Showing the changes to codec/decoder (1 of 6 files). Show all
Unified · Split
| @@ -1,1602 +0,0 @@ | |||
| 1 | /*! | ||
| 2 | * \copy | ||
| 3 | * Copyright (c) 2013, Cisco Systems | ||
| 4 | * All rights reserved. | ||
| 5 | * | ||
| 6 | * Redistribution and use in source and binary forms, with or without | ||
| 7 | * modification, are permitted provided that the following conditions | ||
| 8 | * are met: | ||
| 9 | * | ||
| 10 | * * Redistributions of source code must retain the above copyright | ||
| 11 | * notice, this list of conditions and the following disclaimer. | ||
| 12 | * | ||
| 13 | * * Redistributions in binary form must reproduce the above copyright | ||
| 14 | * notice, this list of conditions and the following disclaimer in | ||
| 15 | * the documentation and/or other materials provided with the | ||
| 16 | * distribution. | ||
| 17 | * | ||
| 18 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | ||
| 19 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | ||
| 20 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS | ||
| 21 | * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE | ||
| 22 | * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, | ||
| 23 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, | ||
| 24 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; | ||
| 25 | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER | ||
| 26 | * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT | ||
| 27 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN | ||
| 28 | * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE | ||
| 29 | * POSSIBILITY OF SUCH DAMAGE. | ||
| 30 | * | ||
| 31 | */ | ||
| 32 | |||
| 33 | #ifdef HAVE_NEON | ||
| 34 | .text | ||
| 35 | #include "arm_arch_common_macro.S" | ||
| 36 | |||
| 37 | #ifdef APPLE_IOS | ||
| 38 | .macro AVERAGE_TWO_8BITS | ||
| 39 | // { // input:dst_d, src_d A and B; working: q13 | ||
| 40 | vaddl.u8 q13, $2, $1 | ||
| 41 | vrshrn.u16 $0, q13, #1 | ||
| 42 | // } | ||
| 43 | .endm | ||
| 44 | |||
| 45 | .macro FILTER_6TAG_8BITS | ||
| 46 | // { // input:src[-2], src[-1], src[0], src[1], src[2], src[3], dst_d, multiplier a/b; working: q12, q13 | ||
| 47 | vaddl.u8 q12, $0, $5 //q12=src[-2]+src[3] | ||
| 48 | vaddl.u8 q13, $2, $3 //src[0]+src[1] | ||
| 49 | vmla.u16 q12, q13, $7 //q12 += 20*(src[0]+src[1]), 2 cycles | ||
| 50 | vaddl.u8 q13, $1, $4 //src[-1]+src[2] | ||
| 51 | vmls.s16 q12, q13, $8 //q12 -= 5*(src[-1]+src[2]), 2 cycles | ||
| 52 | vqrshrun.s16 $6, q12, #5 | ||
| 53 | // } | ||
| 54 | .endm | ||
| 55 | |||
| 56 | .macro FILTER_6TAG_8BITS_AVERAGE_WITH_0 | ||
| 57 | // { // input:src[-2], src[-1], src[0], src[1], src[2], src[3], dst_d, multiplier a/b; working: q12, q13 | ||
| 58 | vaddl.u8 q12, $0, $5 //q12=src[-2]+src[3] | ||
| 59 | vaddl.u8 q13, $2, $3 //src[0]+src[1] | ||
| 60 | vmla.u16 q12, q13, $7 //q12 += 20*(src[0]+src[1]), 2 cycles | ||
| 61 | vaddl.u8 q13, $1, $4 //src[-1]+src[2] | ||
| 62 | vmls.s16 q12, q13, $8 //q12 -= 5*(src[-1]+src[2]), 2 cycles | ||
| 63 | vqrshrun.s16 $6, q12, #5 | ||
| 64 | vaddl.u8 q13, $2, $6 | ||
| 65 | vrshrn.u16 $6, q13, #1 | ||
| 66 | // } | ||
| 67 | .endm | ||
| 68 | |||
| 69 | .macro FILTER_6TAG_8BITS_AVERAGE_WITH_1 | ||
| 70 | // { // input:src[-2], src[-1], src[0], src[1], src[2], src[3], dst_d, multiplier a/b; working: q12, q13 | ||
| 71 | vaddl.u8 q12, $0, $5 //q12=src[-2]+src[3] | ||
| 72 | vaddl.u8 q13, $2, $3 //src[0]+src[1] | ||
| 73 | vmla.u16 q12, q13, $7 //q12 += 20*(src[0]+src[1]), 2 cycles | ||
| 74 | vaddl.u8 q13, $1, $4 //src[-1]+src[2] | ||
| 75 | vmls.s16 q12, q13, $8 //q12 -= 5*(src[-1]+src[2]), 2 cycles | ||
| 76 | vqrshrun.s16 $6, q12, #5 | ||
| 77 | vaddl.u8 q13, $3, $6 | ||
| 78 | vrshrn.u16 $6, q13, #1 | ||
| 79 | // } | ||
| 80 | .endm | ||
| 81 | |||
| 82 | .macro FILTER_6TAG_8BITS_TO_16BITS | ||
| 83 | // { // input:d_src[-2], d_src[-1], d_src[0], d_src[1], d_src[2], d_src[3], dst_q, multiplier a/b; working:q13 | ||
| 84 | vaddl.u8 $6, $0, $5 //dst_q=src[-2]+src[3] | ||
| 85 | vaddl.u8 q13, $2, $3 //src[0]+src[1] | ||
| 86 | vmla.u16 $6, q13, $7 //dst_q += 20*(src[0]+src[1]), 2 cycles | ||
| 87 | vaddl.u8 q13, $1, $4 //src[-1]+src[2] | ||
| 88 | vmls.s16 $6, q13, $8 //dst_q -= 5*(src[-1]+src[2]), 2 cycles | ||
| 89 | // } | ||
| 90 | .endm | ||
| 91 | |||
| 92 | .macro FILTER_3_IN_16BITS_TO_8BITS | ||
| 93 | // { // input:a, b, c, dst_d; | ||
| 94 | vsub.s16 $0, $0, $1 //a-b | ||
| 95 | vshr.s16 $0, $0, #2 //(a-b)/4 | ||
| 96 | vsub.s16 $0, $0, $1 //(a-b)/4-b | ||
| 97 | vadd.s16 $0, $0, $2 //(a-b)/4-b+c | ||
| 98 | vshr.s16 $0, $0, #2 //((a-b)/4-b+c)/4 | ||
| 99 | vadd.s16 $0, $0, $2 //((a-b)/4-b+c)/4+c = (a-5*b+20*c)/16 | ||
| 100 | vqrshrun.s16 $3, $0, #6 //(+32)>>6 | ||
| 101 | // } | ||
| 102 | .endm | ||
| 103 | |||
| 104 | .macro UNPACK_2_16BITS_TO_ABC | ||
| 105 | // { // input:q_src[-2:5], q_src[6:13](avail 8+5)/q_src[6:**](avail 4+5), dst_a, dst_b, dst_c; | ||
| 106 | vext.16 $4, $0, $1, #2 //src[0] | ||
| 107 | vext.16 $3, $0, $1, #3 //src[1] | ||
| 108 | vadd.s16 $4, $3 //c=src[0]+src[1] | ||
| 109 | |||
| 110 | vext.16 $3, $0, $1, #1 //src[-1] | ||
| 111 | vext.16 $2, $0, $1, #4 //src[2] | ||
| 112 | vadd.s16 $3, $2 //b=src[-1]+src[2] | ||
| 113 | |||
| 114 | vext.16 $2, $0, $1, #5 //src[3] | ||
| 115 | vadd.s16 $2, $0 //a=src[-2]+src[3] | ||
| 116 | // } | ||
| 117 | .endm | ||
| 118 | #else | ||
| 119 | .macro AVERAGE_TWO_8BITS arg0, arg1, arg2 | ||
| 120 | // { // input:dst_d, src_d A and B; working: q13 | ||
| 121 | vaddl.u8 q13, \arg2, \arg1 | ||
| 122 | vrshrn.u16 \arg0, q13, #1 | ||
| 123 | // } | ||
| 124 | .endm | ||
| 125 | |||
| 126 | .macro FILTER_6TAG_8BITS arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8 | ||
| 127 | // { // input:src[-2], src[-1], src[0], src[1], src[2], src[3], dst_d, multiplier a/b; working: q12, q13 | ||
| 128 | vaddl.u8 q12, \arg0, \arg5 //q12=src[-2]+src[3] | ||
| 129 | vaddl.u8 q13, \arg2, \arg3 //src[0]+src[1] | ||
| 130 | vmla.u16 q12, q13, \arg7 //q12 += 20*(src[0]+src[1]), 2 cycles | ||
| 131 | vaddl.u8 q13, \arg1, \arg4 //src[-1]+src[2] | ||
| 132 | vmls.s16 q12, q13, \arg8 //q12 -= 5*(src[-1]+src[2]), 2 cycles | ||
| 133 | vqrshrun.s16 \arg6, q12, #5 | ||
| 134 | // } | ||
| 135 | .endm | ||
| 136 | |||
| 137 | .macro FILTER_6TAG_8BITS_AVERAGE_WITH_0 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8 | ||
| 138 | // { // input:src[-2], src[-1], src[0], src[1], src[2], src[3], dst_d, multiplier a/b; working: q12, q13 | ||
| 139 | vaddl.u8 q12, \arg0, \arg5 //q12=src[-2]+src[3] | ||
| 140 | vaddl.u8 q13, \arg2, \arg3 //src[0]+src[1] | ||
| 141 | vmla.u16 q12, q13, \arg7 //q12 += 20*(src[0]+src[1]), 2 cycles | ||
| 142 | vaddl.u8 q13, \arg1, \arg4 //src[-1]+src[2] | ||
| 143 | vmls.s16 q12, q13, \arg8 //q12 -= 5*(src[-1]+src[2]), 2 cycles | ||
| 144 | vqrshrun.s16 \arg6, q12, #5 | ||
| 145 | vaddl.u8 q13, \arg2, \arg6 | ||
| 146 | vrshrn.u16 \arg6, q13, #1 | ||
| 147 | // } | ||
| 148 | .endm | ||
| 149 | |||
| 150 | .macro FILTER_6TAG_8BITS_AVERAGE_WITH_1 arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8 | ||
| 151 | // { // input:src[-2], src[-1], src[0], src[1], src[2], src[3], dst_d, multiplier a/b; working: q12, q13 | ||
| 152 | vaddl.u8 q12, \arg0, \arg5 //q12=src[-2]+src[3] | ||
| 153 | vaddl.u8 q13, \arg2, \arg3 //src[0]+src[1] | ||
| 154 | vmla.u16 q12, q13, \arg7 //q12 += 20*(src[0]+src[1]), 2 cycles | ||
| 155 | vaddl.u8 q13, \arg1, \arg4 //src[-1]+src[2] | ||
| 156 | vmls.s16 q12, q13, \arg8 //q12 -= 5*(src[-1]+src[2]), 2 cycles | ||
| 157 | vqrshrun.s16 \arg6, q12, #5 | ||
| 158 | vaddl.u8 q13, \arg3, \arg6 | ||
| 159 | vrshrn.u16 \arg6, q13, #1 | ||
| 160 | // } | ||
| 161 | .endm | ||
| 162 | |||
