summaryrefslogtreecommitdiffstats
path: root/src/intel/isl/isl_storage_image.c
blob: 2b5b5cdbaed7058849006afb26c02762f6c7b86a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
/*
 * Copyright 2015 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 */

#include "isl_priv.h"
#include "brw_compiler.h"

bool
isl_is_storage_image_format(enum isl_format format)
{
   /* XXX: Maybe we should put this in the CSV? */

   switch (format) {
   case ISL_FORMAT_R32G32B32A32_UINT:
   case ISL_FORMAT_R32G32B32A32_SINT:
   case ISL_FORMAT_R32G32B32A32_FLOAT:
   case ISL_FORMAT_R32_UINT:
   case ISL_FORMAT_R32_SINT:
   case ISL_FORMAT_R32_FLOAT:
   case ISL_FORMAT_R16G16B16A16_UINT:
   case ISL_FORMAT_R16G16B16A16_SINT:
   case ISL_FORMAT_R16G16B16A16_FLOAT:
   case ISL_FORMAT_R32G32_UINT:
   case ISL_FORMAT_R32G32_SINT:
   case ISL_FORMAT_R32G32_FLOAT:
   case ISL_FORMAT_R8G8B8A8_UINT:
   case ISL_FORMAT_R8G8B8A8_SINT:
   case ISL_FORMAT_R16G16_UINT:
   case ISL_FORMAT_R16G16_SINT:
   case ISL_FORMAT_R16G16_FLOAT:
   case ISL_FORMAT_R8G8_UINT:
   case ISL_FORMAT_R8G8_SINT:
   case ISL_FORMAT_R16_UINT:
   case ISL_FORMAT_R16_FLOAT:
   case ISL_FORMAT_R16_SINT:
   case ISL_FORMAT_R8_UINT:
   case ISL_FORMAT_R8_SINT:
   case ISL_FORMAT_R10G10B10A2_UINT:
   case ISL_FORMAT_R10G10B10A2_UNORM:
   case ISL_FORMAT_R11G11B10_FLOAT:
   case ISL_FORMAT_R16G16B16A16_UNORM:
   case ISL_FORMAT_R16G16B16A16_SNORM:
   case ISL_FORMAT_R8G8B8A8_UNORM:
   case ISL_FORMAT_R8G8B8A8_SNORM:
   case ISL_FORMAT_R16G16_UNORM:
   case ISL_FORMAT_R16G16_SNORM:
   case ISL_FORMAT_R8G8_UNORM:
   case ISL_FORMAT_R8G8_SNORM:
   case ISL_FORMAT_R16_UNORM:
   case ISL_FORMAT_R16_SNORM:
   case ISL_FORMAT_R8_UNORM:
   case ISL_FORMAT_R8_SNORM:
      return true;
   default:
      return false;
   }
}

enum isl_format
isl_lower_storage_image_format(const struct isl_device *dev,
                               enum isl_format format)
{
   switch (format) {
   /* These are never lowered.  Up to BDW we'll have to fall back to untyped
    * surface access for 128bpp formats.
    */
   case ISL_FORMAT_R32G32B32A32_UINT:
   case ISL_FORMAT_R32G32B32A32_SINT:
   case ISL_FORMAT_R32G32B32A32_FLOAT:
   case ISL_FORMAT_R32_UINT:
   case ISL_FORMAT_R32_SINT:
   case ISL_FORMAT_R32_FLOAT:
      return format;

   /* From HSW to BDW the only 64bpp format supported for typed access is
    * RGBA_UINT16.  IVB falls back to untyped.
    */
   case ISL_FORMAT_R16G16B16A16_UINT:
   case ISL_FORMAT_R16G16B16A16_SINT:
   case ISL_FORMAT_R16G16B16A16_FLOAT:
   case ISL_FORMAT_R32G32_UINT:
   case ISL_FORMAT_R32G32_SINT:
   case ISL_FORMAT_R32G32_FLOAT:
      return (ISL_DEV_GEN(dev) >= 9 ? format :
              ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R16G16B16A16_UINT :
              ISL_FORMAT_R32G32_UINT);

