1 @/***************************************************************************** 2 @* 3 @* Copyright (C) 2012 Ittiam Systems Pvt Ltd, Bangalore 4 @* 5 @* Licensed under the Apache License, Version 2.0 (the "License"); 6 @* you may not use this file except in compliance with the License. 7 @* You may obtain a copy of the License at: 8 @* 9 @* http://www.apache.org/licenses/LICENSE-2.0 10 @* 11 @* Unless required by applicable law or agreed to in writing, software 12 @* distributed under the License is distributed on an "AS IS" BASIS, 13 @* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 @* See the License for the specific language governing permissions and 15 @* limitations under the License. 16 @* 17 @*****************************************************************************/ 18 @/** 19 @******************************************************************************* 20 @* @file 21 @* ihevc_intra_pred_chroma_mode_27_to_33.s 22 @* 23 @* @brief 24 @* contains function definition for intra prediction interpolation filters 25 @* 26 @* 27 @* @author 28 @* parthiban v 29 @* 30 @* @par list of functions: 31 @* - ihevc_intra_pred_chroma_mode_27_to_33() 32 @* 33 @* @remarksll 34 @* none 35 @* 36 @******************************************************************************* 37 @*/ 38 39 @/** 40 @******************************************************************************* 41 @* 42 @* @brief 43 @* intraprediction for mode 27 to 33 (positive angle, vertical mode ) with 44 @* neighboring samples location pointed by 'pu1_ref' to the tu 45 @* block location pointed by 'pu1_dst' 46 @* 47 @* @par description: 48 @* 49 @* 50 @* @param[in] pu1_src 51 @* uword8 pointer to the source 52 @* 53 @* @param[in] pu1_dst 54 @* uword8 pointer to the destination 55 @* 56 @* @param[in] src_strd 57 @* integer source stride 58 @* 59 @* @param[in] dst_strd 60 @* integer destination stride 61 @* 62 @* @param[in] nt 63 @* integer transform block size 64 @* 65 @* @param[in] mode 66 @* integer intraprediction mode 67 @* 68 @* @returns 69 @* 70 @* @remarks 71 @* none 72 @* 73 @******************************************************************************* 74 @*/ 75 76 @.if intra_pred_chroma_27_t0_33 == c 77 @void ihevc_intra_pred_chroma_mode_27_to_33(uword8 *pu1_ref, 78 @ word32 src_strd, 79 @ uword8 *pu1_dst, 80 @ word32 dst_strd, 81 @ word32 nt, 82 @ word32 mode) 83 84 .equ nt_offset, 104 85 .equ mode_offset, 108 86 87 .text 88 .align 4 89 90 91 92 93 .globl ihevc_intra_pred_chroma_mode_27_to_33_a9q 94 .extern gai4_ihevc_ang_table 95 .extern gau1_ihevc_planar_factor 96 97 gai4_ihevc_ang_table_addr: 98 .long gai4_ihevc_ang_table - ulbl1 - 8 99 100 gau1_ihevc_planar_factor_addr: 101 .long gau1_ihevc_planar_factor - ulbl2 - 8 102 103 104 .type ihevc_intra_pred_chroma_mode_27_to_33_a9q, %function 105 