fbarchard@google.com | ca41005 | 2012-10-14 06:01:19 +0000 | [diff] [blame] | 1 | /* |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 2 | * Copyright (c) 2012 The LibYuv project authors. All Rights Reserved. |
fbarchard@google.com | ca41005 | 2012-10-14 06:01:19 +0000 | [diff] [blame] | 3 | * |
| 4 | * Use of this source code is governed by a BSD-style license |
| 5 | * that can be found in the LICENSE file in the root of the source |
| 6 | * tree. An additional intellectual property rights grant can be found |
| 7 | * in the file PATENTS. All contributing project authors may |
| 8 | * be found in the AUTHORS file in the root of the source tree. |
| 9 | */ |
| 10 | |
| 11 | #include "libyuv/row.h" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 12 | |
fbarchard@google.com | ca41005 | 2012-10-14 06:01:19 +0000 | [diff] [blame] | 13 | #ifdef __cplusplus |
| 14 | namespace libyuv { |
| 15 | extern "C" { |
| 16 | #endif |
| 17 | |
| 18 | #if !defined(YUV_DISABLE_ASM) && defined(__mips__) |
fbarchard@google.com | 834b748 | 2012-12-11 17:59:26 +0000 | [diff] [blame] | 19 | #ifdef HAS_COPYROW_MIPS |
fbarchard@google.com | 6c1b2d3 | 2012-10-26 22:49:18 +0000 | [diff] [blame] | 20 | extern "C" void memcpy_MIPS(uint8* dst, const uint8* src, int count); |
| 21 | void CopyRow_MIPS(const uint8* src, uint8* dst, int count) { |
| 22 | memcpy_MIPS(dst, src, count); |
| 23 | } |
fbarchard@google.com | 834b748 | 2012-12-11 17:59:26 +0000 | [diff] [blame] | 24 | #endif // HAS_COPYROW_MIPS |
| 25 | #endif // __mips__ |
fbarchard@google.com | 6c1b2d3 | 2012-10-26 22:49:18 +0000 | [diff] [blame] | 26 | |
fbarchard@google.com | 834b748 | 2012-12-11 17:59:26 +0000 | [diff] [blame] | 27 | // MIPS DSPR2 functions |
| 28 | #if !defined(YUV_DISABLE_ASM) && defined(__mips_dsp) && (__mips_dsp_rev >= 2) |
fbarchard@google.com | f08ac6b | 2012-11-15 00:21:14 +0000 | [diff] [blame] | 29 | void SplitUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, |
| 30 | int width) { |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 31 | __asm__ __volatile__ ( |
| 32 | ".set push \n" |
| 33 | ".set noreorder \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 34 | "srl $t4, %[width], 4 \n" // multiplies of 16 |
| 35 | "blez $t4, 2f \n" |
| 36 | " andi %[width], %[width], 0xf \n" // residual |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 37 | |
fbarchard@google.com | db694ed | 2012-10-17 21:54:04 +0000 | [diff] [blame] | 38 | "1: \n" |
| 39 | "addiu $t4, $t4, -1 \n" |
| 40 | "lw $t0, 0(%[src_uv]) \n" // V1 | U1 | V0 | U0 |
| 41 | "lw $t1, 4(%[src_uv]) \n" // V3 | U3 | V2 | U2 |
| 42 | "lw $t2, 8(%[src_uv]) \n" // V5 | U5 | V4 | U4 |
| 43 | "lw $t3, 12(%[src_uv]) \n" // V7 | U7 | V6 | U6 |
| 44 | "lw $t5, 16(%[src_uv]) \n" // V9 | U9 | V8 | U8 |
| 45 | "lw $t6, 20(%[src_uv]) \n" // V11 | U11 | V10 | U10 |
| 46 | "lw $t7, 24(%[src_uv]) \n" // V13 | U13 | V12 | U12 |
| 47 | "lw $t8, 28(%[src_uv]) \n" // V15 | U15 | V14 | U14 |
| 48 | "addiu %[src_uv], %[src_uv], 32 \n" |
| 49 | "precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0 |