   /* Up to BDW no SINT or FLOAT formats of less than 32 bits per component
    * are supported.  IVB doesn't support formats with more than one component
    * for typed access.  For 8 and 16 bpp formats IVB relies on the
    * undocumented behavior that typed reads from R_UINT8 and R_UINT16
    * surfaces actually do a 32-bit misaligned read.  The alternative would be
    * to use two surface state entries with different formats for each image,
    * one for reading (using R_UINT32) and another one for writing (using
    * R_UINT8 or R_UINT16), but that would complicate the shaders we generate
    * even more.
    */
   case ISL_FORMAT_R8G8B8A8_UINT:
   case ISL_FORMAT_R8G8B8A8_SINT:
      return (ISL_DEV_GEN(dev) >= 9 ? format :
              ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R8G8B8A8_UINT : ISL_FORMAT_R32_UINT);

   case ISL_FORMAT_R16G16_UINT:
   case ISL_FORMAT_R16G16_SINT:
   case ISL_FORMAT_R16G16_FLOAT:
      return (ISL_DEV_GEN(dev) >= 9 ? format :
              ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R16G16_UINT : ISL_FORMAT_R32_UINT);

   case ISL_FORMAT_R8G8_UINT:
   case ISL_FORMAT_R8G8_SINT:
      return (ISL_DEV_GEN(dev) >= 9 ? format :
              ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R8G8_UINT : ISL_FORMAT_R16_UINT);

   case ISL_FORMAT_R16_UINT:
   case ISL_FORMAT_R16_FLOAT:
   case ISL_FORMAT_R16_SINT:
      return (ISL_DEV_GEN(dev) >= 9 ? format : ISL_FORMAT_R16_UINT);

   case ISL_FORMAT_R8_UINT:
   case ISL_FORMAT_R8_SINT:
      return (ISL_DEV_GEN(dev) >= 9 ? format : ISL_FORMAT_R8_UINT);

   /* Neither the 2/10/10/10 nor the 11/11/10 packed formats are supported
    * by the hardware.
    */
   case ISL_FORMAT_R10G10B10A2_UINT:
   case ISL_FORMAT_R10G10B10A2_UNORM:
   case ISL_FORMAT_R11G11B10_FLOAT:
      return ISL_FORMAT_R32_UINT;

   /* No normalized fixed-point formats are supported by the hardware. */
   case ISL_FORMAT_R16G16B16A16_UNORM:
   case ISL_FORMAT_R16G16B16A16_SNORM:
      return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R16G16B16A16_UINT :
              ISL_FORMAT_R32G32_UINT);

   case ISL_FORMAT_R8G8B8A8_UNORM:
   case ISL_FORMAT_R8G8B8A8_SNORM:
      return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R8G8B8A8_UINT : ISL_FORMAT_R32_UINT);

   case ISL_FORMAT_R16G16_UNORM:
   case ISL_FORMAT_R16G16_SNORM:
      return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R16G16_UINT : ISL_FORMAT_R32_UINT);

   case ISL_FORMAT_R8G8_UNORM:
   case ISL_FORMAT_R8G8_SNORM:
      return (ISL_DEV_GEN(dev) >= 8 || dev->info->is_haswell ?
              ISL_FORMAT_R8G8_UINT : ISL_FORMAT_R16_UINT);

   case ISL_FORMAT_R16_UNORM:
   case ISL_FORMAT_R16_SNORM:
      return ISL_FORMAT_R16_UINT;

   case ISL_FORMAT_R8_UNORM:
   case ISL_FORMAT_R8_SNORM:
      return ISL_FORMAT_R8_UINT;

   default:
      assert(!"Unknown image format");
      return ISL_FORMAT_UNSUPPORTED;
   }
}

