isl: Add a helper for filling a buffer surface state
[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
51 static const uint8_t isl_to_gen_valign[] = {
52 [4] = VALIGN4,
53 [8] = VALIGN8,
54 [16] = VALIGN16,
55 };
56 #else
57 static const uint8_t isl_to_gen_halign[] = {
58 [4] = HALIGN_4,
59 [8] = HALIGN_8,
60 };
61
62 static const uint8_t isl_to_gen_valign[] = {
63 [2] = VALIGN_2,
64 [4] = VALIGN_4,
65 };
66 #endif
67
68 #if GEN_GEN >= 8
69 static const uint8_t isl_to_gen_tiling[] = {
70 [ISL_TILING_LINEAR] = LINEAR,
71 [ISL_TILING_X] = XMAJOR,
72 [ISL_TILING_Y0] = YMAJOR,
73 [ISL_TILING_Yf] = YMAJOR,
74 [ISL_TILING_Ys] = YMAJOR,
75 [ISL_TILING_W] = WMAJOR,
76 };
77 #endif
78
79 #if GEN_GEN >= 8
80 static const uint32_t isl_to_gen_multisample_layout[] = {
81 [ISL_MSAA_LAYOUT_NONE] = MSS,
82 [ISL_MSAA_LAYOUT_INTERLEAVED] = DEPTH_STENCIL,
83 [ISL_MSAA_LAYOUT_ARRAY] = MSS,
84 };
85 #else
86 static const uint32_t isl_to_gen_multisample_layout[] = {
87 [ISL_MSAA_LAYOUT_NONE] = MSFMT_MSS,
88 [ISL_MSAA_LAYOUT_INTERLEAVED] = MSFMT_DEPTH_STENCIL,
89 [ISL_MSAA_LAYOUT_ARRAY] = MSFMT_MSS,
90 };
91 #endif
92
93 static const uint8_t
94 get_surftype(enum isl_surf_dim dim, isl_surf_usage_flags_t usage)
95 {
96 switch (dim) {
97 default:
98 unreachable("bad isl_surf_dim");
99 case ISL_SURF_DIM_1D:
100 assert(!(usage & ISL_SURF_USAGE_CUBE_BIT));
101 return SURFTYPE_1D;
102 case ISL_SURF_DIM_2D:
103 if (usage & ISL_SURF_USAGE_STORAGE_BIT) {
104 /* Storage images are always plain 2-D, not cube */
105 return SURFTYPE_2D;
106 } else if (usage & ISL_SURF_USAGE_CUBE_BIT) {
107 return SURFTYPE_CUBE;
108 } else {
109 return SURFTYPE_2D;
110 }
111 case ISL_SURF_DIM_3D:
112 assert(!(usage & ISL_SURF_USAGE_CUBE_BIT));
113 return SURFTYPE_3D;
114 }
115 }
116
117 /**
118 * Get the values to pack into RENDER_SUFFACE_STATE.SurfaceHorizontalAlignment
119 * and SurfaceVerticalAlignment.
120 */
121 static void
122 get_halign_valign(const struct isl_surf *surf,
123 uint32_t *halign, uint32_t *valign)
124 {
125 if (GEN_GEN >= 9) {
126 if (isl_tiling_is_std_y(surf->tiling) ||
127 surf->dim_layout == ISL_DIM_LAYOUT_GEN9_1D) {
128 /* The hardware ignores the alignment values. Anyway, the surface's
129 * true alignment is likely outside the enum range of HALIGN* and
130 * VALIGN*.
131 */
132 *halign = 0;
133 *valign = 0;
134 } else {
135 /* In Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in units
136 * of surface elements (not pixels nor samples). For compressed formats,
137 * a "surface element" is defined as a compression block. For example,
138 * if SurfaceVerticalAlignment is VALIGN_4 and SurfaceFormat is an ETC2
139 * format (ETC2 has a block height of 4), then the vertical alignment is
140 * 4 compression blocks or, equivalently, 16 pixels.
