_format format);
1407 bool isl_format_supports_typed_reads(const struct gen_device_info *devinfo,
1408 enum isl_format format);
1409 bool isl_format_supports_ccs_d(const struct gen_device_info *devinfo,
1410 enum isl_format format);
1411 bool isl_format_supports_ccs_e(const struct gen_device_info *devinfo,
1412 enum isl_format format);
1413 bool isl_format_supports_multisampling(const struct gen_device_info *devinfo,
1414 enum isl_format format);
1415
1416 bool isl_formats_are_ccs_e_compatible(const struct gen_device_info *devinfo,
1417 enum isl_format format1,
1418 enum isl_format format2);
1419
1420 bool isl_format_has_unorm_channel(enum isl_format fmt) ATTRIBUTE_CONST;
1421 bool isl_format_has_snorm_channel(enum isl_format fmt) ATTRIBUTE_CONST;
1422 bool isl_format_has_ufloat_channel(enum isl_format fmt) ATTRIBUTE_CONST;
1423 bool isl_format_has_sfloat_channel(enum isl_format fmt) ATTRIBUTE_CONST;
1424 bool isl_format_has_uint_channel(enum isl_format fmt) ATTRIBUTE_CONST;
1425 bool isl_format_has_sint_channel(enum isl_format fmt) ATTRIBUTE_CONST;
1426
1427 static inline bool
isl_format_has_normalized_channel(enum isl_format fmt)1428 isl_format_has_normalized_channel(enum isl_format fmt)
1429 {
1430 return isl_format_has_unorm_channel(fmt) ||
1431 isl_format_has_snorm_channel(fmt);
1432 }
1433
1434 static inline bool
isl_format_has_float_channel(enum isl_format fmt)1435 isl_format_has_float_channel(enum isl_format fmt)
1436 {
1437 return isl_format_has_ufloat_channel(fmt) ||
1438 isl_format_has_sfloat_channel(fmt);
1439 }
1440
1441 static inline bool
isl_format_has_int_channel(enum isl_format fmt)1442 isl_format_has_int_channel(enum isl_format fmt)
1443 {
1444 return isl_format_has_uint_channel(fmt) ||
1445 isl_format_has_sint_channel(fmt);
1446 }
1447
1448 unsigned isl_format_get_num_channels(enum isl_format fmt);
1449
1450 uint32_t isl_format_get_depth_format(enum isl_format fmt, bool has_stencil);
1451
1452 static inline bool
isl_format_is_compressed(enum isl_format fmt)1453 isl_format_is_compressed(enum isl_format fmt)
1454 {
1455 const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
1456
1457 return fmtl->txc != ISL_TXC_NONE;
1458 }
1459
1460 static inline bool
isl_format_has_bc_compression(enum isl_format fmt)1461 isl_format_has_bc_compression(enum isl_format fmt)
1462 {
1463 switch (isl_format_get_layout(fmt)->txc) {
1464 case ISL_TXC_DXT1:
1465 case ISL_TXC_DXT3:
1466 case ISL_TXC_DXT5:
1467 return true;
1468 case ISL_TXC_NONE:
1469 case ISL_TXC_FXT1:
1470 case ISL_TXC_RGTC1:
1471 case ISL_TXC_RGTC2:
1472 case ISL_TXC_BPTC:
1473 case ISL_TXC_ETC1:
1474 case ISL_TXC_ETC2:
1475 case ISL_TXC_ASTC:
1476 return false;
1477
1478 case ISL_TXC_HIZ:
1479 case ISL_TXC_MCS:
1480 case ISL_TXC_CCS:
1481 unreachable("Should not be called on an aux surface");
1482 }
1483
1484 unreachable("bad texture compression mode");
1485 return false;
1486 }
1487
1488 static inline bool
isl_format_is_yuv(enum isl_format fmt)1489 isl_format_is_yuv(enum isl_format fmt)