106 ihevc_intra_pred_chroma_mode_27_to_33_a9q: 107 108 stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments 109 vpush {d8 - d15} 110 111 ldr r4,[sp,#nt_offset] @loads nt 112 ldr r5,[sp,#mode_offset] @loads mode 113 ldr r6,gai4_ihevc_ang_table_addr @loads word32 gai4_ihevc_ang_table[35] 114 ulbl1: 115 add r6,r6,pc 116 117 lsl r7,r4,#2 @four_nt 118 119 add r8,r6,r5,lsl #2 @*gai4_ihevc_ang_table[mode] 120 ldr r9,[r8] @intra_pred_ang = gai4_ihevc_ang_table[mode] 121 ldr r1,gau1_ihevc_planar_factor_addr @used for ((row + 1) * intra_pred_ang) row values 122 ulbl2: 123 add r1,r1,pc 124 add r6,r1,#1 125 126 tst r4,#7 127 add r8,r0,r7 @pu1_ref + four_nt 128 mov lr,#0 @row 129 mov r12,r4 130 bne core_loop_4 131 lsl r4,r4,#1 132 b core_loop_8 133 134 core_loop_8: 135 add r8,r8,#2 @pu1_ref_main_idx += (four_nt + 1) 136 vdup.8 d0,r9 @intra_pred_ang 137 mov r12,r4,lsr #4 @divide by 8 138 139 vmov.i8 d1,#32 140 mul r7,r4,r12 141 142 vmov.i16 q3,#31 143 144 mov r1,r8 145 mov r5,r4 146 mov r11,#2 147 148 prologue: 149 vld1.8 {d3},[r6] @loads the row value 150 vmull.u8 q1,d3,d0 @pos = ((row + 1) * intra_pred_ang) 151 vand q2,q1,q3 @dup_const_fract(fract = pos & (31)) 152 vmovn.i16 d4,q2 153 vshrn.u16 d5,q1,#5 @idx = pos >> 5 154 155 vdup.8 d31,d4[0] 156 add r0,r2,r3 157 158 vmov.u32 lr,d5[0] @(i row)extract idx to the r register 159 lsl lr,lr,#1 160 161 vdup.8 d29,d4[1] @(ii) 162 and r9,lr,#0xff @(i row) get the last byte 163 164 add r10,r8,r9 @(i row)*pu1_ref[ref_main_idx] 165 166 asr lr,lr,#8 @(ii)shift by 8 167 vld1.8 {d8},[r10],r11 @(i row)ref_main_idx 168 and r9,lr,#0xff @(ii)get the last byte 169 170 asr lr,lr,#8 @(iii) 171 vld1.8 {d9},[r10] @(i row)ref_main_idx_1 172 add r12,r8,r9 @(ii)*pu1_ref[ref_main_idx] 173 174 and r9,lr,#0xff @(iii) 175 vsub.u8 d30,d1,d31 @32-fract(dup_const_32_fract) 176 add r10,r8,r9 @(iii)*pu1_ref[ref_main_idx] 177 178 vld1.8 {d12},[r12],r11 @(ii)ref_main_idx 179 vmull.u8 q5,d8,d30 @(i row)vmull_u8(ref_main_idx, dup_const_32_fract) 180 181 vld1.8 {d13},[r12] @(ii)ref_main_idx_1 182 vmlal.u8 q5,d9,d31 @(i row)vmull_u8(ref_main_idx_1, dup_const_fract) 183 asr lr,lr,#8 @(iv) 184 185 vdup.8 d27,d4[2] @(iii) 186 vsub.u8 d28,d1,d29 @(ii)32-fract(dup_const_32_fract) 187 and r9,lr,#0xff @(iv) 188 189 vdup.8 d25,d4[3] @(iv) 190 vmull.u8 q7,d12,d28 @(ii)vmull_u8(ref_main_idx, dup_const_32_fract) 191 add r12,r8,r9 @(iv)*pu1_ref[ref_main_idx] 192 193 vld1.8 {d16},[r10],r11 @(iii)ref_main_idx 194 vmlal.u8 q7,d13,d29 @(ii)vmull_u8(ref_main_idx_1, dup_const_fract) 195 196 vld1.8 {d17},[r10] @(iii)ref_main_idx_1 197 vrshrn.i16 d10,q5,#5 @(i row)shift_res = vrshrn_n_u16(add_res, 5) 198 199 vld1.8 {d20},[r12],r11 @(iv)ref_main_idx 200 vsub.u8 d26,d1,d27 @(iii)32-fract(dup_const_32_fract) 201 202 vld1.8 {d21},[r12] @(iv)ref_main_idx_1 203 204 vdup.8 d31,d4[4] @(v) 205 