| 50 | "precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0 |
| 51 | "precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4 |
| 52 | "precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4 |
| 53 | "precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8 |
| 54 | "precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8 |
| 55 | "precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12 |
| 56 | "precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12 |
| 57 | "sw $t9, 0(%[dst_v]) \n" |
| 58 | "sw $t0, 0(%[dst_u]) \n" |
| 59 | "sw $t1, 4(%[dst_v]) \n" |
| 60 | "sw $t2, 4(%[dst_u]) \n" |
| 61 | "sw $t3, 8(%[dst_v]) \n" |
| 62 | "sw $t5, 8(%[dst_u]) \n" |
| 63 | "sw $t6, 12(%[dst_v]) \n" |
| 64 | "sw $t7, 12(%[dst_u]) \n" |
| 65 | "addiu %[dst_v], %[dst_v], 16 \n" |
| 66 | "bgtz $t4, 1b \n" |
| 67 | " addiu %[dst_u], %[dst_u], 16 \n" |
| 68 | |
| 69 | "beqz %[width], 3f \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 70 | " nop \n" |
| 71 | |
fbarchard@google.com | db694ed | 2012-10-17 21:54:04 +0000 | [diff] [blame] | 72 | "2: \n" |
| 73 | "lbu $t0, 0(%[src_uv]) \n" |
| 74 | "lbu $t1, 1(%[src_uv]) \n" |
| 75 | "addiu %[src_uv], %[src_uv], 2 \n" |
| 76 | "addiu %[width], %[width], -1 \n" |
| 77 | "sb $t0, 0(%[dst_u]) \n" |
| 78 | "sb $t1, 0(%[dst_v]) \n" |
| 79 | "addiu %[dst_u], %[dst_u], 1 \n" |
| 80 | "bgtz %[width], 2b \n" |
| 81 | " addiu %[dst_v], %[dst_v], 1 \n" |
| 82 | |
| 83 | "3: \n" |
| 84 | ".set pop \n" |
| 85 | : [src_uv] "+r" (src_uv), |
| 86 | [width] "+r" (width), |
| 87 | [dst_u] "+r" (dst_u), |
| 88 | [dst_v] "+r" (dst_v) |
| 89 | : |
| 90 | : "t0", "t1", "t2", "t3", |
| 91 | "t4", "t5", "t6", "t7", "t8", "t9" |
| 92 | ); |
| 93 | } |
| 94 | |
fbarchard@google.com | f08ac6b | 2012-11-15 00:21:14 +0000 | [diff] [blame] | 95 | void SplitUVRow_Unaligned_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, |
| 96 | uint8* dst_v, int width) { |
fbarchard@google.com | db694ed | 2012-10-17 21:54:04 +0000 | [diff] [blame] | 97 | __asm__ __volatile__ ( |
| 98 | ".set push \n" |
| 99 | ".set noreorder \n" |
| 100 | "srl $t4, %[width], 4 \n" // multiplies of 16 |
| 101 | "blez $t4, 2f \n" |
| 102 | " andi %[width], %[width], 0xf \n" // residual |
| 103 | |
| 104 | "1: \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 105 | "addiu $t4, $t4, -1 \n" |
| 106 | "lwr $t0, 0(%[src_uv]) \n" |
| 107 | "lwl $t0, 3(%[src_uv]) \n" // V1 | U1 | V0 | U0 |
| 108 | "lwr $t1, 4(%[src_uv]) \n" |
| 109 | "lwl $t1, 7(%[src_uv]) \n" // V3 | U3 | V2 | U2 |
| 110 | "lwr $t2, 8(%[src_uv]) \n" |
| 111 | "lwl $t2, 11(%[src_uv]) \n" // V5 | U5 | V4 | U4 |
| 112 | "lwr $t3, 12(%[src_uv]) \n" |
| 113 | "lwl $t3, 15(%[src_uv]) \n" // V7 | U7 | V6 | U6 |
| 114 | "lwr $t5, 16(%[src_uv]) \n" |
| 115 | "lwl $t5, 19(%[src_uv]) \n" // V9 | U9 | V8 | U8 |
| 116 | "lwr $t6, 20(%[src_uv]) \n" |
| 117 | "lwl $t6, 23(%[src_uv]) \n" // V11 | U11 | V10 | U10 |
| 118 | "lwr $t7, 24(%[src_uv]) \n" |
| 119 | "lwl $t7, 27(%[src_uv]) \n" // V13 | U13 | V12 | U12 |