1490 {
1491 const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
1492
1493 return fmtl->colorspace == ISL_COLORSPACE_YUV;
1494 }
1495
1496 static inline bool
isl_format_block_is_1x1x1(enum isl_format fmt)1497 isl_format_block_is_1x1x1(enum isl_format fmt)
1498 {
1499 const struct isl_format_layout *fmtl = isl_format_get_layout(fmt);
1500
1501 return fmtl->bw == 1 && fmtl->bh == 1 && fmtl->bd == 1;
1502 }
1503
1504 static inline bool
isl_format_is_srgb(enum isl_format fmt)1505 isl_format_is_srgb(enum isl_format fmt)
1506 {
1507 return isl_format_layouts[fmt].colorspace == ISL_COLORSPACE_SRGB;
1508 }
1509
1510 enum isl_format isl_format_srgb_to_linear(enum isl_format fmt);
1511
1512 static inline bool
isl_format_is_rgb(enum isl_format fmt)1513 isl_format_is_rgb(enum isl_format fmt)
1514 {
1515 if (isl_format_is_yuv(fmt))
1516 return false;
1517 return isl_format_layouts[fmt].channels.r.bits > 0 &&
1518 isl_format_layouts[fmt].channels.g.bits > 0 &&
1519 isl_format_layouts[fmt].channels.b.bits > 0 &&
1520 isl_format_layouts[fmt].channels.a.bits == 0;
1521 }
1522
1523 enum isl_format isl_format_rgb_to_rgba(enum isl_format rgb) ATTRIBUTE_CONST;
1524 enum isl_format isl_format_rgb_to_rgbx(enum isl_format rgb) ATTRIBUTE_CONST;
1525
1526 bool isl_is_storage_image_format(enum isl_format fmt);
1527
1528 enum isl_format
1529 isl_lower_storage_image_format(const struct gen_device_info *devinfo,
1530 enum isl_format fmt);
1531
1532 /* Returns true if this hardware supports typed load/store on a format with
1533 * the same size as the given format.
1534 */
1535 bool
1536 isl_has_matching_typed_storage_image_format(const struct gen_device_info *devinfo,
1537 enum isl_format fmt);
1538
1539 static inline bool
isl_tiling_is_any_y(enum isl_tiling tiling)1540 isl_tiling_is_any_y(enum isl_tiling tiling)
1541 {
1542 return (1u << tiling) & ISL_TILING_ANY_Y_MASK;
1543 }
1544
1545 static inline bool
isl_tiling_is_std_y(enum isl_tiling tiling)1546 isl_tiling_is_std_y(enum isl_tiling tiling)
1547 {
1548 return (1u << tiling) & ISL_TILING_STD_Y_MASK;
1549 }
1550
1551 uint32_t
1552 isl_tiling_to_i915_tiling(enum isl_tiling tiling);
1553
1554 enum isl_tiling
1555 isl_tiling_from_i915_tiling(uint32_t tiling);
1556
1557 const struct isl_drm_modifier_info * ATTRIBUTE_CONST
1558 isl_drm_modifier_get_info(uint64_t modifier);
1559
1560 static inline bool
isl_drm_modifier_has_aux(uint64_t modifier)1561 isl_drm_modifier_has_aux(uint64_t modifier)
1562 {
1563 return isl_drm_modifier_get_info(modifier)->aux_usage != ISL_AUX_USAGE_NONE;
1564 }
1565
1566 /** Returns the default isl_aux_state for the given modifier.
1567 *
1568 * All modified images are required to be kept out of the AUX_INVALID state
1569 * but they may or may not actually be compressed and may or may not have
1570 * clear color. This function returns the worst case aux_state that we need
1571 * to assume when getting a surface from another process or API.
