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_luma_mode_19_to_25.s 22 @* 23 @* @brief 24 @* contains function definitions for intra prediction dc filtering. 25 @* functions are coded using neon intrinsics and can be compiled using 26 27 @* rvct 28 @* 29 @* @author 30 @* naveen sr 31 @* 32 @* @par list of functions: 33 @* 34 @* 35 @* @remarks 36 @* none 37 @* 38 @******************************************************************************* 39 @*/ 40 @/** 41 @******************************************************************************* 42 @* 43 @* @brief 44 @* luma intraprediction filter for dc input 45 @* 46 @* @par description: 47 @* 48 @* @param[in] pu1_ref 49 @* uword8 pointer to the source 50 @* 51 @* @param[out] pu1_dst 52 @* uword8 pointer to the destination 53 @* 54 @* @param[in] src_strd 55 @* integer source stride 56 @* 57 @* @param[in] dst_strd 58 @* integer destination stride 59 @* 60 @* @param[in] nt 61 @* size of tranform block 62 @* 63 @* @param[in] mode 64 @* type of filtering 65 @* 66 @* @returns 67 @* 68 @* @remarks 69 @* none 70 @* 71 @******************************************************************************* 72 @*/ 73 74 @void ihevc_intra_pred_luma_mode_19_to_25(uword8* pu1_ref, 75 @ word32 src_strd, 76 @ uword8* pu1_dst, 77 @ word32 dst_strd, 78 @ word32 nt, 79 @ word32 mode) 80 @ 81 @**************variables vs registers***************************************** 82 @r0 => *pu1_ref 83 @r1 => src_strd 84 @r2 => *pu1_dst 85 @r3 => dst_strd 86 87 @stack contents from #40 88 @ nt 89 @ mode 90 91 .text 92 .align 4 93 94 95 96 97 .globl ihevc_intra_pred_luma_mode_19_to_25_a9q 98 .extern gai4_ihevc_ang_table 99 .extern gai4_ihevc_inv_ang_table 100 .extern gau1_ihevc_planar_factor 101 102 gai4_ihevc_inv_ang_table_addr: 103 .long gai4_ihevc_inv_ang_table - ulbl1 - 8 104 105 gau1_ihevc_planar_factor_addr: 106 .long gau1_ihevc_planar_factor - ulbl2 - 8 107 108 gai4_ihevc_ang_table_addr_1: 109 .long gai4_ihevc_ang_table - ulbl_1 - 8 110 111 gai4_ihevc_ang_table_addr_2: 112 .long gai4_ihevc_ang_table - ulbl_2 - 8 113 114 .type ihevc_intra_pred_luma_mode_19_to_25_a9q, %function 115 116 ihevc_intra_pred_luma_mode_19_to_25_a9q: 117 118 stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments 119 120 ldr r4,[sp,#40] @loads nt 121 ldr r7, gai4_ihevc_ang_table_addr_1 122 ulbl_1: 123 add r7,r7,pc 124 125 ldr r5,[sp,#44] @mode (19 to 25) 126 ldr r8, gai4_ihevc_inv_ang_table_addr 127 ulbl1: 128 add r8,r8,pc 129 130 add r7, r7, r5, lsl #2 @gai4_ihevc_ang_table[mode] 131 add r8, r8, r5, lsl #2 @gai4_ihevc_inv_ang_table 132 sub r8, r8, #48 @gai4_ihevc_inv_ang_table[mode - 12] 133 134 ldr r7, [r7] @intra_pred_ang 135 sub sp, sp, #132 @ref_temp[2 * max_cu_size + 1] 136 137 ldr r8, [r8] @inv_ang 138 add r6, sp, r4 @ref_temp + nt 139 140 mul r9, r4, r7 @nt*intra_pred_ang 141 142 sub r6, r6, #1 @ref_temp + nt - 1 143 144 add r1, r0, r4, lsl #1 @r1 = &src[2nt] 145 vdup.8 d30, r7 @intra_pred_ang 146 147 mov r7, r4 148 149 asr r9, r9, #5 150 151 vld1.32 d0[0],[r1]! @ pu1_ref[two_nt + k] 152 153 vst1.32 d0[0],[r6]! @ref_temp[k + nt - 1] = pu1_ref[two_nt + k]@ 154 155 subs r7, r7, #4 156 beq end_loop_copy 157 sub r1,#4 158 sub r6,#4 159 subs r7,r7,#4 160 beq loop_copy_8 161 subs r7,r7,#8 162 beq loop_copy_16 163 164 loop_copy_32: 165 vld1.8 d0,[r1]! 166 vld1.8 d1,[r1]! 167 vld1.8 d2,[r1]! 168 vld1.8 d3,[r1]! 169 170 vst1.8 d0,[r6]! 171 vst1.8 d1,[r6]! 172 vst1.8 d2,[r6]! 173 vst1.8 d3,[r6]! 174 b end_loop_copy 175 176 loop_copy_16: 177 vld1.8 d0,[r1]! 178 vld1.8 d1,[r1]! 179 180 vst1.8 d0,[r6]! 181 vst1.8 d1,[r6]! 182 b end_loop_copy 183 184 loop_copy_8: 185 vld1.8 d0,[r1]! 186 vst1.8 d0,[r6]! 187 188 end_loop_copy: 189 190 ldrb r11, [r1] 191 strb r11, [r6] 192 193 cmp r9, #-1 194 bge linear_filtering 195 196 add r6, sp, r4 @ref_temp + nt 197 sub r6, r6, #2 @ref_temp + nt - 2 198 199 mov r12, #0xffffffff 200 201 rsb r9, r9, r12 @count to take care off ref_idx 202 203 add r1, r0, r4, lsl #1 @r1 = &src[2nt] 204 205 mov r7, #128 @inv_ang_sum 206 207 loop_copy_ref_idx: 208 209 add r7, r7, r8 @inv_ang_sum += inv_ang 210 mov r14,r7,lsr #8 211 ldrb r11, [r1, -r14] 212 @ ldrb r11, [r1, -r7, lsr #8] 213 strb r11, [r6], #-1 214 215 subs r9, r9, #1 216 217 bne loop_copy_ref_idx 218 219 220 linear_filtering: 221 @ after copy 222 @ below code is taken from mode 27 to 33 and modified 223 224 ldr r6,gai4_ihevc_ang_table_addr_2 @loads word32 gai4_ihevc_ang_table[35] 225 ulbl_2: 226 add r6,r6,pc 227 228 add r8,r6,r5,lsl #2 @*gai4_ihevc_ang_table[mode] 229 ldr r9,[r8] @intra_pred_ang = gai4_ihevc_ang_table[mode] 230 ldr r1,gau1_ihevc_planar_factor_addr @used for ((row + 1) * intra_pred_ang) row values 231 ulbl2: 232 add r1,r1,pc 233 add r6,r1,#1 234 235 add r8, sp, r4 @ref_temp + nt 236 sub r8,#1 @ref_temp + nt -1 237 238 tst r4,#7 239 mov lr,#0 @row 240 mov r12,r4 241 bne core_loop_4 242 243 core_loop_8: 244 add r8,r8,#1 @pu1_ref_main_idx += (two_nt + 1) 245 vdup.8 d0,r9 @intra_pred_ang 246 mov r12,r4,lsr #3 @divide by 8 247 248 vmov.i8 d1,#32 249 mul r7,r4,r12 250 251 vmov.i16 q3,#31 252 @lsl r12,r3,#3 