1/*
2 * Copyright (c) 2012-2013 Luc Verhaegen <libv@skynet.be>
3 * Copyright (c) 2018 Alyssa Rosenzweig <alyssa@rosenzweig.io>
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, sub license,
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
13 * next paragraph) shall be included in all copies or substantial portions
14 * of the 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 NON-INFRINGEMENT. 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
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
23 */
24
25#include <stdio.h>
26#include "pan_swizzle.h"
27#include "pan_allocate.h"
28
29/* Space a group of 4-bits out. For instance, 0x7 -- that is, 0b111 -- would
30 * become 0b10101 */
31
32static inline int
33space_bits_4(int i)
34{
35        return ((i & 0x8) << 3) |
36               ((i & 0x4) << 2) |
37               ((i & 0x2) << 1) |
38               ((i & 0x1) << 0);
39}
40
41/* Generate lookup table for the space filler curve. Note this is a 1:1
42 * mapping, just with bits twiddled around. */
43
44uint32_t space_filler[16][16];
45uint32_t space_filler_packed4[16][4];
46
47void
48panfrost_generate_space_filler_indices()
49{
50        for (int y = 0; y < 16; ++y) {
51                for (int x = 0; x < 16; ++x) {
52                        space_filler[y][x] =
53                                space_bits_4(y ^ x) | (space_bits_4(y) << 1);
54                }
55
56                for (int q = 0; q < 4; ++q) {
57                        space_filler_packed4[y][q] =
58                                (space_filler[y][(q * 4) + 0] << 0) |
59                                (space_filler[y][(q * 4) + 1] << 8) |
60                                (space_filler[y][(q * 4) + 2] << 16) |
61                                (space_filler[y][(q * 4) + 3] << 24);
62                }
63        }
64}
65
66static void
67swizzle_bpp1_align16(int width, int height, int source_stride, int block_pitch,
68                     const uint8_t *pixels,
69                     uint8_t *ldest)
70{
71        for (int y = 0; y < height; ++y) {
72                {
73                        int block_y = y & ~(0x0f);
74                        int rem_y = y & 0x0f;
75                        uint8_t *block_start_s = ldest + (block_y * block_pitch);
76                        const uint8_t *source_start = pixels + (y * source_stride);
77                        const uint8_t *source_end = source_start + width;
78
79                        /* Operate on blocks of 16 pixels to minimise bookkeeping */
80
81                        for (; source_start < source_end; block_start_s += 16 * 16, source_start += 16) {
82                                const uint32_t *src_32 = (const uint32_t *) source_start;
83
84                                for (int q = 0; q < 4; ++q) {
85                                        uint32_t src = src_32[q];
86                                        uint32_t spaced = space_filler_packed4[rem_y][q];
87                                        uint16_t *bs = (uint16_t *) block_start_s;
88
89                                        int spacedA = (spaced >> 0) & 0xFF;
90                                        int spacedB = (spaced >> 16) & 0xFF;
91
92                                        bs[spacedA >> 1] = (src >> 0) & 0xFFFF;
93                                        bs[spacedB >> 1] = (src >> 16) & 0xFFFF;
94                                }
95                        }
96                }
97
98                ++y;
99
100                if (y >= height)
101                        break;
102
103                {
104                        int block_y = y & ~(0x0f);
105                        int rem_y = y & 0x0f;
106                        uint8_t *block_start_s = ldest + (block_y * block_pitch);
107                        const uint8_t *source_start = pixels + (y * source_stride);
108                        const uint8_t *source_end = source_start + width;
109
110                        /* Operate on blocks of 16 pixels to minimise bookkeeping */
111
112                        for (; source_start < source_end; block_start_s += 16 * 16, source_start += 16) {
113                                const uint32_t *src_32 = (const uint32_t *) source_start;
114
115                                for (int q = 0; q < 4; ++q) {
116                                        uint32_t src = src_32[q];
117                                        uint32_t spaced = space_filler_packed4[rem_y][q];
118
119                                        block_start_s[(spaced >> 0) & 0xFF] = (src >> 0) & 0xFF;
120                                        block_start_s[(spaced >> 8) & 0xFF] = (src >> 8) & 0xFF;
121
122                                        block_start_s[(spaced >> 16) & 0xFF] = (src >> 16) & 0xFF;
123                                        block_start_s[(spaced >> 24) & 0xFF] = (src >> 24) & 0xFF;
124                                }
125                        }
126                }
127
128        }
129}
130
131static void
132swizzle_bpp4_align16(int width, int height, int source_stride, int block_pitch,
133                     const uint32_t *pixels,
134                     uint32_t *ldest)
135{
136        for (int y = 0; y < height; ++y) {
137                int block_y = y & ~(0x0f);
138                int rem_y = y & 0x0f;
139                uint32_t *block_start_s = ldest + (block_y * block_pitch);
140                const uint32_t *source_start = pixels + (y * source_stride);
141                const uint32_t *source_end = source_start + width;
142
143                /* Operate on blocks of 16 pixels to minimise bookkeeping */
144
145                for (; source_start < source_end; block_start_s += 16 * 16, source_start += 16) {
146                        for (int j = 0; j < 16; ++j)
147                                block_start_s[space_filler[rem_y][j]] = source_start[j];
148                }
149        }
150}
151
152void
153panfrost_texture_swizzle(unsigned off_x,
154                         unsigned off_y,
155                         int width, int height, int bytes_per_pixel, int dest_width,
156                         const uint8_t *pixels,
157                         uint8_t *ldest)
158{
159        /* Calculate maximum size, overestimating a bit */
160        int block_pitch = ALIGN(dest_width, 16) >> 4;
161
162        /* Strides must be tight, since we're only ever called indirectly */
163        int source_stride = width * bytes_per_pixel;
164
165        /* Use fast path if available */
166        if (!(off_x || off_y) && (width == dest_width)) {
167                if (bytes_per_pixel == 4 && (ALIGN(width, 16) == width)) {
168                        swizzle_bpp4_align16(width, height, source_stride >> 2, (block_pitch * 256 >> 4), (const uint32_t *) pixels, (uint32_t *) ldest);
169                        return;
170                } else if (bytes_per_pixel == 1 && (ALIGN(width, 16) == width)) {
171                        swizzle_bpp1_align16(width, height, source_stride, (block_pitch * 256 >> 4), pixels, (uint8_t *) ldest);
172                        return;
173                }
174        }
175
176        /* Otherwise, default back on generic path */
177
178        for (int y = 0; y < height; ++y) {
179                int block_y = (y + off_y) >> 4;
180                int rem_y = (y + off_y) & 0x0F;
181                int block_start_s = block_y * block_pitch * 256;
182                int source_start = y * source_stride;
183
184                for (int x = 0; x < width; ++x) {
185                        int block_x_s = ((x + off_x) >> 4) * 256;
186                        int rem_x = (x + off_x) & 0x0F;
187
188                        int index = space_filler[rem_y][rem_x];
189                        const uint8_t *source = &pixels[source_start + bytes_per_pixel * x];
190                        uint8_t *dest = ldest + bytes_per_pixel * (block_start_s + block_x_s + index);
191
192                        for (int b = 0; b < bytes_per_pixel; ++b)
193                                dest[b] = source[b];
194                }
195        }
196}
197