static const struct brw_image_param image_param_defaults = {
   /* Set the swizzling shifts to all-ones to effectively disable
    * swizzling -- See emit_address_calculation() in
    * brw_fs_surface_builder.cpp for a more detailed explanation of
    * these parameters.
    */
   .swizzling = { 0xff, 0xff },
};

void
isl_surf_fill_image_param(const struct isl_device *dev,
                          struct brw_image_param *param,
                          const struct isl_surf *surf,
                          const struct isl_view *view)
{
   *param = image_param_defaults;

   param->size[0] = isl_minify(surf->logical_level0_px.w, view->base_level);
   param->size[1] = isl_minify(surf->logical_level0_px.h, view->base_level);
   if (surf->dim == ISL_SURF_DIM_3D) {
      param->size[2] = isl_minify(surf->logical_level0_px.d, view->base_level);
   } else {
      param->size[2] = surf->logical_level0_px.array_len -
                       view->base_array_layer;
   }

   isl_surf_get_image_offset_el(surf, view->base_level, view->base_array_layer,
                                0, &param->offset[0],  &param->offset[1]);

   const int cpp = isl_format_get_layout(surf->format)->bs;
   param->stride[0] = cpp;
   param->stride[1] = surf->row_pitch / cpp;

   const struct isl_extent3d image_align_sa =
      isl_surf_get_image_alignment_sa(surf);
   if (ISL_DEV_GEN(dev) < 9 && surf->dim == ISL_SURF_DIM_3D) {
      param->stride[2] = isl_align_npot(param->size[0], image_align_sa.w);
      param->stride[3] = isl_align_npot(param->size[1], image_align_sa.h);
   } else {
      param->stride[2] = 0;
      param->stride[3] = isl_surf_get_array_pitch_el_rows(surf);
   }

   switch (surf->tiling) {
   case ISL_TILING_LINEAR:
      /* image_param_defaults is good enough */
      break;

   case ISL_TILING_X:
      /* An X tile is a rectangular block of 512x8 bytes. */
      param->tiling[0] = isl_log2u(512 / cpp);
      param->tiling[1] = isl_log2u(8);

      if (dev->has_bit6_swizzling) {
         /* Right shifts required to swizzle bits 9 and 10 of the memory
          * address with bit 6.
          */
         param->swizzling[0] = 3;
         param->swizzling[1] = 4;
      }
      break;

   case ISL_TILING_Y0:
      /* The layout of a Y-tiled surface in memory isn't really fundamentally
       * different to the layout of an X-tiled surface, we simply pretend that
       * the surface is broken up in a number of smaller 16Bx32 tiles, each
       * one arranged in X-major order just like is the case for X-tiling.
       */
      param->tiling[0] = isl_log2u(16 / cpp);
      param->tiling[1] = isl_log2u(32);

      if (dev->has_bit6_swizzling) {
         /* Right shift required to swizzle bit 9 of the memory address with
          * bit 6.
          */
         param->swizzling[0] = 3;
         param->swizzling[1] = 0xff;
      }
      break;

   default:
      assert(!"Unhandled storage image tiling");
   }

   /* 3D textures are arranged in 2D in memory with 2^lod slices per row.  The
    * address calculation algorithm (emit_address_calculation() in
    * brw_fs_surface_builder.cpp) handles this as a sort of tiling with
    * modulus equal to the LOD.
    */
   param->tiling[2] = (ISL_DEV_GEN(dev) < 9 && surf->dim == ISL_SURF_DIM_3D ?
                       view->base_level : 0);
}

void
isl_buffer_fill_image_param(const struct isl_device *dev,
                            struct brw_image_param *param,
                            enum isl_format format,
                            uint64_t size)
{
   *param = image_param_defaults;

   param->stride[0] = isl_format_layouts[format].bs;
   param->size[0] = size / param->stride[0];
}