| 163 | .macro FILTER_6TAG_8BITS_TO_16BITS arg0, arg1, arg2, arg3, arg4, arg5, arg6, arg7, arg8 | ||
| 164 | // { // input:d_src[-2], d_src[-1], d_src[0], d_src[1], d_src[2], d_src[3], dst_q, multiplier a/b; working:q13 | ||
| 165 | vaddl.u8 \arg6, \arg0, \arg5 //dst_q=src[-2]+src[3] | ||
| 166 | vaddl.u8 q13, \arg2, \arg3 //src[0]+src[1] | ||
| 167 | vmla.u16 \arg6, q13, \arg7 //dst_q += 20*(src[0]+src[1]), 2 cycles | ||
| 168 | vaddl.u8 q13, \arg1, \arg4 //src[-1]+src[2] | ||
| 169 | vmls.s16 \arg6, q13, \arg8 //dst_q -= 5*(src[-1]+src[2]), 2 cycles | ||
| 170 | // } | ||
| 171 | .endm | ||
| 172 | |||
| 173 | .macro FILTER_3_IN_16BITS_TO_8BITS arg0, arg1, arg2, arg3 | ||
| 174 | // { // input:a, b, c, dst_d; | ||
| 175 | vsub.s16 \arg0, \arg0, \arg1 //a-b | ||
| 176 | vshr.s16 \arg0, \arg0, #2 //(a-b)/4 | ||
| 177 | vsub.s16 \arg0, \arg0, \arg1 //(a-b)/4-b | ||
| 178 | vadd.s16 \arg0, \arg0, \arg2 //(a-b)/4-b+c | ||
| 179 | vshr.s16 \arg0, \arg0, #2 //((a-b)/4-b+c)/4 | ||
| 180 | vadd.s16 \arg0, \arg0, \arg2 //((a-b)/4-b+c)/4+c = (a-5*b+20*c)/16 | ||
| 181 | vqrshrun.s16 \arg3, \arg0, #6 //(+32)>>6 | ||
| 182 | // } | ||
| 183 | .endm | ||
| 184 | |||
| 185 | .macro UNPACK_2_16BITS_TO_ABC arg0, arg1, arg2, arg3, arg4 | ||
| 186 | // { // input:q_src[-2:5], q_src[6:13](avail 8+5)/q_src[6:**](avail 4+5), dst_a, dst_b, dst_c; | ||
| 187 | vext.16 \arg4, \arg0, \arg1, #2 //src[0] | ||
| 188 | vext.16 \arg3, \arg0, \arg1, #3 //src[1] | ||
| 189 | vadd.s16 \arg4, \arg3 //c=src[0]+src[1] | ||
| 190 | |||
| 191 | vext.16 \arg3, \arg0, \arg1, #1 //src[-1] | ||
| 192 | vext.16 \arg2, \arg0, \arg1, #4 //src[2] | ||
| 193 | vadd.s16 \arg3,\arg2 //b=src[-1]+src[2] | ||
| 194 | |||
| 195 | vext.16 \arg2, \arg0, \arg1, #5 //src[3] | ||
| 196 | vadd.s16 \arg2, \arg0 //a=src[-2]+src[3] | ||
| 197 | // } | ||
| 198 | .endm | ||
| 199 | #endif | ||
| 200 | |||
| 201 | WELS_ASM_FUNC_BEGIN McHorVer20WidthEq16_neon | ||
| 202 | push {r4} | ||
| 203 | ldr r4, [sp, #4] | ||
| 204 | |||
| 205 | sub r0, #2 | ||
| 206 | vmov.u16 q14, #0x0014 // 20 | ||
| 207 | vshr.u16 q15, q14, #2 // 5 | ||
| 208 | |||
| 209 | w16_h_mc_luma_loop: | ||
| 210 | vld1.u8 {d0,d1,d2}, [r0], r1 //only use 21(16+5); q0=src[-2] | ||
| 211 | pld [r0] | ||
| 212 | pld [r0, #16] | ||
| 213 | |||
| 214 | vext.8 q2, q0, q1, #1 //q2=src[-1] | ||
| 215 | vext.8 q3, q0, q1, #2 //q3=src[0] | ||
| 216 | vext.8 q4, q0, q1, #3 //q4=src[1] | ||
| 217 | vext.8 q5, q0, q1, #4 //q5=src[2] | ||
| 218 | vext.8 q6, q0, q1, #5 //q6=src[3] | ||
| 219 | |||
| 220 | FILTER_6TAG_8BITS d0, d4, d6, d8, d10, d12, d2, q14, q15 | ||
| 221 | |||
| 222 | FILTER_6TAG_8BITS d1, d5, d7, d9, d11, d13, d3, q14, q15 | ||
| 223 | |||
| 224 | sub r4, #1 | ||
| 225 | vst1.u8 {d2, d3}, [r2], r3 //write 16Byte | ||
| 226 | |||
| 227 | cmp r4, #0 | ||
| 228 | bne w16_h_mc_luma_loop | ||
| 229 | pop {r4} | ||
| 230 | WELS_ASM_FUNC_END | ||
| 231 | |||
| 232 | |||
| 233 | WELS_ASM_FUNC_BEGIN McHorVer20WidthEq8_neon | ||
| 234 | push {r4} | ||
| 235 | ldr r4, [sp, #4] | ||
| 236 | |||
| 237 | sub r0, #2 | ||
| 238 | vmov.u16 q14, #0x0014 // 20 | ||
| 239 | vshr.u16 q15, q14, #2 // 5 | ||
| 240 | |||
| 241 | w8_h_mc_luma_loop: | ||
| 242 | vld1.u8 {d0,d1}, [r0], r1 //only use 13(8+5); q0=src[-2] | ||
| 243 | pld [r0] | ||
| 244 | |||
| 245 | vext.8 d2, d0, d1, #1 //d2=src[-1] | ||
| 246 | vext.8 d3, d0, d1, #2 //d3=src[0] | ||
| 247 | vext.8 d4, d0, d1, #3 //d4=src[1] | ||
| 248 | vext.8 d5, d0, d1, #4 //d5=src[2] | ||
| 249 | vext.8 d6, d0, d1, #5 //d6=src[3] | ||
| 250 | |||
| 251 | FILTER_6TAG_8BITS d0, d2, d3, d4, d5, d6, d1, q14, q15 | ||
| 252 | |||
| 253 | sub r4, #1 | ||
| 254 | vst1.u8 {d1}, [r2], r3 | ||
| 255 | |||
| 256 | cmp r4, #0 | ||
| 257 | bne w8_h_mc_luma_loop | ||
| 258 | pop {r4} | ||
| 259 | WELS_ASM_FUNC_END | ||
| 260 | |||
| 261 | |||
| 262 | WELS_ASM_FUNC_BEGIN McHorVer20WidthEq4_neon | ||
| 263 | push {r4, r5, r6} | ||
| 264 | ldr r6, [sp, #12] | ||
| 265 | |||
| 266 | sub r0, #2 | ||
| 267 | vmov.u16 q14, #0x0014 // 20 | ||
| 268 | vshr.u16 q15, q14, #2 // 5 | ||
| 269 | |||
| 270 | w4_h_mc_luma_loop: | ||
| 271 | vld1.u8 {d0, d1}, [r0], r1 //only use 9(4+5);d0: 1st row src[-2:5] | ||
| 272 | pld [r0] | ||
| 273 | vld1.u8 {d2, d3}, [r0], r1 //d2: 2nd row src[-2:5] | ||
| 274 | pld [r0] | ||
| 275 | |||
| 276 | vext.8 d4, d0, d1, #1 //d4: 1st row src[-1:6] | ||
| 277 | vext.8 d5, d2, d3, #1 //d5: 2nd row src[-1:6] | ||
| 278 | vext.8 q3, q2, q2, #1 //src[0:6 *] | ||
| 279 | vext.8 q4, q2, q2, #2 //src[1:6 * *] | ||
| 280 | |||
| 281 | vtrn.32 q3, q4 //q3::d6:1st row [0:3]+[1:4]; d7:2nd row [0:3]+[1:4] | ||
| 282 | vtrn.32 d6, d7 //d6:[0:3]; d7[1:4] | ||
| 283 | vtrn.32 d0, d2 //d0:[-2:1]; d2[2:5] | ||
| 284 | vtrn.32 d4, d5 //d4:[-1:2]; d5[3:6] | ||
| 285 | |||
| 286 | FILTER_6TAG_8BITS d0, d4, d6, d7, d2, d5, d1, q14, q15 | ||
| 287 | |||
| 288 | vmov r4, r5, d1 | ||
| 289 | str r4, [r2], r3 | ||
| 290 | str r5, [r2], r3 | ||
| 291 | |||
| 292 | sub r6, #2 | ||
| 293 | cmp r6, #0 | ||
| 294 | bne w4_h_mc_luma_loop | ||
| 295 | |||
| 296 | pop {r4, r5, r6} | ||
| 297 | WELS_ASM_FUNC_END | ||
| 298 | |||
| 299 | |||
| 300 | WELS_ASM_FUNC_BEGIN McHorVer10WidthEq16_neon | ||
| 301 | push {r4} | ||
| 302 | ldr r4, [sp, #4] | ||
| 303 | |||
| 304 | sub r0, #2 | ||
| 305 | vmov.u16 q14, #0x0014 // 20 | ||
| 306 | vshr.u16 q15, q14, #2 // 5 | ||
| 307 | |||
| 308 | w16_xy_10_mc_luma_loop: | ||
| 309 | vld1.u8 {d0,d1,d2}, [r0], r1 //only use 21(16+5); q0=src[-2] | ||
| 310 | pld [r0] | ||
| 311 | pld [r0, #16] | ||
| 312 | |||
| 313 | vext.8 q2, q0, q1, #1 //q2=src[-1] | ||
| 314 | vext.8 q3, q0, q1, #2 //q3=src[0] | ||
| 315 | vext.8 q4, q0, q1, #3 //q4=src[1] | ||
| 316 | vext.8 q5, q0, q1, #4 //q5=src[2] | ||
| 317 | vext.8 q6, q0, q1, #5 //q6=src[3] | ||
| 318 | |||
| 319 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d4, d6, d8, d10, d12, d2, q14, q15 | ||
| 320 | |||
| 321 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d1, d5, d7, d9, d11, d13, d3, q14, q15 | ||
| 322 | |||
| 323 | sub r4, #1 | ||
| 324 | vst1.u8 {d2, d3}, [r2], r3 //write 16Byte | ||
| 325 | |||
| 326 | cmp r4, #0 | ||
| 327 | bne w16_xy_10_mc_luma_loop | ||
| 328 | pop {r4} | ||
| 329 | WELS_ASM_FUNC_END | ||
| 330 | |||
| 331 | |||
| 332 | WELS_ASM_FUNC_BEGIN McHorVer10WidthEq8_neon | ||
| 333 | push {r4} | ||
| 334 | ldr r4, [sp, #4] | ||
| 335 | |||
| 336 | sub r0, #2 | ||
| 337 | vmov.u16 q14, #0x0014 // 20 | ||
| 338 | vshr.u16 q15, q14, #2 // 5 | ||
| 339 | |||
| 340 | w8_xy_10_mc_luma_loop: | ||
| 341 | vld1.u8 {d0,d1}, [r0], r1 //only use 13(8+5); q0=src[-2] | ||
| 342 | pld [r0] | ||
| 343 | |||
| 344 | vext.8 d2, d0, d1, #1 //d2=src[-1] | ||
| 345 | vext.8 d3, d0, d1, #2 //d3=src[0] | ||
| 346 | vext.8 d4, d0, d1, #3 //d4=src[1] | ||
| 347 | vext.8 d5, d0, d1, #4 //d5=src[2] | ||
| 348 | vext.8 d6, d0, d1, #5 //d6=src[3] | ||
| 349 | |||
| 350 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d2, d3, d4, d5, d6, d1, q14, q15 | ||
| 351 | |||
| 352 | sub r4, #1 | ||
| 353 | vst1.u8 {d1}, [r2], r3 | ||
| 354 | |||
| 355 | cmp r4, #0 | ||
| 356 | bne w8_xy_10_mc_luma_loop | ||
| 357 | pop {r4} | ||
| 358 | WELS_ASM_FUNC_END | ||
| 359 | |||
| 360 | |||
| 361 | WELS_ASM_FUNC_BEGIN McHorVer10WidthEq4_neon | ||
| 362 | push {r4, r5, r6} | ||
| 363 | ldr r6, [sp, #12] | ||
| 364 | |||
| 365 | sub r0, #2 | ||
| 366 | vmov.u16 q14, #0x0014 // 20 | ||
| 367 | vshr.u16 q15, q14, #2 // 5 | ||
| 368 | |||
| 369 | w4_xy_10_mc_luma_loop: | ||
| 370 | vld1.u8 {d0, d1}, [r0], r1 //only use 9(4+5);d0: 1st row src[-2:5] | ||
| 371 | pld [r0] | ||
| 372 | vld1.u8 {d2, d3}, [r0], r1 //d2: 2nd row src[-2:5] | ||
| 373 | pld [r0] | ||
| 374 | |||
| 375 | vext.8 d4, d0, d1, #1 //d4: 1st row src[-1:6] | ||