141 */
142 struct isl_extent3d image_align_el
143 = isl_surf_get_image_alignment_el(surf);
144
145 *halign = isl_to_gen_halign[image_align_el.width];
146 *valign = isl_to_gen_valign[image_align_el.height];
147 }
148 } else {
149 /* Pre-Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in
150 * units of surface samples. For example, if SurfaceVerticalAlignment
151 * is VALIGN_4 and the surface is singlesampled, then for any surface
152 * format (compressed or not) the vertical alignment is
153 * 4 pixels.
154 */
155 struct isl_extent3d image_align_sa
156 = isl_surf_get_image_alignment_sa(surf);
157
158 *halign = isl_to_gen_halign[image_align_sa.width];
159 *valign = isl_to_gen_valign[image_align_sa.height];
160 }
161 }
162
163 #if GEN_GEN >= 8
164 static uint32_t
165 get_qpitch(const struct isl_surf *surf)
166 {
167 switch (surf->dim) {
168 default:
169 assert(!"Bad isl_surf_dim");
170 case ISL_SURF_DIM_1D:
171 if (GEN_GEN >= 9) {
172 /* QPitch is usually expressed as rows of surface elements (where
173 * a surface element is an compression block or a single surface
174 * sample). Skylake 1D is an outlier.
175 *
176 * From the Skylake BSpec >> Memory Views >> Common Surface
177 * Formats >> Surface Layout and Tiling >> 1D Surfaces:
178 *
179 * Surface QPitch specifies the distance in pixels between array
180 * slices.
181 */
182 return isl_surf_get_array_pitch_el(surf);
183 } else {
184 return isl_surf_get_array_pitch_el_rows(surf);
185 }
186 case ISL_SURF_DIM_2D:
187 case ISL_SURF_DIM_3D:
188 if (GEN_GEN >= 9) {
189 return isl_surf_get_array_pitch_el_rows(surf);
190 } else {
191 /* From the Broadwell PRM for RENDER_SURFACE_STATE.QPitch
192 *
193 * "This field must be set to an integer multiple of the Surface
194 * Vertical Alignment. For compressed textures (BC*, FXT1,
195 * ETC*, and EAC* Surface Formats), this field is in units of
196 * rows in the uncompressed surface, and must be set to an
197 * integer multiple of the vertical alignment parameter "j"
198 * defined in the Common Surface Formats section."
199 */
200 return isl_surf_get_array_pitch_sa_rows(surf);
201 }
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 uint32_t halign, valign;
211 get_halign_valign(info->surf, &halign, &valign);
212
213 struct GENX(RENDER_SURFACE_STATE) s = {
214 .SurfaceType = get_surftype(info->surf->dim, info->view->usage),
215 .SurfaceArray = info->surf->phys_level0_sa.array_len > 1,
216 .SurfaceVerticalAlignment = valign,
217 .SurfaceHorizontalAlignment = halign,
218
219 #if GEN_GEN >= 8
220 .TileMode = isl_to_gen_tiling[info->surf->tiling],
221 #else
222 .TiledSurface = info->surf->tiling != ISL_TILING_LINEAR,
223 .TileWalk = info->surf->tiling == ISL_TILING_X ? TILEWALK_XMAJOR :
224 TILEWALK_YMAJOR,
225 #endif
226
227 .VerticalLineStride = 0,
228 .VerticalLineStrideOffset = 0,
229
230 #if GEN_GEN >= 8
231 .SamplerL2BypassModeDisable = true,
232 #endif
233
234 #if GEN_GEN >= 8
235 .RenderCacheReadWriteMode = WriteOnlyCache,
236 #else
237 .RenderCacheReadWriteMode = 0,
238 #endif
239
240 #if GEN_GEN >= 8
241 .CubeFaceEnablePositiveZ = 1,
242 .CubeFaceEnableNegativeZ = 1,
243 .CubeFaceEnablePositiveY = 1,
244 .CubeFaceEnableNegativeY = 1,
