tu: Use an input for the layer when lowering input attachments
[mesa.git] / src / freedreno / fdl / fd5_layout.c
1 /*
2 * Copyright (C) 2018 Rob Clark <robclark@freedesktop.org>
3 * Copyright © 2018-2019 Google, Inc.
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
14 * Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 *
24 * Authors:
25 * Rob Clark <robclark@freedesktop.org>
26 */
27
28 #include <stdio.h>
29
30 #include "freedreno_layout.h"
31
32 void
33 fdl5_layout(struct fdl_layout *layout,
34 enum pipe_format format, uint32_t nr_samples,
35 uint32_t width0, uint32_t height0, uint32_t depth0,
36 uint32_t mip_levels, uint32_t array_size, bool is_3d)
37 {
38 assert(nr_samples > 0);
39 layout->width0 = width0;
40 layout->height0 = height0;
41 layout->depth0 = depth0;
42
43 layout->cpp = util_format_get_blocksize(format);
44 layout->cpp *= nr_samples;
45 layout->cpp_shift = ffs(layout->cpp) - 1;
46
47 layout->format = format;
48 layout->nr_samples = nr_samples;
49 layout->layer_first = !is_3d;
50
51 uint32_t heightalign = layout->cpp == 1 ? 32 : 16;
52 /* in layer_first layout, the level (slice) contains just one
53 * layer (since in fact the layer contains the slices)
54 */
55 uint32_t layers_in_level = layout->layer_first ? 1 : array_size;
56
57 /* use 128 pixel alignment for cpp=1 and cpp=2 */
58 if (layout->cpp < 4 && layout->tile_mode)
59 fdl_set_pitchalign(layout, fdl_cpp_shift(layout) + 7);
60 else
61 fdl_set_pitchalign(layout, fdl_cpp_shift(layout) + 6);
62
63 for (uint32_t level = 0; level < mip_levels; level++) {
64 struct fdl_slice *slice = &layout->slices[level];
65 uint32_t tile_mode = fdl_tile_mode(layout, level);
66 uint32_t pitch = fdl_pitch(layout, level);
67 uint32_t nblocksy = util_format_get_nblocksy(format, u_minify(height0, level));
68
69 if (tile_mode) {
70 nblocksy = align(nblocksy, heightalign);
71 } else {
72 /* The blits used for mem<->gmem work at a granularity of
73 * 32x32, which can cause faults due to over-fetch on the
74 * last level. The simple solution is to over-allocate a
75 * bit the last level to ensure any over-fetch is harmless.
76 * The pitch is already sufficiently aligned, but height
77 * may not be:
78 */
79 if (level == mip_levels - 1)
80 nblocksy = align(nblocksy, 32);
81 }
82
83 slice->offset = layout->size;
84
85 const int alignment = is_3d ? 4096 : 1;
86
87 /* 1d array and 2d array textures must all have the same layer size
88 * for each miplevel on a3xx. 3d textures can have different layer
89 * sizes for high levels, but the hw auto-sizer is buggy (or at least
90 * different than what this code does), so as soon as the layer size
91 * range gets into range, we stop reducing it.
92 */
93 if (is_3d && (
94 level == 1 ||
95 (level > 1 && layout->slices[level - 1].size0 > 0xf000)))
96 slice->size0 = align(nblocksy * pitch, alignment);
97 else if (level == 0 || layout->layer_first || alignment == 1)
98 slice->size0 = align(nblocksy * pitch, alignment);
99 else
100 slice->size0 = layout->slices[level - 1].size0;
101
102 layout->size += slice->size0 * u_minify(depth0, level) * layers_in_level;
103 }
104 }
105