| 120 | "lwr $t8, 28(%[src_uv]) \n" |
| 121 | "lwl $t8, 31(%[src_uv]) \n" // V15 | U15 | V14 | U14 |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 122 | "precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0 |
| 123 | "precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0 |
| 124 | "precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4 |
| 125 | "precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4 |
| 126 | "precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8 |
| 127 | "precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8 |
| 128 | "precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12 |
| 129 | "precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12 |
| 130 | "addiu %[src_uv], %[src_uv], 32 \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 131 | "swr $t9, 0(%[dst_v]) \n" |
| 132 | "swl $t9, 3(%[dst_v]) \n" |
| 133 | "swr $t0, 0(%[dst_u]) \n" |
| 134 | "swl $t0, 3(%[dst_u]) \n" |
| 135 | "swr $t1, 4(%[dst_v]) \n" |
| 136 | "swl $t1, 7(%[dst_v]) \n" |
| 137 | "swr $t2, 4(%[dst_u]) \n" |
| 138 | "swl $t2, 7(%[dst_u]) \n" |
| 139 | "swr $t3, 8(%[dst_v]) \n" |
| 140 | "swl $t3, 11(%[dst_v]) \n" |
| 141 | "swr $t5, 8(%[dst_u]) \n" |
| 142 | "swl $t5, 11(%[dst_u]) \n" |
| 143 | "swr $t6, 12(%[dst_v]) \n" |
| 144 | "swl $t6, 15(%[dst_v]) \n" |
| 145 | "swr $t7, 12(%[dst_u]) \n" |
| 146 | "swl $t7, 15(%[dst_u]) \n" |
| 147 | "addiu %[dst_u], %[dst_u], 16 \n" |
| 148 | "bgtz $t4, 1b \n" |
| 149 | " addiu %[dst_v], %[dst_v], 16 \n" |
| 150 | |
| 151 | "beqz %[width], 3f \n" |
| 152 | " nop \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 153 | |
fbarchard@google.com | db694ed | 2012-10-17 21:54:04 +0000 | [diff] [blame] | 154 | "2: \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 155 | "lbu $t0, 0(%[src_uv]) \n" |
| 156 | "lbu $t1, 1(%[src_uv]) \n" |
| 157 | "addiu %[src_uv], %[src_uv], 2 \n" |
| 158 | "addiu %[width], %[width], -1 \n" |
| 159 | "sb $t0, 0(%[dst_u]) \n" |
| 160 | "sb $t1, 0(%[dst_v]) \n" |
| 161 | "addiu %[dst_u], %[dst_u], 1 \n" |
| 162 | "bgtz %[width], 2b \n" |
| 163 | " addiu %[dst_v], %[dst_v], 1 \n" |
| 164 | |
fbarchard@google.com | db694ed | 2012-10-17 21:54:04 +0000 | [diff] [blame] | 165 | "3: \n" |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 166 | ".set pop \n" |
| 167 | : [src_uv] "+r" (src_uv), |
| 168 | [width] "+r" (width), |
| 169 | [dst_u] "+r" (dst_u), |
| 170 | [dst_v] "+r" (dst_v) |
| 171 | : |
| 172 | : "t0", "t1", "t2", "t3", |
| 173 | "t4", "t5", "t6", "t7", "t8", "t9" |
fbarchard@google.com | ca41005 | 2012-10-14 06:01:19 +0000 | [diff] [blame] | 174 | ); |
| 175 | } |
fbarchard@google.com | bb6bddc | 2012-10-14 06:41:17 +0000 | [diff] [blame] | 176 | |
fbarchard@google.com | 6c1b2d3 | 2012-10-26 22:49:18 +0000 | [diff] [blame] | 177 | void MirrorRow_MIPS_DSPR2(const uint8* src, uint8* dst, int width) { |
| 178 | __asm__ __volatile__ ( |
| 179 | ".set push \n" |
| 180 | ".set noreorder \n" |
| 181 | |
| 182 | "srl $t4, %[width], 4 \n" // multiplies of 16 |
| 183 | "andi $t5, %[width], 0xf \n" |
