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_11_to_17.s 22 @* 23 @* @brief 24 @* contains function definitions for intra prediction chroma mode 11 to 17 25 @* functions are coded using neon intrinsics and can be compiled using 26 27 @* rvct 28 @* 29 @* @author 30 @* akshaya mukund 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_chroma_mode_11_to_17(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_chroma_mode_11_to_17_a9q 98 .extern gai4_ihevc_ang_table 99 .extern gai4_ihevc_inv_ang_table 100 .extern col_for_intra_chroma 101 .extern idx_neg_idx_chroma_11_17 102 103 gai4_ihevc_ang_table_addr: 104 .long gai4_ihevc_ang_table - ulbl1 - 8 105 106 gai4_ihevc_inv_ang_table_addr: 107 .long gai4_ihevc_inv_ang_table - ulbl2 - 8 108 109 idx_neg_idx_chroma_11_17_addr: 110 .long idx_neg_idx_chroma_11_17 - ulbl3 - 8 111 112 col_for_intra_chroma_addr_1: 113 .long col_for_intra_chroma - ulbl4 - 8 114 115 col_for_intra_chroma_addr_2: 116 .long col_for_intra_chroma - ulbl5 - 8 117 118 col_for_intra_chroma_addr_3: 119 .long col_for_intra_chroma - ulbl6 - 8 120 121 .type ihevc_intra_pred_chroma_mode_11_to_17_a9q, %function 122 123 ihevc_intra_pred_chroma_mode_11_to_17_a9q: 124 125 stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments 126 127 ldr r4,[sp,#40] @loads nt 128 ldr r7, gai4_ihevc_ang_table_addr 129 ulbl1: 130 add r7,r7,pc 131 132 ldr r5,[sp,#44] @mode (11 to 17) 133 ldr r8, gai4_ihevc_inv_ang_table_addr 134 ulbl2: 135 add r8,r8,pc 136 137 add r7, r7, r5, lsl #2 @gai4_ihevc_ang_table[mode] 138 add r8, r8, r5, lsl #2 @gai4_ihevc_inv_ang_table[mode - 11] 139 sub r8, r8, #44 140 141 ldr r7, [r7] @intra_pred_ang 142 sub sp, sp, #132 @ref_temp[2 * max_cu_size + 2] 143 144 ldr r8, [r8] @inv_ang 145 add r6, sp, r4, lsl #1 @ref_temp + 2 * nt 146 147 mul r9, r4, r7 @nt*intra_pred_ang 148 149 sub r6, r6, #2 @ref_temp + 2*nt - 2 150 151 add r1, r0, r4, lsl #2 @r1 = &src[4nt] 152 vdup.8 d30, r7 @intra_pred_ang 153 154 mov r7, r4 155 156 sub r1,r1,#6 @address calculation for copying 4 halfwords 157 158 asr r9, r9, #5 159 160 vld1.8 d0,[r1] 161 vrev64.16 d0,d0 162 vst1.8 d0,[r6]! 163 164 sub r1,#8 165 166 subs r7, r7, #4 167 addeq r1,#8 168 beq end_loop_copy 169 subs r7,r7,#4 170 beq loop_copy_8 171 subs r7,r7,#8 172 beq loop_copy_16 173 174 loop_copy_32: 175 sub r1,#24 176 vld1.8 {d0,d1,d2,d3},[r1] 177 178 sub r1,#24 179 vld1.8 {d0,d1,d2,d3},[r1]! 180 181 vrev64.16 d6,d6 182 vrev64.16 d5,d5 183 vrev64.16 d4,d4 184 vrev64.16 d3,d3 185 vrev64.16 d2,d2 186 vrev64.16 d1,d1 187 vrev64.16 d0,d0 188 189 vst1.8 d6,[r6]! 190 vst1.8 d5,[r6]! 191 vst1.8 d4,[r6]! 192 vst1.8 d3,[r6]! 193 vst1.8 d2,[r6]! 194 vst1.8 d1,[r6]! 195 vst1.8 d0,[r6]! 196 197 vld1.8 {d4,d5,d6},[r1]! 198 b end_loop_copy 199 200 loop_copy_16: 201 sub r1,#16 202 vld1.8 {d0,d1,d2},[r1] 203 204 vrev64.16 d2,d2 205 vrev64.16 d1,d1 206 vrev64.16 d0,d0 207 208 vst1.8 d2,[r6]! 209 vst1.8 d1,[r6]! 210 vst1.8 d0,[r6]! 