253 254 mov r1,r8 255 @sub r12,r12,r4 256 mov r5,r4 257 mov r11,#1 258 259 prologue: 260 vld1.8 {d3},[r6] @loads the row value 261 vmull.s8 q1,d3,d0 @pos = ((row + 1) * intra_pred_ang) 262 vand q2,q1,q3 @dup_const_fract(fract = pos & (31)) 263 vmovn.i16 d4,q2 264 vshrn.s16 d5,q1,#5 @idx = pos >> 5 265 266 vdup.8 d31,d4[0] 267 add r0,r2,r3 268 269 vmov.u32 lr,d5[0] @(i row)extract idx to the r register 270 271 vdup.8 d29,d4[1] @(ii) 272 sbfx r9,lr,#0,#8 273 274 add r10,r8,r9 @(i row)*pu1_ref[ref_main_idx] 275 276 vld1.8 {d8},[r10],r11 @(i row)ref_main_idx 277 sbfx r9,lr,#8,#8 278 279 vld1.8 {d9},[r10] @(i row)ref_main_idx_1 280 add r12,r8,r9 @(ii)*pu1_ref[ref_main_idx] 281 282 sbfx r9,lr,#16,#8 283 vsub.u8 d30,d1,d31 @32-fract(dup_const_32_fract) 284 add r10,r8,r9 @(iii)*pu1_ref[ref_main_idx] 285 286 vld1.8 {d12},[r12],r11 @(ii)ref_main_idx 287 vmull.u8 q5,d8,d30 @(i row)vmull_u8(ref_main_idx, dup_const_32_fract) 288 289 vld1.8 {d13},[r12] @(ii)ref_main_idx_1 290 vmlal.u8 q5,d9,d31 @(i row)vmull_u8(ref_main_idx_1, dup_const_fract) 291 292 vdup.8 d27,d4[2] @(iii) 293 vsub.u8 d28,d1,d29 @(ii)32-fract(dup_const_32_fract) 294 sbfx r9,lr,#24,#8 295 296 vdup.8 d25,d4[3] @(iv) 297 vmull.u8 q7,d12,d28 @(ii)vmull_u8(ref_main_idx, dup_const_32_fract) 298 add r12,r8,r9 @(iv)*pu1_ref[ref_main_idx] 299 300 vld1.8 {d16},[r10],r11 @(iii)ref_main_idx 301 vmlal.u8 q7,d13,d29 @(ii)vmull_u8(ref_main_idx_1, dup_const_fract) 302 303 vld1.8 {d17},[r10] @(iii)ref_main_idx_1 304 vrshrn.i16 d10,q5,#5 @(i row)shift_res = vrshrn_n_u16(add_res, 5) 305 306 vld1.8 {d20},[r12],r11 @(iv)ref_main_idx 307 vsub.u8 d26,d1,d27 @(iii)32-fract(dup_const_32_fract) 308 309 vld1.8 {d21},[r12] @(iv)ref_main_idx_1 310 311 vdup.8 d31,d4[4] @(v) 312 vmull.u8 q9,d16,d26 @(iii)vmull_u8(ref_main_idx, dup_const_32_fract) 313 314 vmov.u32 lr,d5[1] @extract idx to the r register 315 vmlal.u8 q9,d17,d27 @(iii)vmull_u8(ref_main_idx_1, dup_const_fract) 316 317 vst1.8 {d10},[r2]! @(i row) 318 vrshrn.i16 d14,q7,#5 @(ii)shift_res = vrshrn_n_u16(add_res, 5) 319 320 sbfx r9,lr,#0,#8 321 vdup.8 d29,d4[5] @(vi) 322 add r10,r8,r9 @(v)*pu1_ref[ref_main_idx] 323 324 vld1.8 {d8},[r10],r11 @(v)ref_main_idx 325 vsub.u8 d24,d1,d25 @(iv)32-fract(dup_const_32_fract) 326 327 vmull.u8 q11,d20,d24 @(iv)vmull_u8(ref_main_idx, dup_const_32_fract) 328 sbfx r9,lr,#8,#8 329 330 vld1.8 {d9},[r10] @(v)ref_main_idx_1 331 vmlal.u8 q11,d21,d25 @(iv)vmull_u8(ref_main_idx_1, dup_const_fract) 332 333 vst1.8 {d14},[r0],r3 @(ii) 334 vrshrn.i16 d18,q9,#5 @(iii)shift_res = vrshrn_n_u16(add_res, 