| 376 | vext.8 d5, d2, d3, #1 //d5: 2nd row src[-1:6] | ||
| 377 | vext.8 q3, q2, q2, #1 //src[0:6 *] | ||
| 378 | vext.8 q4, q2, q2, #2 //src[1:6 * *] | ||
| 379 | |||
| 380 | vtrn.32 q3, q4 //q3::d6:1st row [0:3]+[1:4]; d7:2nd row [0:3]+[1:4] | ||
| 381 | vtrn.32 d6, d7 //d6:[0:3]; d7[1:4] | ||
| 382 | vtrn.32 d0, d2 //d0:[-2:1]; d2[2:5] | ||
| 383 | vtrn.32 d4, d5 //d4:[-1:2]; d5[3:6] | ||
| 384 | |||
| 385 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d4, d6, d7, d2, d5, d1, q14, q15 | ||
| 386 | |||
| 387 | vmov r4, r5, d1 | ||
| 388 | str r4, [r2], r3 | ||
| 389 | str r5, [r2], r3 | ||
| 390 | |||
| 391 | sub r6, #2 | ||
| 392 | cmp r6, #0 | ||
| 393 | bne w4_xy_10_mc_luma_loop | ||
| 394 | |||
| 395 | pop {r4, r5, r6} | ||
| 396 | WELS_ASM_FUNC_END | ||
| 397 | |||
| 398 | |||
| 399 | WELS_ASM_FUNC_BEGIN McHorVer30WidthEq16_neon | ||
| 400 | push {r4} | ||
| 401 | ldr r4, [sp, #4] | ||
| 402 | |||
| 403 | sub r0, #2 | ||
| 404 | vmov.u16 q14, #0x0014 // 20 | ||
| 405 | vshr.u16 q15, q14, #2 // 5 | ||
| 406 | |||
| 407 | w16_xy_30_mc_luma_loop: | ||
| 408 | vld1.u8 {d0,d1,d2}, [r0], r1 //only use 21(16+5); q0=src[-2] | ||
| 409 | pld [r0] | ||
| 410 | pld [r0, #16] | ||
| 411 | |||
| 412 | vext.8 q2, q0, q1, #1 //q2=src[-1] | ||
| 413 | vext.8 q3, q0, q1, #2 //q3=src[0] | ||
| 414 | vext.8 q4, q0, q1, #3 //q4=src[1] | ||
| 415 | vext.8 q5, q0, q1, #4 //q5=src[2] | ||
| 416 | vext.8 q6, q0, q1, #5 //q6=src[3] | ||
| 417 | |||
| 418 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d4, d6, d8, d10, d12, d2, q14, q15 | ||
| 419 | |||
| 420 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d1, d5, d7, d9, d11, d13, d3, q14, q15 | ||
| 421 | |||
| 422 | sub r4, #1 | ||
| 423 | vst1.u8 {d2, d3}, [r2], r3 //write 16Byte | ||
| 424 | |||
| 425 | cmp r4, #0 | ||
| 426 | bne w16_xy_30_mc_luma_loop | ||
| 427 | pop {r4} | ||
| 428 | WELS_ASM_FUNC_END | ||
| 429 | |||
| 430 | |||
| 431 | WELS_ASM_FUNC_BEGIN McHorVer30WidthEq8_neon | ||
| 432 | push {r4} | ||
| 433 | ldr r4, [sp, #4] | ||
| 434 | |||
| 435 | sub r0, #2 | ||
| 436 | vmov.u16 q14, #0x0014 // 20 | ||
| 437 | vshr.u16 q15, q14, #2 // 5 | ||
| 438 | |||
| 439 | w8_xy_30_mc_luma_loop: | ||
| 440 | vld1.u8 {d0,d1}, [r0], r1 //only use 13(8+5); q0=src[-2] | ||
| 441 | pld [r0] | ||
| 442 | |||
| 443 | vext.8 d2, d0, d1, #1 //d2=src[-1] | ||
| 444 | vext.8 d3, d0, d1, #2 //d3=src[0] | ||
| 445 | vext.8 d4, d0, d1, #3 //d4=src[1] | ||
| 446 | vext.8 d5, d0, d1, #4 //d5=src[2] | ||
| 447 | vext.8 d6, d0, d1, #5 //d6=src[3] | ||
| 448 | |||
| 449 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d2, d3, d4, d5, d6, d1, q14, q15 | ||
| 450 | |||
| 451 | sub r4, #1 | ||
| 452 | vst1.u8 {d1}, [r2], r3 | ||
| 453 | |||
| 454 | cmp r4, #0 | ||
| 455 | bne w8_xy_30_mc_luma_loop | ||
| 456 | pop {r4} | ||
| 457 | WELS_ASM_FUNC_END | ||
| 458 | |||
| 459 | |||
| 460 | WELS_ASM_FUNC_BEGIN McHorVer30WidthEq4_neon | ||
| 461 | push {r4, r5, r6} | ||
| 462 | ldr r6, [sp, #12] | ||
| 463 | |||
| 464 | sub r0, #2 | ||
| 465 | vmov.u16 q14, #0x0014 // 20 | ||
| 466 | vshr.u16 q15, q14, #2 // 5 | ||
| 467 | |||
| 468 | w4_xy_30_mc_luma_loop: | ||
| 469 | vld1.u8 {d0, d1}, [r0], r1 //only use 9(4+5);d0: 1st row src[-2:5] | ||
| 470 | pld [r0] | ||
| 471 | vld1.u8 {d2, d3}, [r0], r1 //d2: 2nd row src[-2:5] | ||
| 472 | pld [r0] | ||
| 473 | |||
| 474 | vext.8 d4, d0, d1, #1 //d4: 1st row src[-1:6] | ||
| 475 | vext.8 d5, d2, d3, #1 //d5: 2nd row src[-1:6] | ||
| 476 | vext.8 q3, q2, q2, #1 //src[0:6 *] | ||
| 477 | vext.8 q4, q2, q2, #2 //src[1:6 * *] | ||
| 478 | |||
| 479 | vtrn.32 q3, q4 //q3::d6:1st row [0:3]+[1:4]; d7:2nd row [0:3]+[1:4] | ||
| 480 | vtrn.32 d6, d7 //d6:[0:3]; d7[1:4] | ||
| 481 | vtrn.32 d0, d2 //d0:[-2:1]; d2[2:5] | ||
| 482 | vtrn.32 d4, d5 //d4:[-1:2]; d5[3:6] | ||
| 483 | |||
| 484 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d4, d6, d7, d2, d5, d1, q14, q15 | ||
| 485 | |||
| 486 | vmov r4, r5, d1 | ||
| 487 | str r4, [r2], r3 | ||
| 488 | str r5, [r2], r3 | ||
| 489 | |||
| 490 | sub r6, #2 | ||
| 491 | cmp r6, #0 | ||
| 492 | bne w4_xy_30_mc_luma_loop | ||
| 493 | |||
| 494 | pop {r4, r5, r6} | ||
| 495 | WELS_ASM_FUNC_END | ||
| 496 | |||
| 497 | |||
| 498 | WELS_ASM_FUNC_BEGIN McHorVer01WidthEq16_neon | ||
| 499 | push {r4} | ||
| 500 | ldr r4, [sp, #4] | ||
| 501 | |||
| 502 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 503 | pld [r0] | ||
| 504 | pld [r0, r1] | ||
| 505 | vmov.u16 q14, #0x0014 // 20 | ||
| 506 | vld1.u8 {q0}, [r0], r1 //q0=src[-2] | ||
| 507 | vld1.u8 {q1}, [r0], r1 //q1=src[-1] | ||
| 508 | |||
| 509 | pld [r0] | ||
| 510 | pld [r0, r1] | ||
| 511 | vshr.u16 q15, q14, #2 // 5 | ||
| 512 | vld1.u8 {q2}, [r0], r1 //q2=src[0] | ||
| 513 | vld1.u8 {q3}, [r0], r1 //q3=src[1] | ||
| 514 | vld1.u8 {q4}, [r0], r1 //q4=src[2] | ||
| 515 | |||
| 516 | w16_xy_01_luma_loop: | ||
| 517 | |||
| 518 | vld1.u8 {q5}, [r0], r1 //q5=src[3] | ||
| 519 | |||
| 520 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d2, d4, d6, d8, d10, d12, q14, q15 | ||
| 521 | pld [r0] | ||
| 522 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d1, d3, d5, d7, d9, d11, d13, q14, q15 | ||
| 523 | vld1.u8 {q0}, [r0], r1 //read 2nd row | ||
| 524 | vst1.u8 {q6}, [r2], r3 //write 1st 16Byte | ||
| 525 | |||
| 526 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d2, d4, d6, d8, d10, d0, d12, q14, q15 | ||
| 527 | pld [r0] | ||
| 528 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d3, d5, d7, d9, d11, d1, d13, q14, q15 | ||
| 529 | vld1.u8 {q1}, [r0], r1 //read 3rd row | ||
| 530 | vst1.u8 {q6}, [r2], r3 //write 2nd 16Byte | ||
| 531 | |||
| 532 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d4, d6, d8, d10, d0, d2, d12, q14, q15 | ||
| 533 | pld [r0] | ||
| 534 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d5, d7, d9, d11, d1, d3, d13, q14, q15 | ||
| 535 | vld1.u8 {q2}, [r0], r1 //read 4th row | ||
| 536 | vst1.u8 {q6}, [r2], r3 //write 3rd 16Byte | ||
| 537 | |||
| 538 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d6, d8, d10, d0, d2, d4, d12, q14, q15 | ||
| 539 | pld [r0] | ||
| 540 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d7, d9, d11, d1, d3, d5, d13, q14, q15 | ||
| 541 | vld1.u8 {q3}, [r0], r1 //read 5th row | ||
| 542 | vst1.u8 {q6}, [r2], r3 //write 4th 16Byte | ||
| 543 | |||
| 544 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d8, d10, d0, d2, d4, d6, d12, q14, q15 | ||
| 545 | pld [r0] | ||
| 546 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d9, d11, d1, d3, d5, d7, d13, q14, q15 | ||
| 547 | vld1.u8 {q4}, [r0], r1 //read 6th row | ||
| 548 | vst1.u8 {q6}, [r2], r3 //write 5th 16Byte | ||
| 549 | |||
| 550 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d10, d0, d2, d4, d6, d8, d12, q14, q15 | ||
| 551 | pld [r0] | ||
| 552 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d11, d1, d3, d5, d7, d9, d13, q14, q15 | ||
| 553 | vld1.u8 {q5}, [r0], r1 //read 7th row | ||
| 554 | vst1.u8 {q6}, [r2], r3 //write 6th 16Byte | ||
| 555 | |||
| 556 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d2, d4, d6, d8, d10, d12, q14, q15 | ||
| 557 | pld [r0] | ||
| 558 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d1, d3, d5, d7, d9, d11, d13, q14, q15 | ||
| 559 | vld1.u8 {q0}, [r0], r1 //read 8th row | ||
| 560 | vst1.u8 {q6}, [r2], r3 //write 7th 16Byte | ||
| 561 | |||
| 562 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d2, d4, d6, d8, d10, d0, d12, q14, q15 | ||
| 563 | pld [r0] | ||
| 564 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d3, d5, d7, d9, d11, d1, d13, q14, q15 | ||
| 565 | vst1.u8 {q6}, [r2], r3 //write 8th 16Byte | ||
| 566 | |||
| 567 | //q2, q3, q4, q5, q0 --> q0~q4 | ||
| 568 | vswp q0, q4 | ||
| 569 | vswp q0, q2 | ||