| 184 | "blez $t4, 2f \n" |
| 185 | " addu %[src], %[src], %[width] \n" // src += width |
| 186 | |
| 187 | "1: \n" |
| 188 | "lw $t0, -16(%[src]) \n" // |3|2|1|0| |
| 189 | "lw $t1, -12(%[src]) \n" // |7|6|5|4| |
| 190 | "lw $t2, -8(%[src]) \n" // |11|10|9|8| |
| 191 | "lw $t3, -4(%[src]) \n" // |15|14|13|12| |
| 192 | "wsbh $t0, $t0 \n" // |2|3|0|1| |
| 193 | "wsbh $t1, $t1 \n" // |6|7|4|5| |
| 194 | "wsbh $t2, $t2 \n" // |10|11|8|9| |
| 195 | "wsbh $t3, $t3 \n" // |14|15|12|13| |
| 196 | "rotr $t0, $t0, 16 \n" // |0|1|2|3| |
| 197 | "rotr $t1, $t1, 16 \n" // |4|5|6|7| |
| 198 | "rotr $t2, $t2, 16 \n" // |8|9|10|11| |
| 199 | "rotr $t3, $t3, 16 \n" // |12|13|14|15| |
| 200 | "addiu %[src], %[src], -16 \n" |
| 201 | "addiu $t4, $t4, -1 \n" |
| 202 | "sw $t3, 0(%[dst]) \n" // |15|14|13|12| |
| 203 | "sw $t2, 4(%[dst]) \n" // |11|10|9|8| |
| 204 | "sw $t1, 8(%[dst]) \n" // |7|6|5|4| |
| 205 | "sw $t0, 12(%[dst]) \n" // |3|2|1|0| |
| 206 | "bgtz $t4, 1b \n" |
| 207 | " addiu %[dst], %[dst], 16 \n" |
| 208 | "beqz $t5, 3f \n" |
| 209 | " nop \n" |
| 210 | |
| 211 | "2: \n" |
| 212 | "lbu $t0, -1(%[src]) \n" |
| 213 | "addiu $t5, $t5, -1 \n" |
| 214 | "addiu %[src], %[src], -1 \n" |
| 215 | "sb $t0, 0(%[dst]) \n" |
| 216 | "bgez $t5, 2b \n" |
| 217 | " addiu %[dst], %[dst], 1 \n" |
| 218 | |
| 219 | "3: \n" |
| 220 | ".set pop \n" |
| 221 | : [src] "+r" (src), [dst] "+r" (dst) |
| 222 | : [width] "r" (width) |
| 223 | : "t0", "t1", "t2", "t3", "t4", "t5" |
| 224 | ); |
| 225 | } |
fbarchard@google.com | 6c1b2d3 | 2012-10-26 22:49:18 +0000 | [diff] [blame] | 226 | |
fbarchard@google.com | bdf7cb5 | 2012-11-05 23:40:11 +0000 | [diff] [blame] | 227 | void MirrorUVRow_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, |
fbarchard@google.com | 6c1b2d3 | 2012-10-26 22:49:18 +0000 | [diff] [blame] | 228 | int width) { |
| 229 | int x = 0; |
| 230 | int y = 0; |
| 231 | __asm__ __volatile__ ( |
| 232 | ".set push \n" |
| 233 | ".set noreorder \n" |
| 234 | |
| 235 | "addu $t4, %[width], %[width] \n" |
| 236 | "srl %[x], %[width], 4 \n" |
| 237 | "andi %[y], %[width], 0xf \n" |
| 238 | "blez %[x], 2f \n" |
| 239 | " addu %[src_uv], %[src_uv], $t4 \n" |
| 240 | |
| 241 | "1: \n" |
| 242 | "lw $t0, -32(%[src_uv]) \n" // |3|2|1|0| |
| 243 | "lw $t1, -28(%[src_uv]) \n" // |7|6|5|4| |
| 244 | "lw $t2, -24(%[src_uv]) \n" // |11|10|9|8| |
| 245 | "lw $t3, -20(%[src_uv]) \n" // |15|14|13|12| |
| 246 | "lw $t4, -16(%[src_uv]) \n" // |19|18|17|16| |
| 247 | "lw $t6, -12(%[src_uv]) \n" // |23|22|21|20| |
| 248 | "lw $t7, -8(%[src_uv]) \n" // |27|26|25|24| |
| 249 | "lw $t8, -4(%[src_uv]) \n" // |31|30|29|28| |
| 250 | |
| 251 | "rotr $t0, $t0, 16 \n" // |1|0|3|2| |
| 252 | "rotr $t1, $t1, 16 \n" // |5|4|7|6| |
| 253 | "rotr $t2, $t2, 16 \n" // |9|8|11|10| |
| 254 | "rotr $t3, $t3, 16 \n" // |13|12|15|14| |
| 255 | "rotr $t4, $t4, 16 \n" // |17|16|19|18| |
| 256 | "rotr $t6, $t6, 16 \n" // |21|20|23|22| |
| 257 | "rotr $t7, $t7, 16 \n" // |25|24|27|26| |
| 258 | "rotr $t8, $t8, 16 \n" // |29|28|31|30| |