vmull.u8 q9,d16,d26 @(iii)vmull_u8(ref_main_idx, dup_const_32_fract) 206 207 vmov.u32 lr,d5[1] @extract idx to the r register 208 vmlal.u8 q9,d17,d27 @(iii)vmull_u8(ref_main_idx_1, dup_const_fract) 209 lsl lr,lr,#1 210 211 vst1.8 {d10},[r2]! @(i row) 212 vrshrn.i16 d14,q7,#5 @(ii)shift_res = vrshrn_n_u16(add_res, 5) 213 214 and r9,lr,#0xff @(v) 215 vdup.8 d29,d4[5] @(vi) 216 add r10,r8,r9 @(v)*pu1_ref[ref_main_idx] 217 218 vld1.8 {d8},[r10],r11 @(v)ref_main_idx 219 vsub.u8 d24,d1,d25 @(iv)32-fract(dup_const_32_fract) 220 221 asr lr,lr,#8 @(vi) 222 vmull.u8 q11,d20,d24 @(iv)vmull_u8(ref_main_idx, dup_const_32_fract) 223 and r9,lr,#0xff @(vi) 224 225 vld1.8 {d9},[r10] @(v)ref_main_idx_1 226 vmlal.u8 q11,d21,d25 @(iv)vmull_u8(ref_main_idx_1, dup_const_fract) 227 228 vst1.8 {d14},[r0],r3 @(ii) 229 vrshrn.i16 d18,q9,#5 @(iii)shift_res = vrshrn_n_u16(add_res, 5) 230 231 add r12,r8,r9 @(vi)*pu1_ref[ref_main_idx] 232 vdup.8 d27,d4[6] @(vii) 233 asr lr,lr,#8 @(vii) 234 235 and r9,lr,#0xff @(vii) 236 vsub.u8 d30,d1,d31 @(v)32-fract(dup_const_32_fract) 237 add r10,r8,r9 @(vii)*pu1_ref[ref_main_idx] 238 239 vld1.8 {d12},[r12],r11 @(vi)ref_main_idx 240 vmull.u8 q5,d8,d30 @(v)vmull_u8(ref_main_idx, dup_const_32_fract) 241 242 vld1.8 {d13},[r12] @(vi)ref_main_idx_1 243 vmlal.u8 q5,d9,d31 @(v)vmull_u8(ref_main_idx_1, dup_const_fract) 244 245 vst1.8 {d18},[r0],r3 @(iii) 246 vrshrn.i16 d22,q11,#5 @(iv)shift_res = vrshrn_n_u16(add_res, 5) 247 248 asr lr,lr,#8 @(viii) 249 vdup.8 d25,d4[7] @(viii) 250 and r9,lr,#0xff @(viii) 251 252 vld1.8 {d16},[r10],r11 @(vii)ref_main_idx 253 vsub.u8 d28,d1,d29 @(vi)32-fract(dup_const_32_fract) 254 255 vld1.8 {d17},[r10] @(vii)ref_main_idx_1 256 vmull.u8 q7,d12,d28 @(vi)vmull_u8(ref_main_idx, dup_const_32_fract) 257 258 add r12,r8,r9 @(viii)*pu1_ref[ref_main_idx] 259 vmlal.u8 q7,d13,d29 @(vi)vmull_u8(ref_main_idx_1, dup_const_fract) 260 subs r7,r7,#8 261 262 vst1.8 {d22},[r0],r3 @(iv) 263 vrshrn.i16 d10,q5,#5 @(v)shift_res = vrshrn_n_u16(add_res, 5) 264 265 vld1.8 {d20},[r12],r11 @(viii)ref_main_idx 266 vsub.u8 d26,d1,d27 @(vii)32-fract(dup_const_32_fract) 267 268 vld1.8 {d21},[r12] @(viii)ref_main_idx_1 269 vmull.u8 q9,d16,d26 @(vii)vmull_u8(ref_main_idx, dup_const_32_fract) 270 271 addgt r8,r8,#8 272 vmlal.u8 q9,d17,d27 @(vii)vmull_u8(ref_main_idx_1, dup_const_fract) 273 subgt r4,r4,#8 274 275 vst1.8 {d10},[r0],r3 @(v) 276 vrshrn.i16 d14,q7,#5 @(vi)shift_res = vrshrn_n_u16(add_res, 5) 277 278 beq epilogue 279 280 vld1.8 {d5},[r6] @loads the row value 281 vmull.u8 q1,d5,d0 @pos = ((row + 1) * intra_pred_ang) 282 vand q2,q1,q3 @dup_const_fract(fract = pos & (31)) 283 vmovn.i16 d4,q2 284 vshrn.u16 d3,q1,#5 @idx = pos >> 5 285 vmov.u32 lr,d3[0] @(i)extract idx to the r register 286 lsl lr,lr,#1 287 and r9,lr,#0xff @(i) 288 add r10,r8,r9 @(i)*pu1_ref[ref_main_idx] 