| 570 | vmov q1, q3 | ||
| 571 | vmov q3, q5 //q0~q4 | ||
| 572 | |||
| 573 | sub r4, #8 | ||
| 574 | cmp r4, #0 | ||
| 575 | bne w16_xy_01_luma_loop | ||
| 576 | pop {r4} | ||
| 577 | WELS_ASM_FUNC_END | ||
| 578 | |||
| 579 | |||
| 580 | WELS_ASM_FUNC_BEGIN McHorVer01WidthEq8_neon | ||
| 581 | push {r4} | ||
| 582 | ldr r4, [sp, #4] | ||
| 583 | |||
| 584 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 585 | pld [r0] | ||
| 586 | pld [r0, r1] | ||
| 587 | vmov.u16 q14, #0x0014 // 20 | ||
| 588 | vld1.u8 {d0}, [r0], r1 //d0=src[-2] | ||
| 589 | vld1.u8 {d1}, [r0], r1 //d1=src[-1] | ||
| 590 | |||
| 591 | pld [r0] | ||
| 592 | pld [r0, r1] | ||
| 593 | vshr.u16 q15, q14, #2 // 5 | ||
| 594 | vld1.u8 {d2}, [r0], r1 //d2=src[0] | ||
| 595 | vld1.u8 {d3}, [r0], r1 //d3=src[1] | ||
| 596 | |||
| 597 | vld1.u8 {d4}, [r0], r1 //d4=src[2] | ||
| 598 | vld1.u8 {d5}, [r0], r1 //d5=src[3] | ||
| 599 | |||
| 600 | w8_xy_01_mc_luma_loop: | ||
| 601 | |||
| 602 | pld [r0] | ||
| 603 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d1, d2, d3, d4, d5, d12, q14, q15 | ||
| 604 | vld1.u8 {d0}, [r0], r1 //read 2nd row | ||
| 605 | vst1.u8 {d12}, [r2], r3 //write 1st 8Byte | ||
| 606 | |||
| 607 | pld [r0] | ||
| 608 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d1, d2, d3, d4, d5, d0, d12, q14, q15 | ||
| 609 | vld1.u8 {d1}, [r0], r1 //read 3rd row | ||
| 610 | vst1.u8 {d12}, [r2], r3 //write 2nd 8Byte | ||
| 611 | |||
| 612 | pld [r0] | ||
| 613 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d2, d3, d4, d5, d0, d1, d12, q14, q15 | ||
| 614 | vld1.u8 {d2}, [r0], r1 //read 4th row | ||
| 615 | vst1.u8 {d12}, [r2], r3 //write 3rd 8Byte | ||
| 616 | |||
| 617 | pld [r0] | ||
| 618 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d3, d4, d5, d0, d1, d2, d12, q14, q15 | ||
| 619 | vld1.u8 {d3}, [r0], r1 //read 5th row | ||
| 620 | vst1.u8 {d12}, [r2], r3 //write 4th 8Byte | ||
| 621 | |||
| 622 | //d4, d5, d0, d1, d2, d3 --> d0, d1, d2, d3, d4, d5 | ||
| 623 | vswp q0, q2 | ||
| 624 | vswp q1, q2 | ||
| 625 | |||
| 626 | sub r4, #4 | ||
| 627 | cmp r4, #0 | ||
| 628 | bne w8_xy_01_mc_luma_loop | ||
| 629 | |||
| 630 | pop {r4} | ||
| 631 | WELS_ASM_FUNC_END | ||
| 632 | |||
| 633 | |||
| 634 | WELS_ASM_FUNC_BEGIN McHorVer01WidthEq4_neon | ||
| 635 | push {r4, r5, r6, r7} | ||
| 636 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 637 | pld [r0] | ||
| 638 | pld [r0, r1] | ||
| 639 | vmov.u16 q14, #0x0014 // 20 | ||
| 640 | ldr r4, [r0], r1 //r4=src[-2] | ||
| 641 | ldr r5, [r0], r1 //r5=src[-1] | ||
| 642 | |||
| 643 | pld [r0] | ||
| 644 | pld [r0, r1] | ||
| 645 | vshr.u16 q15, q14, #2 // 5 | ||
| 646 | ldr r6, [r0], r1 //r6=src[0] | ||
| 647 | ldr r7, [r0], r1 //r7=src[1] | ||
| 648 | |||
| 649 | vmov d0, r4, r5 | ||
| 650 | vmov d1, r5, r6 | ||
| 651 | vmov d2, r6, r7 | ||
| 652 | |||
| 653 | ldr r4, [r0], r1 //r4=src[2] | ||
| 654 | vmov d3, r7, r4 | ||
| 655 | ldr r7, [sp, #16] | ||
| 656 | |||
| 657 | w4_xy_01_mc_luma_loop: | ||
| 658 | |||
| 659 | // pld [r0] | ||
| 660 | //using reserving r4 | ||
| 661 | ldr r5, [r0], r1 //r5=src[3] | ||
| 662 | ldr r6, [r0], r1 //r6=src[0] | ||
| 663 | vmov d4, r4, r5 | ||
| 664 | vmov d5, r5, r6 //reserved r6 | ||
| 665 | |||
| 666 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d0, d1, d2, d3, d4, d5, d12, q14, q15 | ||
| 667 | vmov r4, r5, d12 | ||
| 668 | str r4, [r2], r3 //write 1st 4Byte | ||
| 669 | str r5, [r2], r3 //write 2nd 4Byte | ||
| 670 | |||
| 671 | ldr r5, [r0], r1 //r5=src[1] | ||
| 672 | ldr r4, [r0], r1 //r4=src[2] | ||
| 673 | vmov d0, r6, r5 | ||
| 674 | vmov d1, r5, r4 //reserved r4 | ||
| 675 | |||
| 676 | FILTER_6TAG_8BITS_AVERAGE_WITH_0 d2, d3, d4, d5, d0, d1, d12, q14, q15 | ||
| 677 | vmov r5, r6, d12 | ||
| 678 | str r5, [r2], r3 //write 3rd 4Byte | ||
| 679 | str r6, [r2], r3 //write 4th 4Byte | ||
| 680 | |||
| 681 | //d4, d5, d0, d1 --> d0, d1, d2, d3 | ||
| 682 | vmov q1, q0 | ||
| 683 | vmov q0, q2 | ||
| 684 | |||
| 685 | sub r7, #4 | ||
| 686 | cmp r7, #0 | ||
| 687 | bne w4_xy_01_mc_luma_loop | ||
| 688 | |||
| 689 | pop {r4, r5, r6, r7} | ||
| 690 | WELS_ASM_FUNC_END | ||
| 691 | |||
| 692 | |||
| 693 | WELS_ASM_FUNC_BEGIN McHorVer03WidthEq16_neon | ||
| 694 | push {r4} | ||
| 695 | ldr r4, [sp, #4] | ||
| 696 | |||
| 697 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 698 | pld [r0] | ||
| 699 | pld [r0, r1] | ||
| 700 | vmov.u16 q14, #0x0014 // 20 | ||
| 701 | vld1.u8 {q0}, [r0], r1 //q0=src[-2] | ||
| 702 | vld1.u8 {q1}, [r0], r1 //q1=src[-1] | ||
| 703 | |||
| 704 | pld [r0] | ||
| 705 | pld [r0, r1] | ||
| 706 | vshr.u16 q15, q14, #2 // 5 | ||
| 707 | vld1.u8 {q2}, [r0], r1 //q2=src[0] | ||
| 708 | vld1.u8 {q3}, [r0], r1 //q3=src[1] | ||
| 709 | vld1.u8 {q4}, [r0], r1 //q4=src[2] | ||
| 710 | |||
| 711 | w16_xy_03_luma_loop: | ||
| 712 | |||
| 713 | vld1.u8 {q5}, [r0], r1 //q5=src[3] | ||
| 714 | |||
| 715 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d2, d4, d6, d8, d10, d12, q14, q15 | ||
| 716 | pld [r0] | ||
| 717 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d1, d3, d5, d7, d9, d11, d13, q14, q15 | ||
| 718 | vld1.u8 {q0}, [r0], r1 //read 2nd row | ||
| 719 | vst1.u8 {q6}, [r2], r3 //write 1st 16Byte | ||
| 720 | |||
| 721 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d2, d4, d6, d8, d10, d0, d12, q14, q15 | ||
| 722 | pld [r0] | ||
| 723 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d3, d5, d7, d9, d11, d1, d13, q14, q15 | ||
| 724 | vld1.u8 {q1}, [r0], r1 //read 3rd row | ||
| 725 | vst1.u8 {q6}, [r2], r3 //write 2nd 16Byte | ||
| 726 | |||
| 727 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d4, d6, d8, d10, d0, d2, d12, q14, q15 | ||
| 728 | pld [r0] | ||
| 729 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d5, d7, d9, d11, d1, d3, d13, q14, q15 | ||
| 730 | vld1.u8 {q2}, [r0], r1 //read 4th row | ||
| 731 | vst1.u8 {q6}, [r2], r3 //write 3rd 16Byte | ||
| 732 | |||
| 733 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d6, d8, d10, d0, d2, d4, d12, q14, q15 | ||
| 734 | pld [r0] | ||
| 735 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d7, d9, d11, d1, d3, d5, d13, q14, q15 | ||
| 736 | vld1.u8 {q3}, [r0], r1 //read 5th row | ||
| 737 | vst1.u8 {q6}, [r2], r3 //write 4th 16Byte | ||
| 738 | |||
| 739 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d8, d10, d0, d2, d4, d6, d12, q14, q15 | ||
| 740 | pld [r0] | ||
| 741 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d9, d11, d1, d3, d5, d7, d13, q14, q15 | ||
| 742 | vld1.u8 {q4}, [r0], r1 //read 6th row | ||
| 743 | vst1.u8 {q6}, [r2], r3 //write 5th 16Byte | ||
| 744 | |||
| 745 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d10, d0, d2, d4, d6, d8, d12, q14, q15 | ||
| 746 | pld [r0] | ||
| 747 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d11, d1, d3, d5, d7, d9, d13, q14, q15 | ||
| 748 | vld1.u8 {q5}, [r0], r1 //read 7th row | ||
| 749 | vst1.u8 {q6}, [r2], r3 //write 6th 16Byte | ||
| 750 | |||
| 751 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d2, d4, d6, d8, d10, d12, q14, q15 | ||
| 752 | pld [r0] | ||
| 753 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d1, d3, d5, d7, d9, d11, d13, q14, q15 | ||
| 754 | vld1.u8 {q0}, [r0], r1 //read 8th row | ||
| 755 | vst1.u8 {q6}, [r2], r3 //write 7th 16Byte | ||
| 756 | |||
| 757 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d2, d4, d6, d8, d10, d0, d12, q14, q15 | ||
| 758 | pld [r0] | ||
| 759 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d3, d5, d7, d9, d11, d1, d13, q14, q15 | ||
| 760 | vst1.u8 {q6}, [r2], r3 //write 8th 16Byte | ||
| 761 | |||
| 762 | //q2, q3, q4, q5, q0 --> q0~q4 | ||
| 763 | vswp q0, q4 | ||
| 764 | vswp q0, q2 | ||
| 765 | vmov q1, q3 | ||