245 .CubeFaceEnablePositiveX = 1,
246 .CubeFaceEnableNegativeX = 1,
247 #else
248 .CubeFaceEnables = 0x3f,
249 #endif
250
251 #if GEN_GEN >= 8
252 .SurfaceQPitch = get_qpitch(info->surf) >> 2,
253 #endif
254
255 .Width = info->level0_extent_px.width - 1,
256 .Height = info->level0_extent_px.height - 1,
257 .Depth = 0, /* TEMPLATE */
258
259 .SurfacePitch = info->surf->row_pitch - 1,
260 .RenderTargetViewExtent = 0, /* TEMPLATE */
261 .MinimumArrayElement = 0, /* TEMPLATE */
262
263 .MultisampledSurfaceStorageFormat =
264 isl_to_gen_multisample_layout[info->surf->msaa_layout],
265 .NumberofMultisamples = ffs(info->surf->samples) - 1,
266 .MultisamplePositionPaletteIndex = 0, /* UNUSED */
267
268 .XOffset = 0,
269 .YOffset = 0,
270
271 .ResourceMinLOD = 0.0,
272
273 .MIPCountLOD = 0, /* TEMPLATE */
274 .SurfaceMinLOD = 0, /* TEMPLATE */
275
276 #if (GEN_GEN >= 8 || GEN_IS_HASWELL)
277 .ShaderChannelSelectRed = info->view->channel_select[0],
278 .ShaderChannelSelectGreen = info->view->channel_select[1],
279 .ShaderChannelSelectBlue = info->view->channel_select[2],
280 .ShaderChannelSelectAlpha = info->view->channel_select[3],
281 #endif
282
283 .SurfaceBaseAddress = info->address,
284 .MOCS = info->mocs,
285
286 #if GEN_GEN >= 8
287 .AuxiliarySurfaceMode = AUX_NONE,
288 #else
289 .MCSEnable = false,
290 #endif
291 };
292
293 if (info->view->usage & ISL_SURF_USAGE_STORAGE_BIT) {
294 s.SurfaceFormat = isl_lower_storage_image_format(dev, info->view->format);
295 } else {
296 s.SurfaceFormat = info->view->format;
297 }
298
299 switch (s.SurfaceType) {
300 case SURFTYPE_1D:
301 case SURFTYPE_2D:
302 s.MinimumArrayElement = info->view->base_array_layer;
303
304 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
305 *
306 * For SURFTYPE_1D, 2D, and CUBE: The range of this field is reduced
307 * by one for each increase from zero of Minimum Array Element. For
308 * example, if Minimum Array Element is set to 1024 on a 2D surface,
309 * the range of this field is reduced to [0,1023].
310 *
311 * In other words, 'Depth' is the number of array layers.
312 */
313 s.Depth = info->view->array_len - 1;
314
315 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
316 *
317 * For Render Target and Typed Dataport 1D and 2D Surfaces:
318 * This field must be set to the same value as the Depth field.
319 */
320 s.RenderTargetViewExtent = s.Depth;
321 break;
322 case SURFTYPE_CUBE:
323 s.MinimumArrayElement = info->view->base_array_layer;
324 /* Same as SURFTYPE_2D, but divided by 6 */
325 s.Depth = info->view->array_len / 6 - 1;
326 s.RenderTargetViewExtent = s.Depth;
327 break;
328 case SURFTYPE_3D:
329 s.MinimumArrayElement = info->view->base_array_layer;
330
331 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
332 *
333 * If the volume texture is MIP-mapped, this field specifies the
334 * depth of the base MIP level.
335 */
336 s.Depth = info->level0_extent_px.depth - 1;
337
338 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
339 *
340 * For Render Target and Typed Dataport 3D Surfaces: This field
341 * indicates the extent of the accessible 'R' coordinates minus 1 on
342 * the LOD currently being rendered to.