1572 */
1573 static inline enum isl_aux_state
isl_drm_modifier_get_default_aux_state(uint64_t modifier)1574 isl_drm_modifier_get_default_aux_state(uint64_t modifier)
1575 {
1576 const struct isl_drm_modifier_info *mod_info =
1577 isl_drm_modifier_get_info(modifier);
1578
1579 if (!mod_info || mod_info->aux_usage == ISL_AUX_USAGE_NONE)
1580 return ISL_AUX_STATE_AUX_INVALID;
1581
1582 return mod_info->supports_clear_color ? ISL_AUX_STATE_COMPRESSED_CLEAR :
1583 ISL_AUX_STATE_COMPRESSED_NO_CLEAR;
1584 }
1585
1586 struct isl_extent2d ATTRIBUTE_CONST
1587 isl_get_interleaved_msaa_px_size_sa(uint32_t samples);
1588
1589 static inline bool
isl_surf_usage_is_display(isl_surf_usage_flags_t usage)1590 isl_surf_usage_is_display(isl_surf_usage_flags_t usage)
1591 {
1592 return usage & ISL_SURF_USAGE_DISPLAY_BIT;
1593 }
1594
1595 static inline bool
isl_surf_usage_is_depth(isl_surf_usage_flags_t usage)1596 isl_surf_usage_is_depth(isl_surf_usage_flags_t usage)
1597 {
1598 return usage & ISL_SURF_USAGE_DEPTH_BIT;
1599 }
1600
1601 static inline bool
isl_surf_usage_is_stencil(isl_surf_usage_flags_t usage)1602 isl_surf_usage_is_stencil(isl_surf_usage_flags_t usage)
1603 {
1604 return usage & ISL_SURF_USAGE_STENCIL_BIT;
1605 }
1606
1607 static inline bool
isl_surf_usage_is_depth_and_stencil(isl_surf_usage_flags_t usage)1608 isl_surf_usage_is_depth_and_stencil(isl_surf_usage_flags_t usage)
1609 {
1610 return (usage & ISL_SURF_USAGE_DEPTH_BIT) &&
1611 (usage & ISL_SURF_USAGE_STENCIL_BIT);
1612 }
1613
1614 static inline bool
isl_surf_usage_is_depth_or_stencil(isl_surf_usage_flags_t usage)1615 isl_surf_usage_is_depth_or_stencil(isl_surf_usage_flags_t usage)
1616 {
1617 return usage & (ISL_SURF_USAGE_DEPTH_BIT | ISL_SURF_USAGE_STENCIL_BIT);
1618 }
1619
1620 static inline bool
isl_surf_info_is_z16(const struct isl_surf_init_info * info)1621 isl_surf_info_is_z16(const struct isl_surf_init_info *info)
1622 {
1623 return (info->usage & ISL_SURF_USAGE_DEPTH_BIT) &&
1624 (info->format == ISL_FORMAT_R16_UNORM);
1625 }
1626
1627 static inline bool
isl_surf_info_is_z32_float(const struct isl_surf_init_info * info)1628 isl_surf_info_is_z32_float(const struct isl_surf_init_info *info)
1629 {
1630 return (info->usage & ISL_SURF_USAGE_DEPTH_BIT) &&
1631 (info->format == ISL_FORMAT_R32_FLOAT);
1632 }
1633
1634 static inline struct isl_extent2d
isl_extent2d(uint32_t width,uint32_t height)1635 isl_extent2d(uint32_t width, uint32_t height)
1636 {
1637 struct isl_extent2d e = { { 0 } };
1638
1639 e.width = width;
1640 e.height = height;
1641
1642 return e;
1643 }
1644
1645 static inline struct isl_extent3d
isl_extent3d(uint32_t width,uint32_t height,uint32_t depth)1646 isl_extent3d(uint32_t width, uint32_t height, uint32_t depth)
1647 {
1648 struct isl_extent3d e = { { 0 } };
1649
1650 e.width = width;
1651 e.height = height;
1652 e.depth = depth;
1653
1654 return e;
1655 }
1656
1657 static inline struct isl_extent4d
isl_extent4d(uint32_t width,uint32_t height,uint32_t depth,uint32_t array_len)1658 isl_extent4d(uint32_t width, uint32_t height, uint32_t depth,
1659 uint32_t array_len)
1660 {
1661 struct isl_extent4d e = { { 0 } };
1662
1663 e.width = width;
1664 e.height = height;
1665 e.depth = depth;
1666 e.array_len = array_len;
1667
1668 return e;
1669 }
1670
1671 bool isl_color_value_is_zero(union isl_color_value value,
1672 enum isl_format format);
1673
1674 bool isl_color_value_is_zero_one(union isl_color_value value,
1675 enum isl_format format);
1676