| 259 | "precr.qb.ph $t9, $t0, $t1 \n" // |0|2|4|6| |
| 260 | "precrq.qb.ph $t5, $t0, $t1 \n" // |1|3|5|7| |
| 261 | "precr.qb.ph $t0, $t2, $t3 \n" // |8|10|12|14| |
| 262 | "precrq.qb.ph $t1, $t2, $t3 \n" // |9|11|13|15| |
| 263 | "precr.qb.ph $t2, $t4, $t6 \n" // |16|18|20|22| |
| 264 | "precrq.qb.ph $t3, $t4, $t6 \n" // |17|19|21|23| |
| 265 | "precr.qb.ph $t4, $t7, $t8 \n" // |24|26|28|30| |
| 266 | "precrq.qb.ph $t6, $t7, $t8 \n" // |25|27|29|31| |
| 267 | "addiu %[src_uv], %[src_uv], -32 \n" |
| 268 | "addiu %[x], %[x], -1 \n" |
| 269 | "swr $t4, 0(%[dst_u]) \n" |
| 270 | "swl $t4, 3(%[dst_u]) \n" // |30|28|26|24| |
| 271 | "swr $t6, 0(%[dst_v]) \n" |
| 272 | "swl $t6, 3(%[dst_v]) \n" // |31|29|27|25| |
| 273 | "swr $t2, 4(%[dst_u]) \n" |
| 274 | "swl $t2, 7(%[dst_u]) \n" // |22|20|18|16| |
| 275 | "swr $t3, 4(%[dst_v]) \n" |
| 276 | "swl $t3, 7(%[dst_v]) \n" // |23|21|19|17| |
| 277 | "swr $t0, 8(%[dst_u]) \n" |
| 278 | "swl $t0, 11(%[dst_u]) \n" // |14|12|10|8| |
| 279 | "swr $t1, 8(%[dst_v]) \n" |
| 280 | "swl $t1, 11(%[dst_v]) \n" // |15|13|11|9| |
| 281 | "swr $t9, 12(%[dst_u]) \n" |
| 282 | "swl $t9, 15(%[dst_u]) \n" // |6|4|2|0| |
| 283 | "swr $t5, 12(%[dst_v]) \n" |
| 284 | "swl $t5, 15(%[dst_v]) \n" // |7|5|3|1| |
| 285 | "addiu %[dst_v], %[dst_v], 16 \n" |
| 286 | "bgtz %[x], 1b \n" |
| 287 | " addiu %[dst_u], %[dst_u], 16 \n" |
| 288 | "beqz %[y], 3f \n" |
| 289 | " nop \n" |
| 290 | "b 2f \n" |
| 291 | " nop \n" |
| 292 | |
| 293 | "2: \n" |
| 294 | "lbu $t0, -2(%[src_uv]) \n" |
| 295 | "lbu $t1, -1(%[src_uv]) \n" |
| 296 | "addiu %[src_uv], %[src_uv], -2 \n" |
| 297 | "addiu %[y], %[y], -1 \n" |
| 298 | "sb $t0, 0(%[dst_u]) \n" |
| 299 | "sb $t1, 0(%[dst_v]) \n" |
| 300 | "addiu %[dst_u], %[dst_u], 1 \n" |
| 301 | "bgtz %[y], 2b \n" |
| 302 | " addiu %[dst_v], %[dst_v], 1 \n" |
| 303 | |
| 304 | "3: \n" |
| 305 | ".set pop \n" |
| 306 | : [src_uv] "+r" (src_uv), |
| 307 | [dst_u] "+r" (dst_u), |
| 308 | [dst_v] "+r" (dst_v), |
| 309 | [x] "=&r" (x), |
| 310 | [y] "+r" (y) |
| 311 | : [width] "r" (width) |
| 312 | : "t0", "t1", "t2", "t3", "t4", |
| 313 | "t5", "t7", "t8", "t9" |
| 314 | ); |
| 315 | } |
fbarchard@google.com | 6c1b2d3 | 2012-10-26 22:49:18 +0000 | [diff] [blame] | 316 | |
| 317 | // Convert (4 Y and 2 VU) I422 and arrange RGB values into |
| 318 | // t5 = | 0 | B0 | 0 | b0 | |
| 319 | // t4 = | 0 | B1 | 0 | b1 | |
| 320 | // t9 = | 0 | G0 | 0 | g0 | |
| 321 | // t8 = | 0 | G1 | 0 | g1 | |
| 322 | // t2 = | 0 | R0 | 0 | r0 | |
| 323 | // t1 = | 0 | R1 | 0 | r1 | |
| 324 | #define I422ToTransientMipsRGB \ |
| 325 | "lw $t0, 0(%[y_buf]) \n" \ |
| 326 | "lhu $t1, 0(%[u_buf]) \n" \ |
| 327 | "lhu $t2, 0(%[v_buf]) \n" \ |
| 328 | "preceu.ph.qbr $t1, $t1 \n" \ |
| 329 | "preceu.ph.qbr $t2, $t2 \n" \ |
| 330 | "preceu.ph.qbra $t3, $t0 \n" \ |
| 331 | "preceu.ph.qbla $t0, $t0 \n" \ |
| 332 | "subu.ph $t1, $t1, $s5 \n" \ |
| 333 | "subu.ph $t2, $t2, $s5 \n" \ |
| 334 | "subu.ph $t3, $t3, $s4 \n" \ |
| 335 | "subu.ph $t0, $t0, $s4 \n" \ |
| 336 | "mul.ph $t3, $t3, $s0 \n" \ |