| 766 | vmov q3, q5 //q0~q4 | ||
| 767 | |||
| 768 | sub r4, #8 | ||
| 769 | cmp r4, #0 | ||
| 770 | bne w16_xy_03_luma_loop | ||
| 771 | pop {r4} | ||
| 772 | WELS_ASM_FUNC_END | ||
| 773 | |||
| 774 | |||
| 775 | WELS_ASM_FUNC_BEGIN McHorVer03WidthEq8_neon | ||
| 776 | push {r4} | ||
| 777 | ldr r4, [sp, #4] | ||
| 778 | |||
| 779 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 780 | pld [r0] | ||
| 781 | pld [r0, r1] | ||
| 782 | vmov.u16 q14, #0x0014 // 20 | ||
| 783 | vld1.u8 {d0}, [r0], r1 //d0=src[-2] | ||
| 784 | vld1.u8 {d1}, [r0], r1 //d1=src[-1] | ||
| 785 | |||
| 786 | pld [r0] | ||
| 787 | pld [r0, r1] | ||
| 788 | vshr.u16 q15, q14, #2 // 5 | ||
| 789 | vld1.u8 {d2}, [r0], r1 //d2=src[0] | ||
| 790 | vld1.u8 {d3}, [r0], r1 //d3=src[1] | ||
| 791 | |||
| 792 | vld1.u8 {d4}, [r0], r1 //d4=src[2] | ||
| 793 | vld1.u8 {d5}, [r0], r1 //d5=src[3] | ||
| 794 | |||
| 795 | w8_xy_03_mc_luma_loop: | ||
| 796 | |||
| 797 | pld [r0] | ||
| 798 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d1, d2, d3, d4, d5, d12, q14, q15 | ||
| 799 | vld1.u8 {d0}, [r0], r1 //read 2nd row | ||
| 800 | vst1.u8 {d12}, [r2], r3 //write 1st 8Byte | ||
| 801 | |||
| 802 | pld [r0] | ||
| 803 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d1, d2, d3, d4, d5, d0, d12, q14, q15 | ||
| 804 | vld1.u8 {d1}, [r0], r1 //read 3rd row | ||
| 805 | vst1.u8 {d12}, [r2], r3 //write 2nd 8Byte | ||
| 806 | |||
| 807 | pld [r0] | ||
| 808 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d2, d3, d4, d5, d0, d1, d12, q14, q15 | ||
| 809 | vld1.u8 {d2}, [r0], r1 //read 4th row | ||
| 810 | vst1.u8 {d12}, [r2], r3 //write 3rd 8Byte | ||
| 811 | |||
| 812 | pld [r0] | ||
| 813 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d3, d4, d5, d0, d1, d2, d12, q14, q15 | ||
| 814 | vld1.u8 {d3}, [r0], r1 //read 5th row | ||
| 815 | vst1.u8 {d12}, [r2], r3 //write 4th 8Byte | ||
| 816 | |||
| 817 | //d4, d5, d0, d1, d2, d3 --> d0, d1, d2, d3, d4, d5 | ||
| 818 | vswp q0, q2 | ||
| 819 | vswp q1, q2 | ||
| 820 | |||
| 821 | sub r4, #4 | ||
| 822 | cmp r4, #0 | ||
| 823 | bne w8_xy_03_mc_luma_loop | ||
| 824 | |||
| 825 | pop {r4} | ||
| 826 | WELS_ASM_FUNC_END | ||
| 827 | |||
| 828 | |||
| 829 | WELS_ASM_FUNC_BEGIN McHorVer03WidthEq4_neon | ||
| 830 | push {r4, r5, r6, r7} | ||
| 831 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 832 | pld [r0] | ||
| 833 | pld [r0, r1] | ||
| 834 | vmov.u16 q14, #0x0014 // 20 | ||
| 835 | ldr r4, [r0], r1 //r4=src[-2] | ||
| 836 | ldr r5, [r0], r1 //r5=src[-1] | ||
| 837 | |||
| 838 | pld [r0] | ||
| 839 | pld [r0, r1] | ||
| 840 | vshr.u16 q15, q14, #2 // 5 | ||
| 841 | ldr r6, [r0], r1 //r6=src[0] | ||
| 842 | ldr r7, [r0], r1 //r7=src[1] | ||
| 843 | |||
| 844 | vmov d0, r4, r5 | ||
| 845 | vmov d1, r5, r6 | ||
| 846 | vmov d2, r6, r7 | ||
| 847 | |||
| 848 | ldr r4, [r0], r1 //r4=src[2] | ||
| 849 | vmov d3, r7, r4 | ||
| 850 | ldr r7, [sp, #16] | ||
| 851 | |||
| 852 | w4_xy_03_mc_luma_loop: | ||
| 853 | |||
| 854 | // pld [r0] | ||
| 855 | //using reserving r4 | ||
| 856 | ldr r5, [r0], r1 //r5=src[3] | ||
| 857 | ldr r6, [r0], r1 //r6=src[0] | ||
| 858 | vmov d4, r4, r5 | ||
| 859 | vmov d5, r5, r6 //reserved r6 | ||
| 860 | |||
| 861 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d0, d1, d2, d3, d4, d5, d12, q14, q15 | ||
| 862 | vmov r4, r5, d12 | ||
| 863 | str r4, [r2], r3 //write 1st 4Byte | ||
| 864 | str r5, [r2], r3 //write 2nd 4Byte | ||
| 865 | |||
| 866 | ldr r5, [r0], r1 //r5=src[1] | ||
| 867 | ldr r4, [r0], r1 //r4=src[2] | ||
| 868 | vmov d0, r6, r5 | ||
| 869 | vmov d1, r5, r4 //reserved r4 | ||
| 870 | |||
| 871 | FILTER_6TAG_8BITS_AVERAGE_WITH_1 d2, d3, d4, d5, d0, d1, d12, q14, q15 | ||
| 872 | vmov r5, r6, d12 | ||
| 873 | str r5, [r2], r3 //write 3rd 4Byte | ||
| 874 | str r6, [r2], r3 //write 4th 4Byte | ||
| 875 | |||
| 876 | //d4, d5, d0, d1 --> d0, d1, d2, d3 | ||
| 877 | vmov q1, q0 | ||
| 878 | vmov q0, q2 | ||
| 879 | |||
| 880 | sub r7, #4 | ||
| 881 | cmp r7, #0 | ||
| 882 | bne w4_xy_03_mc_luma_loop | ||
| 883 | |||
| 884 | pop {r4, r5, r6, r7} | ||
| 885 | WELS_ASM_FUNC_END | ||
| 886 | |||
| 887 | |||
| 888 | WELS_ASM_FUNC_BEGIN McHorVer02WidthEq16_neon | ||
| 889 | push {r4} | ||
| 890 | ldr r4, [sp, #4] | ||
| 891 | |||
| 892 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 893 | pld [r0] | ||
| 894 | pld [r0, r1] | ||
| 895 | vmov.u16 q14, #0x0014 // 20 | ||
| 896 | vld1.u8 {q0}, [r0], r1 //q0=src[-2] | ||
| 897 | vld1.u8 {q1}, [r0], r1 //q1=src[-1] | ||
| 898 | |||
| 899 | pld [r0] | ||
| 900 | pld [r0, r1] | ||
| 901 | vshr.u16 q15, q14, #2 // 5 | ||
| 902 | vld1.u8 {q2}, [r0], r1 //q2=src[0] | ||
| 903 | vld1.u8 {q3}, [r0], r1 //q3=src[1] | ||
| 904 | vld1.u8 {q4}, [r0], r1 //q4=src[2] | ||
| 905 | |||
| 906 | w16_v_mc_luma_loop: | ||
| 907 | |||
| 908 | vld1.u8 {q5}, [r0], r1 //q5=src[3] | ||
| 909 | |||
| 910 | FILTER_6TAG_8BITS d0, d2, d4, d6, d8, d10, d12, q14, q15 | ||
| 911 | pld [r0] | ||
| 912 | FILTER_6TAG_8BITS d1, d3, d5, d7, d9, d11, d13, q14, q15 | ||
| 913 | vld1.u8 {q0}, [r0], r1 //read 2nd row | ||
| 914 | vst1.u8 {q6}, [r2], r3 //write 1st 16Byte | ||
| 915 | |||
| 916 | FILTER_6TAG_8BITS d2, d4, d6, d8, d10, d0, d12, q14, q15 | ||
| 917 | pld [r0] | ||
| 918 | FILTER_6TAG_8BITS d3, d5, d7, d9, d11, d1, d13, q14, q15 | ||
| 919 | vld1.u8 {q1}, [r0], r1 //read 3rd row | ||
| 920 | vst1.u8 {q6}, [r2], r3 //write 2nd 16Byte | ||
| 921 | |||
| 922 | FILTER_6TAG_8BITS d4, d6, d8, d10, d0, d2, d12, q14, q15 | ||
| 923 | pld [r0] | ||
| 924 | FILTER_6TAG_8BITS d5, d7, d9, d11, d1, d3, d13, q14, q15 | ||
| 925 | vld1.u8 {q2}, [r0], r1 //read 4th row | ||
| 926 | vst1.u8 {q6}, [r2], r3 //write 3rd 16Byte | ||
| 927 | |||
| 928 | FILTER_6TAG_8BITS d6, d8, d10, d0, d2, d4, d12, q14, q15 | ||
| 929 | pld [r0] | ||
| 930 | FILTER_6TAG_8BITS d7, d9, d11, d1, d3, d5, d13, q14, q15 | ||
| 931 | vld1.u8 {q3}, [r0], r1 //read 5th row | ||
| 932 | vst1.u8 {q6}, [r2], r3 //write 4th 16Byte | ||
| 933 | |||
| 934 | FILTER_6TAG_8BITS d8, d10, d0, d2, d4, d6, d12, q14, q15 | ||
| 935 | pld [r0] | ||
| 936 | FILTER_6TAG_8BITS d9, d11, d1, d3, d5, d7, d13, q14, q15 | ||
| 937 | vld1.u8 {q4}, [r0], r1 //read 6th row | ||
| 938 | vst1.u8 {q6}, [r2], r3 //write 5th 16Byte | ||
| 939 | |||
| 940 | FILTER_6TAG_8BITS d10, d0, d2, d4, d6, d8, d12, q14, q15 | ||
| 941 | pld [r0] | ||
| 942 | FILTER_6TAG_8BITS d11, d1, d3, d5, d7, d9, d13, q14, q15 | ||
| 943 | vld1.u8 {q5}, [r0], r1 //read 7th row | ||
| 944 | vst1.u8 {q6}, [r2], r3 //write 6th 16Byte | ||
| 945 | |||
| 946 | FILTER_6TAG_8BITS d0, d2, d4, d6, d8, d10, d12, q14, q15 | ||
| 947 | pld [r0] | ||
| 948 | FILTER_6TAG_8BITS d1, d3, d5, d7, d9, d11, d13, q14, q15 | ||
| 949 | vld1.u8 {q0}, [r0], r1 //read 8th row | ||
| 950 | vst1.u8 {q6}, [r2], r3 //write 7th 16Byte | ||
| 951 | |||
| 952 | FILTER_6TAG_8BITS d2, d4, d6, d8, d10, d0, d12, q14, q15 | ||
| 953 | pld [r0] | ||
| 954 | FILTER_6TAG_8BITS d3, d5, d7, d9, d11, d1, d13, q14, q15 | ||
| 955 | vst1.u8 {q6}, [r2], r3 //write 8th 16Byte | ||
| 956 | |||
| 957 | //q2, q3, q4, q5, q0 --> q0~q4 | ||
| 958 | vswp q0, q4 | ||
| 959 | vswp q0, q2 | ||
| 960 | vmov q1, q3 | ||
| 961 | vmov q3, q5 //q0~q4 | ||
| 962 | |||
| 963 | sub r4, #8 | ||
| 964 | cmp r4, #0 | ||
| 965 | bne w16_v_mc_luma_loop | ||
| 966 | pop {r4} | ||
| 967 | WELS_ASM_FUNC_END | ||
| 968 | |||
| 969 | |||
| 970 | WELS_ASM_FUNC_BEGIN McHorVer02WidthEq8_neon | ||
| 971 | push {r4} | ||
| 972 | ldr r4, [sp, #4] | ||
| 973 | |||
| 974 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 975 | pld [r0] | ||
| 976 | pld [r0, r1] | ||
| 977 | vmov.u16 q14, #0x0014 // 20 | ||
| 978 | vld1.u8 {d0}, [r0], r1 //d0=src[-2] | ||
| 979 | vld1.u8 {d1}, [r0], r1 //d1=src[-1] | ||