289 290 kernel_8_rows: 291 asr lr,lr,#8 @(ii) 292 vdup.8 d31,d4[0] 293 subs r4,r4,#8 294 295 vld1.8 {d8},[r10],r11 @(i)ref_main_idx 296 vsub.u8 d24,d1,d25 @(viii)32-fract(dup_const_32_fract) 297 and r9,lr,#0xff @(ii) 298 addle r6,r6,#8 @increment the row value 299 300 vld1.8 {d9},[r10] @(i)ref_main_idx_1 301 vmull.u8 q11,d20,d24 @(viii)vmull_u8(ref_main_idx, dup_const_32_fract) 302 add r12,r8,r9 @(ii)*pu1_ref[ref_main_idx] 303 304 vld1.8 {d5},[r6] @loads the row value 305 vmlal.u8 q11,d21,d25 @(viii)vmull_u8(ref_main_idx_1, dup_const_fract) 306 asr lr,lr,#8 @(iii) 307 308 vdup.8 d29,d4[1] @(ii) 309 vrshrn.i16 d18,q9,#5 @(vii)shift_res = vrshrn_n_u16(add_res, 5) 310 and r9,lr,#0xff @(iii) 311 312 vst1.8 {d14},[r0],r3 @(vi) 313 vsub.u8 d30,d1,d31 @(i)32-fract(dup_const_32_fract) 314 add r10,r8,r9 @(iii)*pu1_ref[ref_main_idx] 315 316 vld1.8 {d12},[r12],r11 @(ii)ref_main_idx 317 vmull.u8 q5,d8,d30 @(i)vmull_u8(ref_main_idx, dup_const_32_fract) 318 asr lr,lr,#8 @(iv) 319 320 vld1.8 {d13},[r12] @(ii)ref_main_idx_1 321 vmlal.u8 q5,d9,d31 @(i)vmull_u8(ref_main_idx_1, dup_const_fract) 322 and r9,lr,#0xff @(iv) 323 324 vmov.u32 lr,d3[1] @extract idx to the r register 325 vrshrn.i16 d22,q11,#5 @(viii)shift_res = vrshrn_n_u16(add_res, 5) 326 327 vdup.8 d27,d4[2] @(iii) 328 vsub.u8 d28,d1,d29 @(ii)32-fract(dup_const_32_fract) 329 movle r4,r5 @reload nt 330 331 vld1.8 {d16},[r10],r11 @(iii)ref_main_idx 332 vmull.u8 q7,d12,d28 @(ii)vmull_u8(ref_main_idx, dup_const_32_fract) 333 add r12,r8,r9 @(iv)*pu1_ref[ref_main_idx] 334 335 vst1.8 {d18},[r0],r3 @(vii) 336 vmlal.u8 q7,d13,d29 @(ii)vmull_u8(ref_main_idx_1, dup_const_fract) 337 338 vld1.8 {d17},[r10] @(iii)ref_main_idx_1 339 vrshrn.i16 d10,q5,#5 @(i)shift_res = vrshrn_n_u16(add_res, 5) 340 341 vdup.8 d25,d4[3] @(iv) 342 vmull.u8 q1,d5,d0 @pos = ((row + 1) * intra_pred_ang) 343 344 vst1.8 {d22},[r0] @(viii) 345 vsub.u8 d26,d1,d27 @(iii)32-fract(dup_const_32_fract) 346 347 vld1.8 {d20},[r12],r11 @(iv)ref_main_idx 348 vmull.u8 q9,d16,d26 @(iii)vmull_u8(ref_main_idx, dup_const_32_fract) 349 lsl lr,lr,#1 350 351 vld1.8 {d21},[r12] @(iv)ref_main_idx_1 352 vmlal.u8 q9,d17,d27 @(iii)vmull_u8(ref_main_idx_1, dup_const_fract) 353 add r0,r2,r3 354 355 vdup.8 d31,d4[4] @(v) 356 vrshrn.i16 d14,q7,#5 @(ii)shift_res = vrshrn_n_u16(add_res, 5) 357 and r9,lr,#0xff @(v) 358 359 vst1.8 {d10},[r2]! @(i) 360 vsub.u8 d24,d1,d25 @(iv)32-fract(dup_const_32_fract) 361 add r10,r8,r9 @(v)*pu1_ref[ref_main_idx] 362 363 vdup.8 d29,d4[5] @(vi) 364 vmull.u8 q11,d20,d24 @(iv)vmull_u8(ref_main_idx, dup_const_32_fract) 365 asr lr,lr,#8 @(vi) 366 367 vdup.8 d27,d4[6] @(vii) 368 vmlal.u8 q11,d21,d25 @(iv)vmull_u8(ref_main_idx_1, dup_const_fract) 369 and r9,lr,#0xff @(vi) 370 371 vdup.8 d25,d4[7] @(viii) 372 vrshrn.i16 d18,q9,#5 @(iii)shift_res = vrshrn_n_u16(add_res, 5) 