5) 335 336 add r12,r8,r9 @(vi)*pu1_ref[ref_main_idx] 337 vdup.8 d27,d4[6] @(vii) 338 339 sbfx r9,lr,#16,#8 340 vsub.u8 d30,d1,d31 @(v)32-fract(dup_const_32_fract) 341 add r10,r8,r9 @(vii)*pu1_ref[ref_main_idx] 342 343 vld1.8 {d12},[r12],r11 @(vi)ref_main_idx 344 vmull.u8 q5,d8,d30 @(v)vmull_u8(ref_main_idx, dup_const_32_fract) 345 346 vld1.8 {d13},[r12] @(vi)ref_main_idx_1 347 vmlal.u8 q5,d9,d31 @(v)vmull_u8(ref_main_idx_1, dup_const_fract) 348 349 vst1.8 {d18},[r0],r3 @(iii) 350 vrshrn.i16 d22,q11,#5 @(iv)shift_res = vrshrn_n_u16(add_res, 5) 351 352 vdup.8 d25,d4[7] @(viii) 353 sbfx r9,lr,#24,#8 354 355 vld1.8 {d16},[r10],r11 @(vii)ref_main_idx 356 vsub.u8 d28,d1,d29 @(vi)32-fract(dup_const_32_fract) 357 358 vld1.8 {d17},[r10] @(vii)ref_main_idx_1 359 vmull.u8 q7,d12,d28 @(vi)vmull_u8(ref_main_idx, dup_const_32_fract) 360 361 add r12,r8,r9 @(viii)*pu1_ref[ref_main_idx] 362 vmlal.u8 q7,d13,d29 @(vi)vmull_u8(ref_main_idx_1, dup_const_fract) 363 subs r4,r4,#8 364 365 vst1.8 {d22},[r0],r3 @(iv) 366 vrshrn.i16 d10,q5,#5 @(v)shift_res = vrshrn_n_u16(add_res, 5) 367 368 vld1.8 {d20},[r12],r11 @(viii)ref_main_idx 369 vsub.u8 d26,d1,d27 @(vii)32-fract(dup_const_32_fract) 370 371 vld1.8 {d21},[r12] @(viii)ref_main_idx_1 372 vmull.u8 q9,d16,d26 @(vii)vmull_u8(ref_main_idx, dup_const_32_fract) 373 374 addgt r8,r8,#8 375 vmlal.u8 q9,d17,d27 @(vii)vmull_u8(ref_main_idx_1, dup_const_fract) 376 subgt r7,r7,#8 377 378 vst1.8 {d10},[r0],r3 @(v) 379 vrshrn.i16 d14,q7,#5 @(vi)shift_res = vrshrn_n_u16(add_res, 5) 380 381 beq epilogue 382 383 vld1.8 {d5},[r6] @loads the row value 384 vmull.s8 q1,d5,d0 @pos = ((row + 1) * intra_pred_ang) 385 vand q2,q1,q3 @dup_const_fract(fract = pos & (31)) 386 vmovn.i16 d4,q2 387 vshrn.s16 d3,q1,#5 @idx = pos >> 5 388 vmov.u32 lr,d3[0] @(i)extract idx to the r register 389 sbfx r9,lr,#0,#8 390 add r10,r8,r9 @(i)*pu1_ref[ref_main_idx] 391 392 kernel_8_rows: 393 vdup.8 d31,d4[0] 394 subs r4,r4,#8 395 sbfx r9,lr,#8,#8 396 397 vld1.8 {d8},[r10],r11 @(i)ref_main_idx 398 vsub.u8 d24,d1,d25 @(viii)32-fract(dup_const_32_fract) 399 400 addle r6,r6,#8 @increment the row value 401 add r12,r8,r9 @(ii)*pu1_ref[ref_main_idx] 402 403 vld1.8 {d9},[r10] @(i)ref_main_idx_1 404 vmull.u8 q11,d20,d24 @(viii)vmull_u8(ref_main_idx, dup_const_32_fract) 405 406 vld1.8 {d5},[r6] @loads the row value 407 vmlal.u8 q11,d21,d25 @(viii)vmull_u8(ref_main_idx_1, dup_const_fract) 408 409 vdup.8 d29,d4[1] @(ii) 410 vrshrn.i16 d18,q9,#5 @(vii)shift_res = vrshrn_n_u16(add_res, 5) 411 412 sbfx r9,lr,#16,#8 413 414 