| 337 | "mul.ph $t0, $t0, $s0 \n" \ |
| 338 | "shll.ph $t4, $t1, 0x7 \n" \ |
| 339 | "subu.ph $t4, $t4, $t1 \n" \ |
| 340 | "mul.ph $t6, $t1, $s1 \n" \ |
| 341 | "mul.ph $t1, $t2, $s2 \n" \ |
| 342 | "addq_s.ph $t5, $t4, $t3 \n" \ |
| 343 | "addq_s.ph $t4, $t4, $t0 \n" \ |
| 344 | "shra.ph $t5, $t5, 6 \n" \ |
| 345 | "shra.ph $t4, $t4, 6 \n" \ |
| 346 | "addiu %[u_buf], 2 \n" \ |
| 347 | "addiu %[v_buf], 2 \n" \ |
| 348 | "addu.ph $t6, $t6, $t1 \n" \ |
| 349 | "mul.ph $t1, $t2, $s3 \n" \ |
| 350 | "addu.ph $t9, $t6, $t3 \n" \ |
| 351 | "addu.ph $t8, $t6, $t0 \n" \ |
| 352 | "shra.ph $t9, $t9, 6 \n" \ |
| 353 | "shra.ph $t8, $t8, 6 \n" \ |
| 354 | "addu.ph $t2, $t1, $t3 \n" \ |
| 355 | "addu.ph $t1, $t1, $t0 \n" \ |
| 356 | "shra.ph $t2, $t2, 6 \n" \ |
| 357 | "shra.ph $t1, $t1, 6 \n" \ |
| 358 | "subu.ph $t5, $t5, $s5 \n" \ |
| 359 | "subu.ph $t4, $t4, $s5 \n" \ |
| 360 | "subu.ph $t9, $t9, $s5 \n" \ |
| 361 | "subu.ph $t8, $t8, $s5 \n" \ |
| 362 | "subu.ph $t2, $t2, $s5 \n" \ |
| 363 | "subu.ph $t1, $t1, $s5 \n" \ |
| 364 | "shll_s.ph $t5, $t5, 8 \n" \ |
| 365 | "shll_s.ph $t4, $t4, 8 \n" \ |
| 366 | "shll_s.ph $t9, $t9, 8 \n" \ |
| 367 | "shll_s.ph $t8, $t8, 8 \n" \ |
| 368 | "shll_s.ph $t2, $t2, 8 \n" \ |
| 369 | "shll_s.ph $t1, $t1, 8 \n" \ |
| 370 | "shra.ph $t5, $t5, 8 \n" \ |
| 371 | "shra.ph $t4, $t4, 8 \n" \ |
| 372 | "shra.ph $t9, $t9, 8 \n" \ |
| 373 | "shra.ph $t8, $t8, 8 \n" \ |
| 374 | "shra.ph $t2, $t2, 8 \n" \ |
| 375 | "shra.ph $t1, $t1, 8 \n" \ |
| 376 | "addu.ph $t5, $t5, $s5 \n" \ |
| 377 | "addu.ph $t4, $t4, $s5 \n" \ |
| 378 | "addu.ph $t9, $t9, $s5 \n" \ |
| 379 | "addu.ph $t8, $t8, $s5 \n" \ |
| 380 | "addu.ph $t2, $t2, $s5 \n" \ |
| 381 | "addu.ph $t1, $t1, $s5 \n" |
| 382 | |
| 383 | void I422ToARGBRow_MIPS_DSPR2(const uint8* y_buf, |
| 384 | const uint8* u_buf, |
| 385 | const uint8* v_buf, |
| 386 | uint8* rgb_buf, |
| 387 | int width) { |
| 388 | __asm__ __volatile__ ( |
| 389 | ".set push \n" |
| 390 | ".set noreorder \n" |
| 391 | "beqz %[width], 2f \n" |
| 392 | " repl.ph $s0, 74 \n" // |YG|YG| = |74|74| |
| 393 | "repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25| |
| 394 | "repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52| |
| 395 | "repl.ph $s3, 102 \n" // |VR|VR| = |102|102| |
| 396 | "repl.ph $s4, 16 \n" // |0|16|0|16| |
| 397 | "repl.ph $s5, 128 \n" // |128|128| // clipping |
| 398 | "lui $s6, 0xff00 \n" |
| 399 | "ori $s6, 0xff00 \n" // |ff|00|ff|00|ff| |
| 400 | "1: \n" |
| 401 | I422ToTransientMipsRGB |
| 402 | // Arranging into argb format |
| 403 | "precr.qb.ph $t4, $t8, $t4 \n" // |G1|g1|B1|b1| |
| 404 | "precr.qb.ph $t5, $t9, $t5 \n" // |G0|g0|B0|b0| |
| 405 | "addiu %[width], -4 \n" |
| 406 | "precrq.qb.ph $t8, $t4, $t5 \n" // |G1|B1|G0|B0| |
| 407 | "precr.qb.ph $t9, $t4, $t5 \n" // |g1|b1|g0|b0| |
| 408 | "precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0| |
| 409 | |
| 410 | "addiu %[y_buf], 4 \n" |
| 411 | "preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0| |