1677 #define isl_surf_init(dev, surf, ...) \
1678 isl_surf_init_s((dev), (surf), \
1679 &(struct isl_surf_init_info) { __VA_ARGS__ });
1680
1681 bool
1682 isl_surf_init_s(const struct isl_device *dev,
1683 struct isl_surf *surf,
1684 const struct isl_surf_init_info *restrict info);
1685
1686 void
1687 isl_surf_get_tile_info(const struct isl_surf *surf,
1688 struct isl_tile_info *tile_info);
1689
1690 bool
1691 isl_surf_get_hiz_surf(const struct isl_device *dev,
1692 const struct isl_surf *surf,
1693 struct isl_surf *hiz_surf);
1694
1695 bool
1696 isl_surf_get_mcs_surf(const struct isl_device *dev,
1697 const struct isl_surf *surf,
1698 struct isl_surf *mcs_surf);
1699
1700 bool
1701 isl_surf_get_ccs_surf(const struct isl_device *dev,
1702 const struct isl_surf *surf,
1703 struct isl_surf *ccs_surf,
1704 uint32_t row_pitch /**< Ignored if 0 */);
1705
1706 #define isl_surf_fill_state(dev, state, ...) \
1707 isl_surf_fill_state_s((dev), (state), \
1708 &(struct isl_surf_fill_state_info) { __VA_ARGS__ });
1709
1710 void
1711 isl_surf_fill_state_s(const struct isl_device *dev, void *state,
1712 const struct isl_surf_fill_state_info *restrict info);
1713
1714 #define isl_buffer_fill_state(dev, state, ...) \
1715 isl_buffer_fill_state_s((dev), (state), \
1716 &(struct isl_buffer_fill_state_info) { __VA_ARGS__ });
1717
1718 void
1719 isl_buffer_fill_state_s(const struct isl_device *dev, void *state,
1720 const struct isl_buffer_fill_state_info *restrict info);
1721
1722 void
1723 isl_null_fill_state(const struct isl_device *dev, void *state,
1724 struct isl_extent3d size);
1725
1726 #define isl_emit_depth_stencil_hiz(dev, batch, ...) \
1727 isl_emit_depth_stencil_hiz_s((dev), (batch), \
1728 &(struct isl_depth_stencil_hiz_emit_info) { __VA_ARGS__ })
1729
1730 void
1731 isl_emit_depth_stencil_hiz_s(const struct isl_device *dev, void *batch,
1732 const struct isl_depth_stencil_hiz_emit_info *restrict info);
1733
1734 void
1735 isl_surf_fill_image_param(const struct isl_device *dev,
1736 struct brw_image_param *param,
1737 const struct isl_surf *surf,
1738 const struct isl_view *view);
1739
1740 void
1741 isl_buffer_fill_image_param(const struct isl_device *dev,
1742 struct brw_image_param *param,
1743 enum isl_format format,
1744 uint64_t size);
1745
1746 /**
1747 * Alignment of the upper-left sample of each subimage, in units of surface
1748 * elements.
1749 */
1750 static inline struct isl_extent3d
isl_surf_get_image_alignment_el(const struct isl_surf * surf)1751 isl_surf_get_image_alignment_el(const struct isl_surf *surf)
1752 {
1753 return surf->image_alignment_el;
1754 }
1755
1756 /**
1757 * Alignment of the upper-left sample of each subimage, in units of surface
1758 * samples.
1759 */
1760 static inline struct isl_extent3d
isl_surf_get_image_alignment_sa(const struct isl_surf * surf)1761 isl_surf_get_image_alignment_sa(const struct isl_surf *surf)
1762 {
1763 const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
1764
1765 return isl_extent3d(fmtl->bw * surf->image_alignment_el.w,
1766 fmtl->bh * surf->image_alignment_el.h,
1767 fmtl->bd * surf->image_alignment_el.d);
1768 }
1769
1770 /**
1771 * Pitch between vertically adjacent surface elements, in bytes.
1772 */
1773 static inline uint32_t
isl_surf_get_row_pitch(const struct isl_surf * surf)1774 isl_surf_get_row_pitch(const struct isl_surf *surf)
1775 {
1776 return surf->row_pitch;
1777 }
1778
1779 /**
1780 * Pitch between vertically adjacent surface elements, in units of surface elements.