vst1.8 {d14},[r0],r3 @(vi) 415 vsub.u8 d30,d1,d31 @(i)32-fract(dup_const_32_fract) 416 417 add r10,r8,r9 @(iii)*pu1_ref[ref_main_idx] 418 419 vld1.8 {d12},[r12],r11 @(ii)ref_main_idx 420 vmull.u8 q5,d8,d30 @(i)vmull_u8(ref_main_idx, dup_const_32_fract) 421 422 vld1.8 {d13},[r12] @(ii)ref_main_idx_1 423 vmlal.u8 q5,d9,d31 @(i)vmull_u8(ref_main_idx_1, dup_const_fract) 424 425 sbfx r9,lr,#24,#8 426 movle r4,r5 @reload nt 427 428 vmov.u32 lr,d3[1] @extract idx to the r register 429 vrshrn.i16 d22,q11,#5 @(viii)shift_res = vrshrn_n_u16(add_res, 5) 430 431 vdup.8 d27,d4[2] @(iii) 432 vsub.u8 d28,d1,d29 @(ii)32-fract(dup_const_32_fract) 433 add r12,r8,r9 @(iv)*pu1_ref[ref_main_idx] 434 435 vld1.8 {d16},[r10],r11 @(iii)ref_main_idx 436 vmull.u8 q7,d12,d28 @(ii)vmull_u8(ref_main_idx, dup_const_32_fract) 437 438 vst1.8 {d18},[r0],r3 @(vii) 439 vmlal.u8 q7,d13,d29 @(ii)vmull_u8(ref_main_idx_1, dup_const_fract) 440 441 vld1.8 {d17},[r10] @(iii)ref_main_idx_1 442 vrshrn.i16 d10,q5,#5 @(i)shift_res = vrshrn_n_u16(add_res, 5) 443 444 vdup.8 d25,d4[3] @(iv) 445 vmull.s8 q1,d5,d0 @pos = ((row + 1) * intra_pred_ang) 446 447 vst1.8 {d22},[r0] @(viii) 448 vsub.u8 d26,d1,d27 @(iii)32-fract(dup_const_32_fract) 449 450 vld1.8 {d20},[r12],r11 @(iv)ref_main_idx 451 vmull.u8 q9,d16,d26 @(iii)vmull_u8(ref_main_idx, dup_const_32_fract) 452 453 vld1.8 {d21},[r12] @(iv)ref_main_idx_1 454 vmlal.u8 q9,d17,d27 @(iii)vmull_u8(ref_main_idx_1, dup_const_fract) 455 456 sbfx r9,lr,#0,#8 457 add r0,r2,r3 458 459 vdup.8 d31,d4[4] @(v) 460 vrshrn.i16 d14,q7,#5 @(ii)shift_res = vrshrn_n_u16(add_res, 5) 461 462 add r10,r8,r9 @(v)*pu1_ref[ref_main_idx] 463 sbfx r9,lr,#8,#8 464 465 vst1.8 {d10},[r2]! @(i) 466 vsub.u8 d24,d1,d25 @(iv)32-fract(dup_const_32_fract) 467 468 vdup.8 d29,d4[5] @(vi) 469 vmull.u8 q11,d20,d24 @(iv)vmull_u8(ref_main_idx, dup_const_32_fract) 470 471 vdup.8 d27,d4[6] @(vii) 472 vmlal.u8 q11,d21,d25 @(iv)vmull_u8(ref_main_idx_1, dup_const_fract) 473 474 add r12,r8,r9 @(vi)*pu1_ref[ref_main_idx] 475 sbfx r9,lr,#16,#8 476 477 vdup.8 d25,d4[7] @(viii) 478 vrshrn.i16 d18,q9,#5 @(iii)shift_res = vrshrn_n_u16(add_res, 5) 479 480 vld1.8 {d8},[r10],r11 @(v)ref_main_idx 481 vand q2,q1,q3 @dup_const_fract(fract = pos & (31)) 482 483 vld1.8 {d9},[r10] @(v)ref_main_idx_1 484 vshrn.s16 d3,q1,#5 @idx = pos >> 5 485 486 vst1.8 {d14},[r0],r3 @(ii) 487 vrshrn.i16 d22,q11,#5 @(iv)shift_res = vrshrn_n_u16(add_res, 5) 488 489 add r10,r8,r9 @(vii)*pu1_ref[ref_main_idx] 490 sbfx r9,lr,#24,#8 491 492 vld1.8 {d12},[r12],r11 @(vi)ref_main_idx 493 