| 412 | "preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0| |
| 413 | "or $t1, $t1, $s6 \n" // |ff|R1|ff|R0| |
| 414 | "or $t2, $t2, $s6 \n" // |ff|r1|ff|r0| |
| 415 | "precrq.ph.w $t0, $t2, $t9 \n" // |ff|r1|g1|b1| |
| 416 | "precrq.ph.w $t3, $t1, $t8 \n" // |ff|R1|G1|B1| |
| 417 | "sll $t9, $t9, 16 \n" |
| 418 | "sll $t8, $t8, 16 \n" |
| 419 | "packrl.ph $t2, $t2, $t9 \n" // |ff|r0|g0|b0| |
| 420 | "packrl.ph $t1, $t1, $t8 \n" // |ff|R0|G0|B0| |
| 421 | // Store results. |
| 422 | "sw $t2, 0(%[rgb_buf]) \n" |
| 423 | "sw $t0, 4(%[rgb_buf]) \n" |
| 424 | "sw $t1, 8(%[rgb_buf]) \n" |
| 425 | "sw $t3, 12(%[rgb_buf]) \n" |
| 426 | "bnez %[width], 1b \n" |
| 427 | " addiu %[rgb_buf], 16 \n" |
| 428 | "2: \n" |
| 429 | ".set pop \n" |
| 430 | :[y_buf] "+r" (y_buf), |
| 431 | [u_buf] "+r" (u_buf), |
| 432 | [v_buf] "+r" (v_buf), |
| 433 | [width] "+r" (width), |
| 434 | [rgb_buf] "+r" (rgb_buf) |
| 435 | : |
| 436 | : "t0", "t1", "t2", "t3", "t4", "t5", |
| 437 | "t6", "t7", "t8", "t9", |
| 438 | "s0", "s1", "s2", "s3", |
| 439 | "s4", "s5", "s6" |
| 440 | ); |
| 441 | } |
| 442 | |
| 443 | void I422ToABGRRow_MIPS_DSPR2(const uint8* y_buf, |
| 444 | const uint8* u_buf, |
| 445 | const uint8* v_buf, |
| 446 | uint8* rgb_buf, |
| 447 | int width) { |
| 448 | __asm__ __volatile__ ( |
| 449 | ".set push \n\t" |
| 450 | ".set noreorder \n\t" |
| 451 | "beqz %[width], 2f \n\t" |
| 452 | " repl.ph $s0, 74 \n\t" // |YG|YG| = |74|74| |
| 453 | "repl.ph $s1, -25 \n\t" // |UG|UG| = |-25|-25| |
| 454 | "repl.ph $s2, -52 \n\t" // |VG|VG| = |-52|-52| |
| 455 | "repl.ph $s3, 102 \n\t" // |VR|VR| = |102|102| |
| 456 | "repl.ph $s4, 16 \n\t" // |0|16|0|16| |
| 457 | "repl.ph $s5, 128 \n\t" // |128|128| |
| 458 | "lui $s6, 0xff00 \n\t" |
| 459 | "ori $s6, 0xff00 \n\t" // |ff|00|ff|00| |
| 460 | "1: \n" |
| 461 | I422ToTransientMipsRGB |
| 462 | // Arranging into abgr format |
| 463 | "precr.qb.ph $t0, $t8, $t1 \n\t" // |G1|g1|R1|r1| |
| 464 | "precr.qb.ph $t3, $t9, $t2 \n\t" // |G0|g0|R0|r0| |
| 465 | "precrq.qb.ph $t8, $t0, $t3 \n\t" // |G1|R1|G0|R0| |
| 466 | "precr.qb.ph $t9, $t0, $t3 \n\t" // |g1|r1|g0|r0| |
| 467 | |
| 468 | "precr.qb.ph $t2, $t4, $t5 \n\t" // |B1|b1|B0|b0| |
| 469 | "addiu %[width], -4 \n\t" |
| 470 | "addiu %[y_buf], 4 \n\t" |
| 471 | "preceu.ph.qbla $t1, $t2 \n\t" // |0 |B1|0 |B0| |
| 472 | "preceu.ph.qbra $t2, $t2 \n\t" // |0 |b1|0 |b0| |
| 473 | "or $t1, $t1, $s6 \n\t" // |ff|B1|ff|B0| |
| 474 | "or $t2, $t2, $s6 \n\t" // |ff|b1|ff|b0| |
| 475 | "precrq.ph.w $t0, $t2, $t9 \n\t" // |ff|b1|g1|r1| |
| 476 | "precrq.ph.w $t3, $t1, $t8 \n\t" // |ff|B1|G1|R1| |
| 477 | "sll $t9, $t9, 16 \n\t" |
| 478 | "sll $t8, $t8, 16 \n\t" |
| 479 | "packrl.ph $t2, $t2, $t9 \n\t" // |ff|b0|g0|r0| |
| 480 | "packrl.ph $t1, $t1, $t8 \n\t" // |ff|B0|G0|R0| |
| 481 | // Store results. |
| 482 | "sw $t2, 0(%[rgb_buf]) \n\t" |
| 483 | "sw $t0, 4(%[rgb_buf]) \n\t" |
| 484 | "sw $t1, 8(%[rgb_buf]) \n\t" |
| 485 | "sw $t3, 12(%[rgb_buf]) \n\t" |
| 486 | "bnez %[width], 1b \n\t" |
| 487 | " addiu %[rgb_buf], 16 \n\t" |
| 488 | "2: \n\t" |
