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 /* indexed by cpp: */
  33 static const struct {
  34         unsigned pitchalign;
  35         unsigned heightalign;
  36 } tile_alignment[] = {
  37         [1]  = { 128, 32 },
  38         [2]  = { 128, 16 },
  39         [3]  = { 128, 16 },
  40         [4]  = {  64, 16 },
  41         [8]  = {  64, 16 },
  42         [12] = {  64, 16 },
  43         [16] = {  64, 16 },
  44 };
  45
  46 void
  47 fdl5_layout(struct fdl_layout *layout,
  48                 enum pipe_format format, uint32_t nr_samples,
  49                 uint32_t width0, uint32_t height0, uint32_t depth0,
  50                 uint32_t mip_levels, uint32_t array_size, bool is_3d)
  51 {
  52         const struct util_format_description *format_desc =
  53                 util_format_description(format);
  54
  55         assert(nr_samples > 0);
  56         layout->width0 = width0;
  57         layout->height0 = height0;
  58         layout->depth0 = depth0;
  59
  60         layout->cpp = util_format_get_blocksize(format);
  61         layout->cpp *= nr_samples;
  62         layout->cpp_shift = ffs(layout->cpp) - 1;
  63
  64         layout->format = format;
  65         layout->nr_samples = nr_samples;
  66         layout->layer_first = !is_3d;
  67
  68         uint32_t pitchalign;
  69         uint32_t heightalign;
  70         uint32_t width = width0;
  71         uint32_t height = height0;
  72         uint32_t depth = depth0;
  73         /* in layer_first layout, the level (slice) contains just one
  74          * layer (since in fact the layer contains the slices)
  75          */
  76         uint32_t layers_in_level = layout->layer_first ? 1 : array_size;
  77
  78         heightalign = tile_alignment[layout->cpp].heightalign;
  79
  80         for (uint32_t level = 0; level < mip_levels; level++) {
  81                 struct fdl_slice *slice = &layout->slices[level];
  82                 uint32_t tile_mode = fdl_tile_mode(layout, level);
  83                 uint32_t aligned_height = height;
  84                 uint32_t blocks;
  85
  86                 if (tile_mode) {
  87                         pitchalign = tile_alignment[layout->cpp].pitchalign;
  88                         aligned_height = align(aligned_height, heightalign);
  89                 } else {
  90                         pitchalign = 64;
  91
  92                         /* The blits used for mem<->gmem work at a granularity of
  93                          * 32x32, which can cause faults due to over-fetch on the
  94                          * last level.  The simple solution is to over-allocate a
  95                          * bit the last level to ensure any over-fetch is harmless.
  96                          * The pitch is already sufficiently aligned, but height
  97                          * may not be:
  98                          */
  99                         if (level == mip_levels - 1)
 100                                 aligned_height = align(aligned_height, 32);
 101                 }
 102
 103                 unsigned pitch_pixels;
 104                 if (format_desc->layout == UTIL_FORMAT_LAYOUT_ASTC)
 105                         pitch_pixels =
 106                                 util_align_npot(width, pitchalign * util_format_get_blockwidth(format));
 107                 else
 108                         pitch_pixels = align(width, pitchalign);
 109
 110                 slice->offset = layout->size;
 111                 blocks = util_format_get_nblocks(format, pitch_pixels, aligned_height);
 112                 slice->pitch = util_format_get_nblocksx(format, pitch_pixels) *
 113                         layout->cpp;
 114
 115                 const int alignment = is_3d ? 4096 : 1;
 116
 117                 /* 1d array and 2d array textures must all have the same layer size
 118                  * for each miplevel on a3xx. 3d textures can have different layer
 119                  * sizes for high levels, but the hw auto-sizer is buggy (or at least
 120                  * different than what this code does), so as soon as the layer size
 121                  * range gets into range, we stop reducing it.
 122                  */
 123                 if (is_3d && (
 124                                         level == 1 ||
 125                                         (level > 1 && layout->slices[level - 1].size0 > 0xf000)))
 126                         slice->size0 = align(blocks * layout->cpp, alignment);
 127                 else if (level == 0 || layout->layer_first || alignment == 1)
 128                         slice->size0 = align(blocks * layout->cpp, alignment);
 129                 else
 130                         slice->size0 = layout->slices[level - 1].size0;
 131
 132                 layout->size += slice->size0 * depth * layers_in_level;
 133
 134                 width = u_minify(width, 1);
 135                 height = u_minify(height, 1);
 136                 depth = u_minify(depth, 1);
 137         }
 138 }
 139