343 */
344 s.RenderTargetViewExtent = info->level0_extent_px.depth - 1;
345 break;
346 default:
347 unreachable(!"bad SurfaceType");
348 }
349
350 if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) {
351 /* For render target surfaces, the hardware interprets field
352 * MIPCount/LOD as LOD. The Broadwell PRM says:
353 *
354 * MIPCountLOD defines the LOD that will be rendered into.
355 * SurfaceMinLOD is ignored.
356 */
357 s.MIPCountLOD = info->view->base_level;
358 s.SurfaceMinLOD = 0;
359 } else {
360 /* For non render target surfaces, the hardware interprets field
361 * MIPCount/LOD as MIPCount. The range of levels accessible by the
362 * sampler engine is [SurfaceMinLOD, SurfaceMinLOD + MIPCountLOD].
363 */
364 s.SurfaceMinLOD = info->view->base_level;
365 s.MIPCountLOD = MAX(info->view->levels, 1) - 1;
366 }
367
368 #if 0
369 if (GEN_GEN == 8) {
370 if (isl_format_is_integer(info->view->format)) {
371 for (unsigned i = 0; i < 4; i++) {
372 assert(info->clear_color.u32[i] == 0 ||
373 info->clear_color.u32[i] == 1);
374 }
375 } else {
376 for (unsigned i = 0; i < 4; i++) {
377 assert(info->clear_color.f32[i] == 0.0f ||
378 info->clear_color.f32[i] == 1.0f);
379 }
380 }
381 s.RedClearColor = info->clear_color.u32[0] != 0,
382 s.GreenClearColor = info->clear_color.u32[1] != 0,
383 s.BlueClearColor = info->clear_color.u32[2] != 0,
384 s.AlphaClearColor = info->clear_color.u32[3] != 0,
385 } else {
386 .RedClearColor = info->clear_color.u32[0],
387 .GreenClearColor = info->clear_color.u32[1],
388 .BlueClearColor = info->clear_color.u32[2],
389 .AlphaClearColor = info->clear_color.u32[3],
390 }
391 #endif
392
393 GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &s);
394 }
395
396 void
397 isl_genX(buffer_fill_state_s)(void *state,
398 const struct isl_buffer_fill_state_info *restrict info)
399 {
400 uint32_t num_elements = info->size / info->stride;
401
402 struct GENX(RENDER_SURFACE_STATE) surface_state = {
403 .SurfaceType = SURFTYPE_BUFFER,
404 .SurfaceArray = false,
405 .SurfaceFormat = info->format,
406 .SurfaceVerticalAlignment = isl_to_gen_valign[4],
407 .SurfaceHorizontalAlignment = isl_to_gen_halign[4],
408 .Height = ((num_elements - 1) >> 7) & 0x3fff,
409 .Width = (num_elements - 1) & 0x7f,
410 .Depth = ((num_elements - 1) >> 21) & 0x3f,
411 .SurfacePitch = info->stride - 1,
412 .NumberofMultisamples = MULTISAMPLECOUNT_1,
413
414 #if (GEN_GEN >= 8)
415 .TileMode = LINEAR,
416 #else
417 .TiledSurface = false,
418 #endif
419
420 #if (GEN_GEN >= 8)
421 .SamplerL2BypassModeDisable = true,
422 .RenderCacheReadWriteMode = WriteOnlyCache,
423 #else
424 .RenderCacheReadWriteMode = 0,
425 #endif
426
427 .MOCS = info->mocs,
428
429 #if (GEN_GEN >= 8 || GEN_IS_HASWELL)
430 .ShaderChannelSelectRed = SCS_RED,
431 .ShaderChannelSelectGreen = SCS_GREEN,
432 .ShaderChannelSelectBlue = SCS_BLUE,
433 .ShaderChannelSelectAlpha = SCS_ALPHA,
434 #endif
435 .SurfaceBaseAddress = info->address,
436 };
437
438 GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &surface_state);
439 }