| 489 | ".set pop \n\t" |
| 490 | :[y_buf] "+r" (y_buf), |
| 491 | [u_buf] "+r" (u_buf), |
| 492 | [v_buf] "+r" (v_buf), |
| 493 | [width] "+r" (width), |
| 494 | [rgb_buf] "+r" (rgb_buf) |
| 495 | : |
| 496 | : "t0", "t1", "t2", "t3", "t4", "t5", |
| 497 | "t6", "t7", "t8", "t9", |
| 498 | "s0", "s1", "s2", "s3", |
| 499 | "s4", "s5", "s6" |
| 500 | ); |
| 501 | } |
| 502 | |
| 503 | void I422ToBGRARow_MIPS_DSPR2(const uint8* y_buf, |
| 504 | const uint8* u_buf, |
| 505 | const uint8* v_buf, |
| 506 | uint8* rgb_buf, |
| 507 | int width) { |
| 508 | __asm__ __volatile__ ( |
| 509 | ".set push \n" |
| 510 | ".set noreorder \n" |
| 511 | "beqz %[width], 2f \n" |
| 512 | " repl.ph $s0, 74 \n" // |YG|YG| = |74 |74 | |
| 513 | "repl.ph $s1, -25 \n" // |UG|UG| = |-25|-25| |
| 514 | "repl.ph $s2, -52 \n" // |VG|VG| = |-52|-52| |
| 515 | "repl.ph $s3, 102 \n" // |VR|VR| = |102|102| |
| 516 | "repl.ph $s4, 16 \n" // |0|16|0|16| |
| 517 | "repl.ph $s5, 128 \n" // |128|128| |
| 518 | "lui $s6, 0xff \n" |
| 519 | "ori $s6, 0xff \n" // |00|ff|00|ff| |
| 520 | "1: \n" |
| 521 | I422ToTransientMipsRGB |
| 522 | // Arranging into bgra format |
| 523 | "precr.qb.ph $t4, $t4, $t8 \n" // |B1|b1|G1|g1| |
| 524 | "precr.qb.ph $t5, $t5, $t9 \n" // |B0|b0|G0|g0| |
| 525 | "precrq.qb.ph $t8, $t4, $t5 \n" // |B1|G1|B0|G0| |
| 526 | "precr.qb.ph $t9, $t4, $t5 \n" // |b1|g1|b0|g0| |
| 527 | |
| 528 | "precr.qb.ph $t2, $t1, $t2 \n" // |R1|r1|R0|r0| |
| 529 | "addiu %[width], -4 \n" |
| 530 | "addiu %[y_buf], 4 \n" |
| 531 | "preceu.ph.qbla $t1, $t2 \n" // |0 |R1|0 |R0| |
| 532 | "preceu.ph.qbra $t2, $t2 \n" // |0 |r1|0 |r0| |
| 533 | "sll $t1, $t1, 8 \n" // |R1|0 |R0|0 | |
| 534 | "sll $t2, $t2, 8 \n" // |r1|0 |r0|0 | |
| 535 | "or $t1, $t1, $s6 \n" // |R1|ff|R0|ff| |
| 536 | "or $t2, $t2, $s6 \n" // |r1|ff|r0|ff| |
| 537 | "precrq.ph.w $t0, $t9, $t2 \n" // |b1|g1|r1|ff| |
| 538 | "precrq.ph.w $t3, $t8, $t1 \n" // |B1|G1|R1|ff| |
| 539 | "sll $t1, $t1, 16 \n" |
| 540 | "sll $t2, $t2, 16 \n" |
| 541 | "packrl.ph $t2, $t9, $t2 \n" // |b0|g0|r0|ff| |
| 542 | "packrl.ph $t1, $t8, $t1 \n" // |B0|G0|R0|ff| |
| 543 | // Store results. |
| 544 | "sw $t2, 0(%[rgb_buf]) \n" |
| 545 | "sw $t0, 4(%[rgb_buf]) \n" |
| 546 | "sw $t1, 8(%[rgb_buf]) \n" |
| 547 | "sw $t3, 12(%[rgb_buf]) \n" |
| 548 | "bnez %[width], 1b \n" |
| 549 | " addiu %[rgb_buf], 16 \n" |
| 550 | "2: \n" |
| 551 | ".set pop \n" |
| 552 | :[y_buf] "+r" (y_buf), |
| 553 | [u_buf] "+r" (u_buf), |
| 554 | [v_buf] "+r" (v_buf), |
| 555 | [width] "+r" (width), |
| 556 | [rgb_buf] "+r" (rgb_buf) |
| 557 | : |
| 558 | : "t0", "t1", "t2", "t3", "t4", "t5", |
| 559 | "t6", "t7", "t8", "t9", |
| 560 | "s0", "s1", "s2", "s3", |
| 561 | "s4", "s5", "s6" |
| 562 | ); |
| 563 | } |
fbarchard@google.com | 834b748 | 2012-12-11 17:59:26 +0000 | [diff] [blame] | 564 | #endif // __mips_dsp_rev >= 2 |
fbarchard@google.com | ca41005 | 2012-10-14 06:01:19 +0000 | [diff] [blame] | 565 | |
| 566 | #ifdef __cplusplus |
| 567 | } // extern "C" |
| 568 | } // namespace libyuv |
| 569 | #endif |