1781 */
1782 static inline uint32_t
isl_surf_get_row_pitch_el(const struct isl_surf * surf)1783 isl_surf_get_row_pitch_el(const struct isl_surf *surf)
1784 {
1785 const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
1786
1787 assert(surf->row_pitch % (fmtl->bpb / 8) == 0);
1788 return surf->row_pitch / (fmtl->bpb / 8);
1789 }
1790
1791 /**
1792 * Pitch between physical array slices, in rows of surface elements.
1793 */
1794 static inline uint32_t
isl_surf_get_array_pitch_el_rows(const struct isl_surf * surf)1795 isl_surf_get_array_pitch_el_rows(const struct isl_surf *surf)
1796 {
1797 return surf->array_pitch_el_rows;
1798 }
1799
1800 /**
1801 * Pitch between physical array slices, in units of surface elements.
1802 */
1803 static inline uint32_t
isl_surf_get_array_pitch_el(const struct isl_surf * surf)1804 isl_surf_get_array_pitch_el(const struct isl_surf *surf)
1805 {
1806 return isl_surf_get_array_pitch_el_rows(surf) *
1807 isl_surf_get_row_pitch_el(surf);
1808 }
1809
1810 /**
1811 * Pitch between physical array slices, in rows of surface samples.
1812 */
1813 static inline uint32_t
isl_surf_get_array_pitch_sa_rows(const struct isl_surf * surf)1814 isl_surf_get_array_pitch_sa_rows(const struct isl_surf *surf)
1815 {
1816 const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
1817 return fmtl->bh * isl_surf_get_array_pitch_el_rows(surf);
1818 }
1819
1820 /**
1821 * Pitch between physical array slices, in bytes.
1822 */
1823 static inline uint32_t
isl_surf_get_array_pitch(const struct isl_surf * surf)1824 isl_surf_get_array_pitch(const struct isl_surf *surf)
1825 {
1826 return isl_surf_get_array_pitch_sa_rows(surf) * surf->row_pitch;
1827 }
1828
1829 /**
1830 * Calculate the offset, in units of surface samples, to a subimage in the
1831 * surface.
1832 *
1833 * @invariant level < surface levels
1834 * @invariant logical_array_layer < logical array length of surface
1835 * @invariant logical_z_offset_px < logical depth of surface at level
1836 */
1837 void
1838 isl_surf_get_image_offset_sa(const struct isl_surf *surf,
1839 uint32_t level,
1840 uint32_t logical_array_layer,
1841 uint32_t logical_z_offset_px,
1842 uint32_t *x_offset_sa,
1843 uint32_t *y_offset_sa);
1844
1845 /**
1846 * Calculate the offset, in units of surface elements, to a subimage in the
1847 * surface.
1848 *
1849 * @invariant level < surface levels
1850 * @invariant logical_array_layer < logical array length of surface
1851 * @invariant logical_z_offset_px < logical depth of surface at level
1852 */
1853 void
1854 isl_surf_get_image_offset_el(const struct isl_surf *surf,
1855 uint32_t level,
1856 uint32_t logical_array_layer,
1857 uint32_t logical_z_offset_px,
1858 uint32_t *x_offset_el,
1859 uint32_t *y_offset_el);
1860
1861 /**
1862 * Calculate the offset, in bytes and intratile surface samples, to a
1863 * subimage in the surface.
1864 *
1865 * This is equivalent to calling isl_surf_get_image_offset_el, passing the
1866 * result to isl_tiling_get_intratile_offset_el, and converting the tile
1867 * offsets to samples.
1868 *
1869 * @invariant level < surface levels
1870 * @invariant logical_array_layer < logical array length of surface
1871 * @invariant logical_z_offset_px < logical depth of surface at level
1872 */
1873 void
1874 isl_surf_get_image_offset_B_tile_sa(const struct isl_surf *surf,
1875 uint32_t level,
1876 uint32_t logical_array_layer,
1877 uint32_t logical_z_offset_px,
1878 uint32_t *offset_B,
1879 uint32_t *x_offset_sa,
1880 uint32_t *y_offset_sa);
1881
1882 /**
1883 * Create an isl_surf that represents a particular subimage in the surface.