vsub.u8 d30,d1,d31 @(v)32-fract(dup_const_32_fract) 494 495 vld1.8 {d13},[r12] @(vi)ref_main_idx_1 496 vmull.u8 q5,d8,d30 @(v)vmull_u8(ref_main_idx, dup_const_32_fract) 497 498 vmov.u32 lr,d3[0] @(i)extract idx to the r register 499 vmlal.u8 q5,d9,d31 @(v)vmull_u8(ref_main_idx_1, dup_const_fract) 500 501 add r12,r8,r9 @(viii)*pu1_ref[ref_main_idx] 502 movle r8,r1 @reload the source to pu1_src+2nt 503 504 vld1.8 {d16},[r10],r11 @(vii)ref_main_idx 505 vsub.u8 d28,d1,d29 @(vi)32-fract(dup_const_32_fract) 506 507 vst1.8 {d18},[r0],r3 @(iii) 508 vmull.u8 q7,d12,d28 @(vi)vmull_u8(ref_main_idx, dup_const_32_fract) 509 510 vld1.8 {d17},[r10] @(vii)ref_main_idx_1 511 vmlal.u8 q7,d13,d29 @(vi)vmull_u8(ref_main_idx_1, dup_const_fract) 512 513 vld1.8 {d20},[r12],r11 @(viii)ref_main_idx 514 vrshrn.i16 d10,q5,#5 @(v)shift_res = vrshrn_n_u16(add_res, 5) 515 516 vld1.8 {d21},[r12] @(viii)ref_main_idx_1 517 vsub.u8 d26,d1,d27 @(vii)32-fract(dup_const_32_fract) 518 519 addgt r8,r8,#8 @increment the source next set 8 columns in same row 520 lslle r12,r3,#3 521 suble r12,r12,r5 522 523 vst1.8 {d22},[r0],r3 @(iv) 524 vmull.u8 q9,d16,d26 @(vii)vmull_u8(ref_main_idx, dup_const_32_fract) 525 526 vst1.8 {d10},[r0],r3 @(v) 527 vmlal.u8 q9,d17,d27 @(vii)vmull_u8(ref_main_idx_1, dup_const_fract) 528 529 addle r2,r2,r12 @increment the dst pointer to 8*dst_strd - nt 530 sbfx r9,lr,#0,#8 531 532 vmovn.i16 d4,q2 533 vrshrn.i16 d14,q7,#5 @(vi)shift_res = vrshrn_n_u16(add_res, 5) 534 535 subs r7,r7,#8 536 add r10,r8,r9 @(i)*pu1_ref[ref_main_idx] 537 538 bne kernel_8_rows 539 540 epilogue: 541 vst1.8 {d14},[r0],r3 @(vi) 542 vrshrn.i16 d18,q9,#5 @(vii)shift_res = vrshrn_n_u16(add_res, 5) 543 544 vsub.u8 d24,d1,d25 @(viii)32-fract(dup_const_32_fract) 545 vmull.u8 q11,d20,d24 @(viii)vmull_u8(ref_main_idx, dup_const_32_fract) 546 vmlal.u8 q11,d21,d25 @(viii)vmull_u8(ref_main_idx_1, dup_const_fract) 547 548 vst1.8 {d18},[r0],r3 @(vii) 549 vrshrn.i16 d22,q11,#5 @(viii)shift_res = vrshrn_n_u16(add_res, 5) 550 551 vst1.8 {d22},[r0],r3 @(viii) 552 b end_loops 553 554 core_loop_4: 555 add r6,r8,#1 @pu1_ref_main_idx += 1 556 mov r8,#0 557 558 add r5,r8,#1 @row + 1 559 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 560 mov lr,r5,asr #5 @if(fract_prev > fract) 561 and r5,r5,#31 @fract = pos & (31) 562 add r10,r6,lr @pu1_ref_main_idx += 1 563 add r11,r10,#1 @pu1_ref_main_idx_1 += 1 564 vdup.8 d0,r5 @dup_const_fract 565 rsb r4,r5,#32 566 vdup.8 d1,r4 @dup_const_32_fract 567 568 @inner_loop_4 569 vld1.32 {d2[0]},[r10] @ref_main_idx 570 add