373 add r12,r8,r9 @(vi)*pu1_ref[ref_main_idx] 374 375 vld1.8 {d8},[r10],r11 @(v)ref_main_idx 376 vand q2,q1,q3 @dup_const_fract(fract = pos & (31)) 377 asr lr,lr,#8 @(vii) 378 379 vld1.8 {d9},[r10] @(v)ref_main_idx_1 380 vshrn.u16 d3,q1,#5 @idx = pos >> 5 381 and r9,lr,#0xff @(vii) 382 383 vst1.8 {d14},[r0],r3 @(ii) 384 vrshrn.i16 d22,q11,#5 @(iv)shift_res = vrshrn_n_u16(add_res, 5) 385 asr lr,lr,#8 @(viii) 386 387 vld1.8 {d12},[r12],r11 @(vi)ref_main_idx 388 vsub.u8 d30,d1,d31 @(v)32-fract(dup_const_32_fract) 389 add r10,r8,r9 @(vii)*pu1_ref[ref_main_idx] 390 391 vld1.8 {d13},[r12] @(vi)ref_main_idx_1 392 vmull.u8 q5,d8,d30 @(v)vmull_u8(ref_main_idx, dup_const_32_fract) 393 and r9,lr,#0xff @(viii) 394 395 vmov.u32 lr,d3[0] @(i)extract idx to the r register 396 vmlal.u8 q5,d9,d31 @(v)vmull_u8(ref_main_idx_1, dup_const_fract) 397 add r12,r8,r9 @(viii)*pu1_ref[ref_main_idx] 398 399 vld1.8 {d16},[r10],r11 @(vii)ref_main_idx 400 vsub.u8 d28,d1,d29 @(vi)32-fract(dup_const_32_fract) 401 402 vst1.8 {d18},[r0],r3 @(iii) 403 vmull.u8 q7,d12,d28 @(vi)vmull_u8(ref_main_idx, dup_const_32_fract) 404 movle r8,r1 @reload the source to pu1_src+2nt 405 406 vld1.8 {d17},[r10] @(vii)ref_main_idx_1 407 vmlal.u8 q7,d13,d29 @(vi)vmull_u8(ref_main_idx_1, dup_const_fract) 408 addgt r8,r8,#8 @increment the source next set 8 columns in same row 409 410 vld1.8 {d20},[r12],r11 @(viii)ref_main_idx 411 vrshrn.i16 d10,q5,#5 @(v)shift_res = vrshrn_n_u16(add_res, 5) 412 413 vld1.8 {d21},[r12] @(viii)ref_main_idx_1 414 vsub.u8 d26,d1,d27 @(vii)32-fract(dup_const_32_fract) 415 lslle r12,r3,#3 416 417 vst1.8 {d22},[r0],r3 @(iv) 418 vmull.u8 q9,d16,d26 @(vii)vmull_u8(ref_main_idx, dup_const_32_fract) 419 suble r12,r12,r5 420 421 vst1.8 {d10},[r0],r3 @(v) 422 vmlal.u8 q9,d17,d27 @(vii)vmull_u8(ref_main_idx_1, dup_const_fract) 423 addle r2,r2,r12 @increment the dst pointer to 8*dst_strd - nt 424 425 vmovn.i16 d4,q2 426 vrshrn.i16 d14,q7,#5 @(vi)shift_res = vrshrn_n_u16(add_res, 5) 427 lsl lr,lr,#1 428 429 and r9,lr,#0xff @(i) 430 subs r7,r7,#8 431 add r10,r8,r9 @(i)*pu1_ref[ref_main_idx] 432 433 bne kernel_8_rows 434 435 epilogue: 436 vst1.8 {d14},[r0],r3 @(vi) 437 vrshrn.i16 d18,q9,#5 @(vii)shift_res = vrshrn_n_u16(add_res, 5) 438 439 vsub.u8 d24,d1,d25 @(viii)32-fract(dup_const_32_fract) 440 vmull.u8 q11,d20,d24 @(viii)vmull_u8(ref_main_idx, dup_const_32_fract) 441 vmlal.u8 q11,d21,d25 @(viii)vmull_u8(ref_main_idx_1, dup_const_fract) 442 443 vst1.8 {d18},[r0],r3 @(vii) 444 vrshrn.i16 d22,q11,#5 @(viii)shift_res = vrshrn_n_u16(add_res, 5) 445 446 vst1.8 {d22},[r0],r3 @(viii) 447 b end_loops 448 449 core_loop_4: 450 add r10,r8,#2 @pu1_ref_main_idx += (four_nt + 1) 451 add r11,r8,#4 @pu1_ref_main_idx_1 += (four_nt + 2) 452 mov r8,#0 453 454 add r5,r8,#1 @row + 1 455 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 