| 980 | |||
| 981 | pld [r0] | ||
| 982 | pld [r0, r1] | ||
| 983 | vshr.u16 q15, q14, #2 // 5 | ||
| 984 | vld1.u8 {d2}, [r0], r1 //d2=src[0] | ||
| 985 | vld1.u8 {d3}, [r0], r1 //d3=src[1] | ||
| 986 | |||
| 987 | vld1.u8 {d4}, [r0], r1 //d4=src[2] | ||
| 988 | vld1.u8 {d5}, [r0], r1 //d5=src[3] | ||
| 989 | |||
| 990 | w8_v_mc_luma_loop: | ||
| 991 | |||
| 992 | pld [r0] | ||
| 993 | FILTER_6TAG_8BITS d0, d1, d2, d3, d4, d5, d12, q14, q15 | ||
| 994 | vld1.u8 {d0}, [r0], r1 //read 2nd row | ||
| 995 | vst1.u8 {d12}, [r2], r3 //write 1st 8Byte | ||
| 996 | |||
| 997 | pld [r0] | ||
| 998 | FILTER_6TAG_8BITS d1, d2, d3, d4, d5, d0, d12, q14, q15 | ||
| 999 | vld1.u8 {d1}, [r0], r1 //read 3rd row | ||
| 1000 | vst1.u8 {d12}, [r2], r3 //write 2nd 8Byte | ||
| 1001 | |||
| 1002 | pld [r0] | ||
| 1003 | FILTER_6TAG_8BITS d2, d3, d4, d5, d0, d1, d12, q14, q15 | ||
| 1004 | vld1.u8 {d2}, [r0], r1 //read 4th row | ||
| 1005 | vst1.u8 {d12}, [r2], r3 //write 3rd 8Byte | ||
| 1006 | |||
| 1007 | pld [r0] | ||
| 1008 | FILTER_6TAG_8BITS d3, d4, d5, d0, d1, d2, d12, q14, q15 | ||
| 1009 | vld1.u8 {d3}, [r0], r1 //read 5th row | ||
| 1010 | vst1.u8 {d12}, [r2], r3 //write 4th 8Byte | ||
| 1011 | |||
| 1012 | //d4, d5, d0, d1, d2, d3 --> d0, d1, d2, d3, d4, d5 | ||
| 1013 | vswp q0, q2 | ||
| 1014 | vswp q1, q2 | ||
| 1015 | |||
| 1016 | sub r4, #4 | ||
| 1017 | cmp r4, #0 | ||
| 1018 | bne w8_v_mc_luma_loop | ||
| 1019 | |||
| 1020 | pop {r4} | ||
| 1021 | WELS_ASM_FUNC_END | ||
| 1022 | |||
| 1023 | |||
| 1024 | WELS_ASM_FUNC_BEGIN McHorVer02WidthEq4_neon | ||
| 1025 | push {r4, r5, r6, r7} | ||
| 1026 | sub r0, r0, r1, lsl #1 //src[-2*src_stride] | ||
| 1027 | pld [r0] | ||
| 1028 | pld [r0, r1] | ||
| 1029 | vmov.u16 q14, #0x0014 // 20 | ||
| 1030 | ldr r4, [r0], r1 //r4=src[-2] | ||
| 1031 | ldr r5, [r0], r1 //r5=src[-1] | ||
| 1032 | |||
| 1033 | pld [r0] | ||
| 1034 | pld [r0, r1] | ||
| 1035 | vshr.u16 q15, q14, #2 // 5 | ||
| 1036 | ldr r6, [r0], r1 //r6=src[0] | ||
| 1037 | ldr r7, [r0], r1 //r7=src[1] | ||
| 1038 | |||
| 1039 | vmov d0, r4, r5 | ||
| 1040 | vmov d1, r5, r6 | ||
| 1041 | vmov d2, r6, r7 | ||
| 1042 | |||
| 1043 | ldr r4, [r0], r1 //r4=src[2] | ||
| 1044 | vmov d3, r7, r4 | ||
| 1045 | ldr r7, [sp, #16] | ||
| 1046 | |||
| 1047 | w4_v_mc_luma_loop: | ||
| 1048 | |||
| 1049 | // pld [r0] | ||
| 1050 | //using reserving r4 | ||
| 1051 | ldr r5, [r0], r1 //r5=src[3] | ||
| 1052 | ldr r6, [r0], r1 //r6=src[0] | ||
| 1053 | vmov d4, r4, r5 | ||
| 1054 | vmov d5, r5, r6 //reserved r6 | ||
| 1055 | |||
| 1056 | FILTER_6TAG_8BITS d0, d1, d2, d3, d4, d5, d12, q14, q15 | ||
| 1057 | vmov r4, r5, d12 | ||
| 1058 | str r4, [r2], r3 //write 1st 4Byte | ||
| 1059 | str r5, [r2], r3 //write 2nd 4Byte | ||
| 1060 | |||
| 1061 | ldr r5, [r0], r1 //r5=src[1] | ||
| 1062 | ldr r4, [r0], r1 //r4=src[2] | ||
| 1063 | vmov d0, r6, r5 | ||
| 1064 | vmov d1, r5, r4 //reserved r4 | ||
| 1065 | |||
| 1066 | FILTER_6TAG_8BITS d2, d3, d4, d5, d0, d1, d12, q14, q15 | ||
| 1067 | vmov r5, r6, d12 | ||
| 1068 | str r5, [r2], r3 //write 3rd 4Byte | ||
| 1069 | str r6, [r2], r3 //write 4th 4Byte | ||
| 1070 | |||
| 1071 | //d4, d5, d0, d1 --> d0, d1, d2, d3 | ||
| 1072 | vmov q1, q0 | ||
| 1073 | vmov q0, q2 | ||
| 1074 | |||
| 1075 | sub r7, #4 | ||
| 1076 | cmp r7, #0 | ||
| 1077 | bne w4_v_mc_luma_loop | ||
| 1078 | |||
| 1079 | pop {r4, r5, r6, r7} | ||
| 1080 | WELS_ASM_FUNC_END | ||
| 1081 | |||
| 1082 | |||
| 1083 | WELS_ASM_FUNC_BEGIN McHorVer22WidthEq16_neon | ||
| 1084 | push {r4} | ||
| 1085 | ldr r4, [sp, #4] | ||
| 1086 | |||
| 1087 | sub r0, #2 //src[-2] | ||
| 1088 | sub r0, r0, r1, lsl #1 //src[-2*src_stride-2] | ||
| 1089 | pld [r0] | ||
| 1090 | pld [r0, r1] | ||
| 1091 | |||
| 1092 | vmov.u16 q14, #0x0014 // 20 | ||
| 1093 | vld1.u8 {d0-d2}, [r0], r1 //use 21(16+5), =src[-2] | ||
| 1094 | vld1.u8 {d3-d5}, [r0], r1 //use 21(16+5), =src[-1] | ||
| 1095 | |||
| 1096 | pld [r0] | ||
| 1097 | pld [r0, r1] | ||
| 1098 | vshr.u16 q15, q14, #2 // 5 | ||
| 1099 | |||
| 1100 | vld1.u8 {d6-d8}, [r0], r1 //use 21(16+5), =src[0] | ||
| 1101 | vld1.u8 {d9-d11}, [r0], r1 //use 21(16+5), =src[1] | ||
| 1102 | pld [r0] | ||
| 1103 | pld [r0, r1] | ||
| 1104 | vld1.u8 {d12-d14}, [r0], r1 //use 21(16+5), =src[2] | ||
| 1105 | |||
| 1106 | w16_hv_mc_luma_loop: | ||
| 1107 | |||
| 1108 | vld1.u8 {d15-d17}, [r0], r1 //use 21(16+5), =src[3] | ||
| 1109 | //the 1st row | ||
| 1110 | pld [r0] | ||
| 1111 | // vertical filtered into q9/q10 | ||
| 1112 | FILTER_6TAG_8BITS_TO_16BITS d0, d3, d6, d9, d12, d15, q9, q14, q15 // 8 avail | ||
| 1113 | FILTER_6TAG_8BITS_TO_16BITS d1, d4, d7,d10, d13, d16,q10, q14, q15 // 8 avail | ||
| 1114 | // horizon filtered | ||
| 1115 | UNPACK_2_16BITS_TO_ABC q9, q10, q11, q12, q13 | ||
| 1116 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d0 //output to q0[0] | ||
| 1117 | |||
| 1118 | // vertical filtered into q10/q11 | ||
| 1119 | FILTER_6TAG_8BITS_TO_16BITS d2, d5, d8,d11, d14, d17,q11, q14, q15 // only 5 avail | ||
| 1120 | // horizon filtered | ||
| 1121 | UNPACK_2_16BITS_TO_ABC q10, q11, q9, q12, q13 | ||
| 1122 | FILTER_3_IN_16BITS_TO_8BITS q9, q12, q13, d1 //output to q0[1] | ||
| 1123 | vst1.u8 {q0}, [r2], r3 //write 16Byte | ||
| 1124 | |||
| 1125 | |||
| 1126 | vld1.u8 {d0-d2}, [r0], r1 //read 2nd row | ||
| 1127 | //the 2nd row | ||
| 1128 | pld [r0] | ||
| 1129 | // vertical filtered into q9/q10 | ||
| 1130 | FILTER_6TAG_8BITS_TO_16BITS d3, d6, d9, d12, d15, d0, q9, q14, q15 // 8 avail | ||
| 1131 | FILTER_6TAG_8BITS_TO_16BITS d4, d7,d10, d13, d16, d1,q10, q14, q15 // 8 avail | ||
| 1132 | // horizon filtered | ||
| 1133 | UNPACK_2_16BITS_TO_ABC q9, q10, q11, q12, q13 | ||
| 1134 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d3 //output to d3 | ||
| 1135 | |||
| 1136 | // vertical filtered into q10/q11 | ||
| 1137 | FILTER_6TAG_8BITS_TO_16BITS d5, d8,d11, d14, d17, d2,q11, q14, q15 // only 5 avail | ||
| 1138 | // horizon filtered | ||
| 1139 | UNPACK_2_16BITS_TO_ABC q10, q11, q9, q12, q13 | ||
| 1140 | FILTER_3_IN_16BITS_TO_8BITS q9, q12, q13, d4 //output to d4 | ||
| 1141 | |||
| 1142 | vst1.u8 {d3, d4}, [r2], r3 //write 16Byte | ||
| 1143 | |||
| 1144 | vld1.u8 {d3-d5}, [r0], r1 //read 3rd row | ||
| 1145 | //the 3rd row | ||
| 1146 | pld [r0] | ||
| 1147 | // vertical filtered into q9/q10 | ||
| 1148 | FILTER_6TAG_8BITS_TO_16BITS d6, d9, d12, d15, d0, d3, q9, q14, q15 // 8 avail | ||
| 1149 | FILTER_6TAG_8BITS_TO_16BITS d7,d10, d13, d16, d1, d4,q10, q14, q15 // 8 avail | ||
| 1150 | // horizon filtered | ||
| 1151 | UNPACK_2_16BITS_TO_ABC q9, q10, q11, q12, q13 | ||
| 1152 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d6 //output to d6 | ||
| 1153 | |||
| 1154 | // vertical filtered into q10/q11 | ||
| 1155 | FILTER_6TAG_8BITS_TO_16BITS d8,d11, d14, d17, d2, d5,q11, q14, q15 // only 5 avail | ||
| 1156 | // horizon filtered | ||
| 1157 | UNPACK_2_16BITS_TO_ABC q10, q11, q9, q12, q13 | ||
| 1158 | FILTER_3_IN_16BITS_TO_8BITS q9, q12, q13, d7 //output to d7 | ||
| 1159 | vst1.u8 {d6, d7}, [r2], r3 //write 16Byte | ||
| 1160 | |||
| 1161 | vld1.u8 {d6-d8}, [r0], r1 //read 4th row | ||
| 1162 | //the 4th row | ||
| 1163 | pld [r0] | ||
| 1164 | // vertical filtered into q9/q10 | ||
| 1165 | FILTER_6TAG_8BITS_TO_16BITS d9, d12, d15, d0, d3, d6, q9, q14, q15 // 8 avail | ||
| 1166 | FILTER_6TAG_8BITS_TO_16BITS d10, d13, d16, d1, d4, d7,q10, q14, q15 // 8 avail | ||
| 1167 | // horizon filtered | ||
| 1168 | UNPACK_2_16BITS_TO_ABC q9, q10, q11, q12, q13 | ||
| 1169 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d9 //output to d9 | ||
| 1170 | // vertical filtered into q10/q11 | ||
| 1171 | FILTER_6TAG_8BITS_TO_16BITS d11, d14, d17, d2, d5, d8,q11, q14, q15 // only 5 avail | ||