r8,r8,#1 571 @ mov lr,r5 @fract_prev = fract 572 573 vld1.32 {d3[0]},[r11] @ref_main_idx_1 574 add r5,r8,#1 @row + 1 575 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 576 mov lr,r5,asr #5 @ pos >> 5 577 and r5,r5,#31 @fract = pos & (31) 578 add r10,r6,lr @pu1_ref_main_idx += 1 579 add r11,r10,#1 @pu1_ref_main_idx_1 += 1 580 581 vdup.8 d6,r5 @dup_const_fract 582 vmull.u8 q2,d2,d1 @vmull_u8(ref_main_idx, dup_const_32_fract) 583 584 rsb r4,r5,#32 585 vdup.8 d7,r4 @dup_const_32_fract 586 vmlal.u8 q2,d3,d0 @vmull_u8(ref_main_idx_1, dup_const_fract) 587 588 vld1.32 {d8[0]},[r10] @ref_main_idx 589 add r8,r8,#1 590 591 vld1.32 {d9[0]},[r11] @ref_main_idx_1 592 vrshrn.i16 d4,q2,#5 @shift_res = vrshrn_n_u16(add_res, 5) 593 594 @ mov lr,r5 @fract_prev = fract 595 add r5,r8,#1 @row + 1 596 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 597 mov lr,r5,asr #5 @if(fract_prev > fract) 598 and r5,r5,#31 @fract = pos & (31) 599 add r10,r6,lr @ref_main + idx 600 add r11,r10,#1 @pu1_ref_main_idx_1 += 1 601 602 vdup.8 d12,r5 @dup_const_fract 603 vmull.u8 q5,d8,d7 @vmull_u8(ref_main_idx, dup_const_32_fract) 604 605 rsb r4,r5,#32 606 vdup.8 d13,r4 @dup_const_32_fract 607 vmlal.u8 q5,d9,d6 @vmull_u8(ref_main_idx_1, dup_const_fract) 608 609 vld1.32 {d14[0]},[r10] @ref_main_idx 610 add r8,r8,#1 611 612 vst1.32 {d4[0]},[r2],r3 613 vrshrn.i16 d10,q5,#5 @shift_res = vrshrn_n_u16(add_res, 5) 614 615 vld1.32 {d15[0]},[r11] @ref_main_idx_1 616 @ mov lr,r5 @fract_prev = fract 617 add r5,r8,#1 @row + 1 618 mul r5,r5,r9 @pos = ((row + 1) * intra_pred_ang) 619 mov lr,r5,asr #5 @if(fract_prev > fract) 620 and r5,r5,#31 @fract = pos & (31) 621 add r10,r6,lr @pu1_ref_main_idx += 1 622 add r11,r10,#1 @pu1_ref_main_idx_1 += 1 623 624 vdup.8 d18,r5 @dup_const_fract 625 vmull.u8 q8,d14,d13 @vmull_u8(ref_main_idx, dup_const_32_fract) 626 627 rsb r4,r5,#32 628 vdup.8 d19,r4 @dup_const_32_fract 629 vmlal.u8 q8,d15,d12 @vmull_u8(ref_main_idx_1, dup_const_fract) 630 631 vld1.32 {d20[0]},[r10] @ref_main_idx 632 633 vst1.32 {d10[0]},[r2],r3 634 vrshrn.i16 d16,q8,#5 @shift_res = vrshrn_n_u16(add_res, 5) 635 vld1.32 {d21[0]},[r11] @ref_main_idx_1 636 637 vmull.u8 q11,d20,d19 @vmull_u8(ref_main_idx, dup_const_32_fract) 638 vmlal.u8 q11,d21,d18 @vmull_u8(ref_main_idx_1, dup_const_fract) 639 640 vst1.32 {d16[0]},[r2],r3 641 vrshrn.i16 d22,q11,#5 @shift_res = vrshrn_n_u16(add_res, 5) 642 643 vst1.32 {d22[0]},[r2],r3 644 645 end_loops: 646 add sp, sp, #132 647 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 648 649 650 651 652 653 654