456 and r5,r5,#31 @fract = pos & (31) 457 cmp lr,r5 @if(fract_prev > fract) 458 addgt r10,r10,#2 @pu1_ref_main_idx += 2 459 add r11,r10,#2 @pu1_ref_main_idx_1 += 2 460 vdup.8 d0,r5 @dup_const_fract 461 rsb r4,r5,#32 462 vdup.8 d1,r4 @dup_const_32_fract 463 464 @inner_loop_4 465 vld1.8 {d2},[r10] @ref_main_idx 466 add r8,r8,#1 467 mov lr,r5 @fract_prev = fract 468 469 vld1.8 {d3},[r11] @ref_main_idx_1 470 add r5,r8,#1 @row + 1 471 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 472 and r5,r5,#31 @fract = pos & (31) 473 cmp lr,r5 @if(fract_prev > fract) 474 addgt r10,r10,#2 @pu1_ref_main_idx += 1 475 add r11,r10,#2 @pu1_ref_main_idx_1 += 1 476 477 vdup.8 d6,r5 @dup_const_fract 478 vmull.u8 q2,d2,d1 @vmull_u8(ref_main_idx, dup_const_32_fract) 479 480 rsb r4,r5,#32 481 vdup.8 d7,r4 @dup_const_32_fract 482 vmlal.u8 q2,d3,d0 @vmull_u8(ref_main_idx_1, dup_const_fract) 483 484 vld1.8 {d8},[r10] @ref_main_idx 485 add r8,r8,#1 486 487 vld1.8 {d9},[r11] @ref_main_idx_1 488 vrshrn.i16 d4,q2,#5 @shift_res = vrshrn_n_u16(add_res, 5) 489 490 mov lr,r5 @fract_prev = fract 491 add r5,r8,#1 @row + 1 492 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 493 and r5,r5,#31 @fract = pos & (31) 494 cmp lr,r5 @if(fract_prev > fract) 495 addgt r10,r10,#2 @pu1_ref_main_idx += 1 496 add r11,r10,#2 @pu1_ref_main_idx_1 += 1 497 498 vdup.8 d12,r5 @dup_const_fract 499 vmull.u8 q5,d8,d7 @vmull_u8(ref_main_idx, dup_const_32_fract) 500 501 rsb r4,r5,#32 502 vdup.8 d13,r4 @dup_const_32_fract 503 vmlal.u8 q5,d9,d6 @vmull_u8(ref_main_idx_1, dup_const_fract) 504 505 vld1.8 {d14},[r10] @ref_main_idx 506 add r8,r8,#1 507 508 vst1.8 {d4},[r2],r3 509 vrshrn.i16 d10,q5,#5 @shift_res = vrshrn_n_u16(add_res, 5) 510 511 vld1.8 {d15},[r11] @ref_main_idx_1 512 mov lr,r5 @fract_prev = fract 513 add r5,r8,#1 @row + 1 514 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 515 and r5,r5,#31 @fract = pos & (31) 516 cmp lr,r5 @if(fract_prev > fract) 517 addgt r10,r10,#2 @pu1_ref_main_idx += 1 518 add r11,r10,#2 @pu1_ref_main_idx_1 += 1 519 520 vdup.8 d18,r5 @dup_const_fract 521 vmull.u8 q8,d14,d13 @vmull_u8(ref_main_idx, dup_const_32_fract) 522 523 rsb r4,r5,#32 524 vdup.8 d19,r4 @dup_const_32_fract 525 vmlal.u8 q8,d15,d12 @vmull_u8(ref_main_idx_1, dup_const_fract) 526 527 vld1.8 {d20},[r10] @ref_main_idx 528 529 vst1.8 {d10},[r2],r3 530 vrshrn.i16 d16,q8,#5 @shift_res = vrshrn_n_u16(add_res, 5) 531 vld1.8 {d21},[r11] @ref_main_idx_1 532 533 vmull.u8 q11,d20,d19 @vmull_u8(ref_main_idx, dup_const_32_fract) 534 vmlal.u8 q11,d21,d18 @vmull_u8(ref_main_idx_1, dup_const_fract) 535 536 vst1.8 {d16},[r2],r3 537 vrshrn.i16 d22,q11,#5 @shift_res = vrshrn_n_u16(add_res, 5) 538 539 vst1.8 {d22},[r2],r3 540 541 end_loops: 542 vpop {d8 - d15} 543 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 544 545 546 547 548