| 1172 | // horizon filtered | ||
| 1173 | UNPACK_2_16BITS_TO_ABC q10, q11, q9, q12, q13 | ||
| 1174 | FILTER_3_IN_16BITS_TO_8BITS q9, q12, q13, d10 //output to d10 | ||
| 1175 | vst1.u8 {d9, d10}, [r2], r3 //write 16Byte | ||
| 1176 | |||
| 1177 | //d12~d17(q6~q8), d0~d8(q0~q3+d8), --> d0~d14 | ||
| 1178 | vswp q0, q6 | ||
| 1179 | vswp q6, q3 | ||
| 1180 | vmov q5, q2 | ||
| 1181 | vmov q2, q8 | ||
| 1182 | |||
| 1183 | vmov d20,d8 | ||
| 1184 | vmov q4, q1 | ||
| 1185 | vmov q1, q7 | ||
| 1186 | vmov d14,d20 | ||
| 1187 | |||
| 1188 | sub r4, #4 | ||
| 1189 | cmp r4, #0 | ||
| 1190 | bne w16_hv_mc_luma_loop | ||
| 1191 | pop {r4} | ||
| 1192 | WELS_ASM_FUNC_END | ||
| 1193 | |||
| 1194 | |||
| 1195 | WELS_ASM_FUNC_BEGIN McHorVer22WidthEq8_neon | ||
| 1196 | push {r4} | ||
| 1197 | ldr r4, [sp, #4] | ||
| 1198 | |||
| 1199 | sub r0, #2 //src[-2] | ||
| 1200 | sub r0, r0, r1, lsl #1 //src[-2*src_stride-2] | ||
| 1201 | pld [r0] | ||
| 1202 | pld [r0, r1] | ||
| 1203 | |||
| 1204 | vmov.u16 q14, #0x0014 // 20 | ||
| 1205 | vld1.u8 {q0}, [r0], r1 //use 13(8+5), =src[-2] | ||
| 1206 | vld1.u8 {q1}, [r0], r1 //use 13(8+5), =src[-1] | ||
| 1207 | |||
| 1208 | pld [r0] | ||
| 1209 | pld [r0, r1] | ||
| 1210 | vshr.u16 q15, q14, #2 // 5 | ||
| 1211 | |||
| 1212 | vld1.u8 {q2}, [r0], r1 //use 13(8+5), =src[0] | ||
| 1213 | vld1.u8 {q3}, [r0], r1 //use 13(8+5), =src[1] | ||
| 1214 | pld [r0] | ||
| 1215 | pld [r0, r1] | ||
| 1216 | vld1.u8 {q4}, [r0], r1 //use 13(8+5), =src[2] | ||
| 1217 | |||
| 1218 | w8_hv_mc_luma_loop: | ||
| 1219 | |||
| 1220 | vld1.u8 {q5}, [r0], r1 //use 13(8+5), =src[3] | ||
| 1221 | //the 1st row | ||
| 1222 | pld [r0] | ||
| 1223 | // vertical filtered into q6/q7 | ||
| 1224 | FILTER_6TAG_8BITS_TO_16BITS d0, d2, d4, d6, d8, d10, q6, q14, q15 // 8 avail | ||
| 1225 | FILTER_6TAG_8BITS_TO_16BITS d1, d3, d5, d7, d9, d11, q7, q14, q15 // 5 avail | ||
| 1226 | // horizon filtered | ||
| 1227 | UNPACK_2_16BITS_TO_ABC q6, q7, q11, q12, q13 | ||
| 1228 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d12 //output to q6[0] | ||
| 1229 | vst1.u8 d12, [r2], r3 //write 8Byte | ||
| 1230 | |||
| 1231 | vld1.u8 {q0}, [r0], r1 //read 2nd row | ||
| 1232 | //the 2nd row | ||
| 1233 | pld [r0] | ||
| 1234 | // vertical filtered into q6/q7 | ||
| 1235 | FILTER_6TAG_8BITS_TO_16BITS d2, d4, d6, d8, d10, d0, q6, q14, q15 // 8 avail | ||
| 1236 | FILTER_6TAG_8BITS_TO_16BITS d3, d5, d7, d9, d11, d1, q7, q14, q15 // 5 avail | ||
| 1237 | // horizon filtered | ||
| 1238 | UNPACK_2_16BITS_TO_ABC q6, q7, q11, q12, q13 | ||
| 1239 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d12 //output to q6[0] | ||
| 1240 | vst1.u8 d12, [r2], r3 //write 8Byte | ||
| 1241 | |||
| 1242 | vld1.u8 {q1}, [r0], r1 //read 3rd row | ||
| 1243 | //the 3rd row | ||
| 1244 | pld [r0] | ||
| 1245 | // vertical filtered into q6/q7 | ||
| 1246 | FILTER_6TAG_8BITS_TO_16BITS d4, d6, d8, d10, d0, d2, q6, q14, q15 // 8 avail | ||
| 1247 | FILTER_6TAG_8BITS_TO_16BITS d5, d7, d9, d11, d1, d3, q7, q14, q15 // 5 avail | ||
| 1248 | // horizon filtered | ||
| 1249 | UNPACK_2_16BITS_TO_ABC q6, q7, q11, q12, q13 | ||
| 1250 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d12 //output to q6[0] | ||
| 1251 | vst1.u8 d12, [r2], r3 //write 8Byte | ||
| 1252 | |||
| 1253 | vld1.u8 {q2}, [r0], r1 //read 4th row | ||
| 1254 | //the 4th row | ||
| 1255 | pld [r0] | ||
| 1256 | // vertical filtered into q6/q7 | ||
| 1257 | FILTER_6TAG_8BITS_TO_16BITS d6, d8, d10, d0, d2, d4, q6, q14, q15 // 8 avail | ||
| 1258 | FILTER_6TAG_8BITS_TO_16BITS d7, d9, d11, d1, d3, d5, q7, q14, q15 // 5 avail | ||
| 1259 | // horizon filtered | ||
| 1260 | UNPACK_2_16BITS_TO_ABC q6, q7, q11, q12, q13 | ||
| 1261 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d12 //output to q6[0] | ||
| 1262 | vst1.u8 d12, [r2], r3 //write 8Byte | ||
| 1263 | |||
| 1264 | //q4~q5, q0~q2, --> q0~q4 | ||
| 1265 | vswp q0, q4 | ||
| 1266 | vswp q2, q4 | ||
| 1267 | vmov q3, q1 | ||
| 1268 | vmov q1, q5 | ||
| 1269 | |||
| 1270 | sub r4, #4 | ||
| 1271 | cmp r4, #0 | ||
| 1272 | bne w8_hv_mc_luma_loop | ||
| 1273 | pop {r4} | ||
| 1274 | WELS_ASM_FUNC_END | ||
| 1275 | |||
| 1276 | |||
| 1277 | WELS_ASM_FUNC_BEGIN McHorVer22WidthEq4_neon | ||
| 1278 | push {r4 ,r5, r6} | ||
| 1279 | ldr r6, [sp, #12] | ||
| 1280 | |||
| 1281 | sub r0, #2 //src[-2] | ||
| 1282 | sub r0, r0, r1, lsl #1 //src[-2*src_stride-2] | ||
| 1283 | pld [r0] | ||
| 1284 | pld [r0, r1] | ||
| 1285 | |||
| 1286 | vmov.u16 q14, #0x0014 // 20 | ||
| 1287 | vld1.u8 {q0}, [r0], r1 //use 9(4+5), =src[-2] | ||
| 1288 | vld1.u8 {q1}, [r0], r1 //use 9(4+5), =src[-1] | ||
| 1289 | |||
| 1290 | pld [r0] | ||
| 1291 | pld [r0, r1] | ||
| 1292 | vshr.u16 q15, q14, #2 // 5 | ||
| 1293 | |||
| 1294 | vld1.u8 {q2}, [r0], r1 //use 9(4+5), =src[0] | ||
| 1295 | vld1.u8 {q3}, [r0], r1 //use 9(4+5), =src[1] | ||
| 1296 | pld [r0] | ||
| 1297 | pld [r0, r1] | ||
| 1298 | vld1.u8 {q4}, [r0], r1 //use 9(4+5), =src[2] | ||
| 1299 | |||
| 1300 | w4_hv_mc_luma_loop: | ||
| 1301 | |||
| 1302 | vld1.u8 {q5}, [r0], r1 //use 9(4+5), =src[3] | ||
| 1303 | vld1.u8 {q6}, [r0], r1 //use 9(4+5), =src[4] | ||
| 1304 | |||
| 1305 | //the 1st&2nd row | ||
| 1306 | pld [r0] | ||
| 1307 | pld [r0, r1] | ||
| 1308 | // vertical filtered | ||
| 1309 | FILTER_6TAG_8BITS_TO_16BITS d0, d2, d4, d6, d8, d10, q7, q14, q15 // 8 avail | ||
| 1310 | FILTER_6TAG_8BITS_TO_16BITS d1, d3, d5, d7, d9, d11, q8, q14, q15 // 1 avail | ||
| 1311 | |||
| 1312 | FILTER_6TAG_8BITS_TO_16BITS d2, d4, d6, d8,d10, d12, q9, q14, q15 // 8 avail | ||
| 1313 | FILTER_6TAG_8BITS_TO_16BITS d3, d5, d7, d9,d11, d13,q10, q14, q15 // 1 avail | ||
| 1314 | // horizon filtered | ||
| 1315 | UNPACK_2_16BITS_TO_ABC q7, q8, q11, q12, q13 //4 avail | ||
| 1316 | UNPACK_2_16BITS_TO_ABC q9,q10, q0, q7, q8 //4 avail | ||
| 1317 | |||
| 1318 | vmov d23, d0 | ||
| 1319 | vmov d25, d14 | ||
| 1320 | vmov d27, d16 | ||
| 1321 | |||
| 1322 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d22 //output to q11[0] | ||
| 1323 | vmov r4 ,r5, d22 | ||
| 1324 | str r4, [r2], r3 //write 4Byte | ||
| 1325 | str r5, [r2], r3 //write 4Byte | ||
| 1326 | |||
| 1327 | //the 3rd&4th row | ||
| 1328 | vld1.u8 {q0}, [r0], r1 //use 9(4+5), =src[3] | ||
| 1329 | vld1.u8 {q1}, [r0], r1 //use 9(4+5), =src[4] | ||
| 1330 | pld [r0] | ||
| 1331 | pld [r0, r1] | ||
| 1332 | // vertical filtered | ||
| 1333 | FILTER_6TAG_8BITS_TO_16BITS d4, d6, d8, d10, d12, d0, q7, q14, q15 // 8 avail | ||
| 1334 | FILTER_6TAG_8BITS_TO_16BITS d5, d7, d9, d11, d13, d1, q8, q14, q15 // 1 avail | ||
| 1335 | |||
| 1336 | FILTER_6TAG_8BITS_TO_16BITS d6, d8,d10, d12, d0, d2, q9, q14, q15 // 8 avail | ||
| 1337 | FILTER_6TAG_8BITS_TO_16BITS d7, d9,d11, d13, d1, d3,q10, q14, q15 // 1 avail | ||
| 1338 | // horizon filtered | ||
| 1339 | UNPACK_2_16BITS_TO_ABC q7, q8, q11, q12, q13 //4 avail | ||
| 1340 | UNPACK_2_16BITS_TO_ABC q9,q10, q2, q7, q8 //4 avail | ||
| 1341 | |||
| 1342 | vmov d23, d4 | ||
| 1343 | vmov d25, d14 | ||
| 1344 | vmov d27, d16 | ||
| 1345 | |||
| 1346 | FILTER_3_IN_16BITS_TO_8BITS q11, q12, q13, d22 //output to q11[0] | ||
| 1347 | vmov r4 ,r5, d22 | ||
| 1348 | str r4, [r2], r3 //write 4Byte | ||
| 1349 | str r5, [r2], r3 //write 4Byte | ||
| 1350 | |||
| 1351 | //q4~q6, q0~q1, --> q0~q4 | ||
| 1352 | vswp q4, q0 | ||
| 1353 | vmov q3, q4 | ||
| 1354 | vmov q4, q1 | ||
| 1355 | vmov q1, q5 | ||
| 1356 | vmov q2, q6 | ||
| 1357 | |||
| 1358 | sub r6, #4 | ||
| 1359 | cmp r6, #0 | ||
| 1360 | bne w4_hv_mc_luma_loop | ||
| 1361 | |||
| 1362 | pop {r4, r5, r6} | ||
| 1363 | WELS_ASM_FUNC_END | ||
| 1364 | |||
| 1365 | |||
| 1366 | WELS_ASM_FUNC_BEGIN McCopyWidthEq16_neon | ||