1884 *
1885 * The newly created surface will have a single miplevel and array slice. The
1886 * surface lives at the returned byte and intratile offsets, in samples.
1887 *
1888 * It is safe to call this function with surf == image_surf.
1889 *
1890 * @invariant level < surface levels
1891 * @invariant logical_array_layer < logical array length of surface
1892 * @invariant logical_z_offset_px < logical depth of surface at level
1893 */
1894 void
1895 isl_surf_get_image_surf(const struct isl_device *dev,
1896 const struct isl_surf *surf,
1897 uint32_t level,
1898 uint32_t logical_array_layer,
1899 uint32_t logical_z_offset_px,
1900 struct isl_surf *image_surf,
1901 uint32_t *offset_B,
1902 uint32_t *x_offset_sa,
1903 uint32_t *y_offset_sa);
1904
1905 /**
1906 * @brief Calculate the intratile offsets to a surface.
1907 *
1908 * In @a base_address_offset return the offset from the base of the surface to
1909 * the base address of the first tile of the subimage. In @a x_offset_B and
1910 * @a y_offset_rows, return the offset, in units of bytes and rows, from the
1911 * tile's base to the subimage's first surface element. The x and y offsets
1912 * are intratile offsets; that is, they do not exceed the boundary of the
1913 * surface's tiling format.
1914 */
1915 void
1916 isl_tiling_get_intratile_offset_el(enum isl_tiling tiling,
1917 uint32_t bpb,
1918 uint32_t row_pitch,
1919 uint32_t total_x_offset_el,
1920 uint32_t total_y_offset_el,
1921 uint32_t *base_address_offset,
1922 uint32_t *x_offset_el,
1923 uint32_t *y_offset_el);
1924
1925 static inline void
isl_tiling_get_intratile_offset_sa(enum isl_tiling tiling,enum isl_format format,uint32_t row_pitch,uint32_t total_x_offset_sa,uint32_t total_y_offset_sa,uint32_t * base_address_offset,uint32_t * x_offset_sa,uint32_t * y_offset_sa)1926 isl_tiling_get_intratile_offset_sa(enum isl_tiling tiling,
1927 enum isl_format format,
1928 uint32_t row_pitch,
1929 uint32_t total_x_offset_sa,
1930 uint32_t total_y_offset_sa,
1931 uint32_t *base_address_offset,
1932 uint32_t *x_offset_sa,
1933 uint32_t *y_offset_sa)
1934 {
1935 const struct isl_format_layout *fmtl = isl_format_get_layout(format);
1936
1937 /* For computing the intratile offsets, we actually want a strange unit
1938 * which is samples for multisampled surfaces but elements for compressed
1939 * surfaces.
1940 */
1941 assert(total_x_offset_sa % fmtl->bw == 0);
1942 assert(total_y_offset_sa % fmtl->bh == 0);
1943 const uint32_t total_x_offset = total_x_offset_sa / fmtl->bw;
1944 const uint32_t total_y_offset = total_y_offset_sa / fmtl->bh;
1945
1946 isl_tiling_get_intratile_offset_el(tiling, fmtl->bpb, row_pitch,
1947 total_x_offset, total_y_offset,
1948 base_address_offset,
1949 x_offset_sa, y_offset_sa);
1950 *x_offset_sa *= fmtl->bw;
1951 *y_offset_sa *= fmtl->bh;
1952 }
1953
1954 /**
1955 * @brief Get value of 3DSTATE_DEPTH_BUFFER.SurfaceFormat
1956 *
1957 * @pre surf->usage has ISL_SURF_USAGE_DEPTH_BIT
1958 * @pre surf->format must be a valid format for depth surfaces
1959 */
1960 uint32_t
1961 isl_surf_get_depth_format(const struct isl_device *dev,
1962 const struct isl_surf *surf);
1963
1964 #ifdef __cplusplus
1965 }
1966 #endif
1967
1968 #endif /* ISL_H */
1969