| 1367 | push {r4} | ||
| 1368 | ldr r4, [sp, #4] | ||
| 1369 | w16_copy_loop: | ||
| 1370 | vld1.u8 {q0}, [r0], r1 | ||
| 1371 | sub r4, #2 | ||
| 1372 | vld1.u8 {q1}, [r0], r1 | ||
| 1373 | vst1.u8 {q0}, [r2], r3 | ||
| 1374 | cmp r4, #0 | ||
| 1375 | vst1.u8 {q1}, [r2], r3 | ||
| 1376 | bne w16_copy_loop | ||
| 1377 | |||
| 1378 | pop {r4} | ||
| 1379 | WELS_ASM_FUNC_END | ||
| 1380 | |||
| 1381 | |||
| 1382 | WELS_ASM_FUNC_BEGIN McCopyWidthEq8_neon | ||
| 1383 | push {r4} | ||
| 1384 | ldr r4, [sp, #4] | ||
| 1385 | w8_copy_loop: | ||
| 1386 | vld1.u8 {d0}, [r0], r1 | ||
| 1387 | vld1.u8 {d1}, [r0], r1 | ||
| 1388 | vst1.u8 {d0}, [r2], r3 | ||
| 1389 | vst1.u8 {d1}, [r2], r3 | ||
| 1390 | sub r4, #2 | ||
| 1391 | cmp r4, #0 | ||
| 1392 | bne w8_copy_loop | ||
| 1393 | |||
| 1394 | pop {r4} | ||
| 1395 | WELS_ASM_FUNC_END | ||
| 1396 | |||
| 1397 | |||
| 1398 | WELS_ASM_FUNC_BEGIN McCopyWidthEq4_neon | ||
| 1399 | push {r4, r5, r6} | ||
| 1400 | ldr r4, [sp, #12] | ||
| 1401 | w4_copy_loop: | ||
| 1402 | ldr r5, [r0], r1 | ||
| 1403 | ldr r6, [r0], r1 | ||
| 1404 | str r5, [r2], r3 | ||
| 1405 | str r6, [r2], r3 | ||
| 1406 | |||
| 1407 | sub r4, #2 | ||
| 1408 | cmp r4, #0 | ||
| 1409 | bne w4_copy_loop | ||
| 1410 | |||
| 1411 | pop {r4, r5, r6} | ||
| 1412 | WELS_ASM_FUNC_END | ||
| 1413 | |||
| 1414 | |||
| 1415 | WELS_ASM_FUNC_BEGIN PixelAvgWidthEq16_neon | ||
| 1416 | push {r4} | ||
| 1417 | ldr r4, [sp, #4] | ||
| 1418 | w16_pix_avg_loop: | ||
| 1419 | vld1.u8 {q0}, [r2]! | ||
| 1420 | vld1.u8 {q1}, [r3]! | ||
| 1421 | vld1.u8 {q2}, [r2]! | ||
| 1422 | vld1.u8 {q3}, [r3]! | ||
| 1423 | |||
| 1424 | vld1.u8 {q4}, [r2]! | ||
| 1425 | vld1.u8 {q5}, [r3]! | ||
| 1426 | vld1.u8 {q6}, [r2]! | ||
| 1427 | vld1.u8 {q7}, [r3]! | ||
| 1428 | |||
| 1429 | AVERAGE_TWO_8BITS d0, d0, d2 | ||
| 1430 | AVERAGE_TWO_8BITS d1, d1, d3 | ||
| 1431 | vst1.u8 {q0}, [r0], r1 | ||
| 1432 | |||
| 1433 | AVERAGE_TWO_8BITS d4, d4, d6 | ||
| 1434 | AVERAGE_TWO_8BITS d5, d5, d7 | ||
| 1435 | vst1.u8 {q2}, [r0], r1 | ||
| 1436 | |||
| 1437 | AVERAGE_TWO_8BITS d8, d8, d10 | ||
| 1438 | AVERAGE_TWO_8BITS d9, d9, d11 | ||
| 1439 | vst1.u8 {q4}, [r0], r1 | ||
| 1440 | |||
| 1441 | AVERAGE_TWO_8BITS d12, d12, d14 | ||
| 1442 | AVERAGE_TWO_8BITS d13, d13, d15 | ||
| 1443 | vst1.u8 {q6}, [r0], r1 | ||
| 1444 | |||
| 1445 | sub r4, #4 | ||
| 1446 | cmp r4, #0 | ||
| 1447 | bne w16_pix_avg_loop | ||
| 1448 | |||
| 1449 | pop {r4} | ||
| 1450 | WELS_ASM_FUNC_END | ||
| 1451 | |||
| 1452 | |||
| 1453 | WELS_ASM_FUNC_BEGIN PixelAvgWidthEq8_neon | ||
| 1454 | push {r4, r5} | ||
| 1455 | ldr r4, [sp, #8] | ||
| 1456 | mov r5, #16 | ||
| 1457 | w8_pix_avg_loop: | ||
| 1458 | |||
| 1459 | vld1.u8 {d0}, [r2], r5 | ||
| 1460 | vld1.u8 {d2}, [r3], r5 | ||
| 1461 | vld1.u8 {d1}, [r2], r5 | ||
| 1462 | vld1.u8 {d3}, [r3], r5 | ||
| 1463 | |||
| 1464 | AVERAGE_TWO_8BITS d0, d0, d2 | ||
| 1465 | AVERAGE_TWO_8BITS d1, d1, d3 | ||
| 1466 | vst1.u8 {d0}, [r0], r1 | ||
| 1467 | vst1.u8 {d1}, [r0], r1 | ||
| 1468 | |||
| 1469 | vld1.u8 {d4}, [r2], r5 | ||
| 1470 | vld1.u8 {d6}, [r3], r5 | ||
| 1471 | vld1.u8 {d5}, [r2], r5 | ||
| 1472 | vld1.u8 {d7}, [r3], r5 | ||
| 1473 | |||
| 1474 | AVERAGE_TWO_8BITS d4, d4, d6 | ||
| 1475 | AVERAGE_TWO_8BITS d5, d5, d7 | ||
| 1476 | vst1.u8 {d4}, [r0], r1 | ||
| 1477 | vst1.u8 {d5}, [r0], r1 | ||
| 1478 | |||
| 1479 | sub r4, #4 | ||
| 1480 | cmp r4, #0 | ||
| 1481 | bne w8_pix_avg_loop | ||
| 1482 | |||
| 1483 | pop {r4, r5} | ||
| 1484 | WELS_ASM_FUNC_END | ||
| 1485 | |||
| 1486 | |||
| 1487 | WELS_ASM_FUNC_BEGIN PixelAvgWidthEq4_neon | ||
| 1488 | push {r4-r8} | ||
| 1489 | ldr r4, [sp, #20] | ||
| 1490 | w4_pix_avg_loop: | ||
| 1491 | |||
| 1492 | ldr r5, [r2] | ||
| 1493 | ldr r6, [r2, #16] | ||
| 1494 | ldr r7, [r3] | ||
| 1495 | ldr r8, [r3, #16] | ||
| 1496 | add r2, #32 | ||
| 1497 | add r3, #32 | ||
| 1498 | |||
| 1499 | vmov d0, r5, r6 | ||
| 1500 | vmov d1, r7, r8 | ||
| 1501 | AVERAGE_TWO_8BITS d0, d0, d1 | ||
| 1502 | vmov r5, r6, d0 | ||
| 1503 | |||
| 1504 | str r5, [r0], r1 | ||
| 1505 | str r6, [r0], r1 | ||
| 1506 | |||
| 1507 | sub r4, #2 | ||
| 1508 | cmp r4, #0 | ||
| 1509 | bne w4_pix_avg_loop | ||
| 1510 | |||
| 1511 | pop {r4-r8} | ||
| 1512 | WELS_ASM_FUNC_END | ||
| 1513 | |||
| 1514 | WELS_ASM_FUNC_BEGIN McChromaWidthEq8_neon | ||
| 1515 | push {r4, r5} | ||
| 1516 | ldr r4, [sp, #8] | ||
| 1517 | ldr r5, [sp, #12] | ||
| 1518 | // normal case: {cA*src[x] + cB*src[x+1]} + {cC*src[x+stride] + cD*srcp[x+stride+1]} | ||
| 1519 | // we can opti it by adding vert only/ hori only cases, to be continue | ||
| 1520 | vld1.u8 {d31}, [r4] //load A/B/C/D | ||
| 1521 | vld1.u8 {q0}, [r0], r1 //src[x] | ||
| 1522 | |||
| 1523 | vdup.u8 d28, d31[0] //A | ||
| 1524 | vdup.u8 d29, d31[1] //B | ||
| 1525 | vdup.u8 d30, d31[2] //C | ||
| 1526 | vdup.u8 d31, d31[3] //D | ||
| 1527 | |||
| 1528 | vext.u8 d1, d0, d1, #1 //src[x+1] | ||
| 1529 | |||
| 1530 | w8_mc_chroma_loop: // each two pxl row | ||
| 1531 | vld1.u8 {q1}, [r0], r1 //src[x+stride] | ||
| 1532 | vld1.u8 {q2}, [r0], r1 //src[x+2*stride] | ||
| 1533 | vext.u8 d3, d2, d3, #1 //src[x+stride+1] | ||
| 1534 | vext.u8 d5, d4, d5, #1 //src[x+2*stride+1] | ||
| 1535 | |||
| 1536 | vmull.u8 q3, d0, d28 //(src[x] * A) | ||
| 1537 | vmlal.u8 q3, d1, d29 //+=(src[x+1] * B) | ||
| 1538 | vmlal.u8 q3, d2, d30 //+=(src[x+stride] * C) | ||
| 1539 | vmlal.u8 q3, d3, d31 //+=(src[x+stride+1] * D) | ||
| 1540 | vrshrn.u16 d6, q3, #6 | ||
| 1541 | vst1.u8 d6, [r2], r3 | ||
| 1542 | |||
| 1543 | vmull.u8 q3, d2, d28 //(src[x] * A) | ||
| 1544 | vmlal.u8 q3, d3, d29 //+=(src[x+1] * B) | ||
| 1545 | vmlal.u8 q3, d4, d30 //+=(src[x+stride] * C) | ||
| 1546 | vmlal.u8 q3, d5, d31 //+=(src[x+stride+1] * D) | ||
| 1547 | vrshrn.u16 d6, q3, #6 | ||
| 1548 | vst1.u8 d6, [r2], r3 | ||
| 1549 | |||
| 1550 | vmov q0, q2 | ||
| 1551 | sub r5, #2 | ||
| 1552 | cmp r5, #0 | ||
| 1553 | bne w8_mc_chroma_loop | ||
| 1554 | |||
| 1555 | pop {r4, r5} | ||
| 1556 | WELS_ASM_FUNC_END | ||
| 1557 | |||
| 1558 | |||
| 1559 | WELS_ASM_FUNC_BEGIN McChromaWidthEq4_neon | ||
| 1560 | |||
| 1561 | push {r4, r5, r6} | ||
| 1562 | ldr r4, [sp, #12] | ||
| 1563 | ldr r6, [sp, #16] | ||
| 1564 | // normal case: {cA*src[x] + cB*src[x+1]} + {cC*src[x+stride] + cD*srcp[x+stride+1]} | ||
| 1565 | // we can opti it by adding vert only/ hori only cases, to be continue | ||
| 1566 | vld1.u8 {d31}, [r4] //load A/B/C/D | ||
| 1567 | |||
| 1568 | vdup.u8 d28, d31[0] //A | ||
| 1569 | vdup.u8 d29, d31[1] //B | ||
| 1570 | vdup.u8 d30, d31[2] //C | ||
| 1571 | vdup.u8 d31, d31[3] //D | ||
| 1572 | |||
| 1573 | w4_mc_chroma_loop: // each two pxl row | ||
| 1574 | vld1.u8 {d0}, [r0], r1 //a::src[x] | ||
| 1575 | vld1.u8 {d2}, [r0], r1 //b::src[x+stride] | ||
| 1576 | vld1.u8 {d4}, [r0] //c::src[x+2*stride] | ||
| 1577 | |||
| 1578 | vshr.u64 d1, d0, #8 | ||
| 1579 | vshr.u64 d3, d2, #8 | ||
| 1580 | vshr.u64 d5, d4, #8 | ||
| 1581 | |||
| 1582 | vmov q3, q1 //b::[0:7]+b::[1~8] | ||
| 1583 | vtrn.32 q0, q1 //d0{a::[0:3]+b::[0:3]}; d1{a::[1:4]+b::[1:4]} | ||
| 1584 | vtrn.32 q3, q2 //d6{b::[0:3]+c::[0:3]}; d7{b::[1:4]+c::[1:4]} | ||
| 1585 | |||
| 1586 | vmull.u8 q1, d0, d28 //(src[x] * A) | ||
| 1587 | vmlal.u8 q1, d1, d29 //+=(src[x+1] * B) | ||
| 1588 | vmlal.u8 q1, d6, d30 //+=(src[x+stride] * C) | ||
| 1589 | vmlal.u8 q1, d7, d31 //+=(src[x+stride+1] * D) | ||
| 1590 | |||
| 1591 | vrshrn.u16 d2, q1, #6 | ||
| 1592 | vmov r4, r5, d2 | ||
| 1593 | str r4, [r2], r3 | ||
| 1594 | str r5, [r2], r3 | ||
| 1595 | |||
| 1596 | sub r6, #2 | ||
| 1597 | cmp r6, #0 | ||
| 1598 | bne w4_mc_chroma_loop | ||
| 1599 | |||
| 1600 | pop {r4, r5, r6} | ||
| 1601 | WELS_ASM_FUNC_END | ||
| 1602 | #endif | ||