211 212 b end_loop_copy 213 loop_copy_8: 214 vld1.8 d0,[r1] 215 vrev64.16 d0,d0 216 vst1.8 d0,[r6]! 217 end_loop_copy: 218 sub r1,#2 219 220 ldrh r11, [r1], #-2 221 strh r11, [r6], #2 222 223 cmp r9, #-1 224 bge prologue_8_16_32 225 226 add r6, sp, r4, lsl #1 @ref_temp + 2 * nt 227 sub r6, r6, #4 @ref_temp + 2 * nt - 2 - 2 228 229 mov r12, #0xffffffff 230 231 rsb r9, r9, r12 @count to take care off ref_idx 232 233 add r1, r0, r4, lsl #2 @r1 = &src[4nt] 234 235 mov r7, #128 @inv_ang_sum 236 237 loop_copy_ref_idx: 238 239 add r7, r7, r8 @inv_ang_sum += inv_ang 240 241 mov r0,r7, lsr #8 242 mov r0,r0, lsl #1 243 244 ldrh r11, [r1, r0] 245 strh r11, [r6], #-2 246 247 subs r9, r9, #1 248 249 bne loop_copy_ref_idx 250 251 prologue_8_16_32: 252 253 ldr r14, col_for_intra_chroma_addr_1 254 ulbl4: 255 add r14,r14,pc 256 257 lsr r10, r4, #3 258 vld1.8 d31, [r14]! 259 mul r10, r4, r10 @block counter (dec by #8) 260 261 mov r11, r4, lsl #1 @col counter to be inc/dec by #8 262 vmull.s8 q11, d30, d31 @(col+1)*intra_pred_angle [0:7](col) 263 264 sub r7, r5, #11 265 ldr r12, idx_neg_idx_chroma_11_17_addr @load least idx table 266 ulbl3: 267 add r12,r12,pc 268 269 add r12, r12, r7, lsl #4 270 mov r8, r12 271 272 mov r7, #8 273 sub r7, r7, r3, lsl #3 @r7 = 8-8r3 274 275 ldr r9, [r8] 276 mov r9,r9,lsl #1 277 add r1, sp, r4, lsl #1 @ref_temp + 2nt 278 279 vmovn.s16 d6, q11 280 vdup.8 d26, r9 @least idx added to final idx values 281 sub r1, r1, #2 @ref_temp + 2nt - 2 282 283 add r6, r1, r9 284 285 vld1.8 {d0,d1,d2,d3}, [r6] @stores the 32 values reqd based on indices values (from least idx) 286 vshr.s16 q11, q11, #5 287 288 @ mov r0, #31 289 vmov.i8 d29, #31 @contains #31 for vand operation 290 291 @ mov r0, #32 292 vmov.i8 d28, #32 293 294 vqmovn.s16 d8, q11 295 vshl.s8 d8, d8, #1 @ 2 * idx 296 297 vand d6, d6, d29 @fract values in d1/ idx values in d0 298 299 @ mov r0, #2 300 vmov.i8 d29, #2 @contains #2 for adding to get ref_main_idx + 1 301 302 mov r0,#0x100 @ idx value for v is +1 of u 303 vdup.u16 d27,r0 304 vadd.u8 d27,d27,d29 305 mov r0,#0 306 307 vadd.s8 d8, d8, d27 @ref_main_idx (add row) 308 vsub.s8 d8, d8, d26 @ref_main_idx (row 0) 309 vadd.s8 d9, d8, d29 @ref_main_idx + 1 (row 0) 310 vtbl.8 d12, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 0) 311 vsub.s8 d7, d28, d6 @32-fract 312 313 vtbl.8 d13, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 0) 314 vadd.s8 d4, d8, d29 @ref_main_idx (row 1) 315 vadd.s8 d5, d9, d29 @ref_main_idx + 1 (row 1) 316 317 @ mov r0, #4 @ 2 *(row * 2 ) 318 vmov.i8 d29, #4 319 320 vtbl.8 d16, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 1) 321 vmull.u8 q12, d12, d7 @mul (row 0) 322 vmlal.u8 q12, d13, d6 @mul (row 0) 323 324 vtbl.8 d17, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 1) 325 vadd.s8 d8, d8, d29 @ref_main_idx (row 2) 326 vadd.s8 d9, d9, d29 @ref_main_idx + 1 (row 2) 327 328 vrshrn.i16 d24, q12, #5 @round shft (row 0) 329 330 vtbl.8 d14, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 2) 331 vmull.u8 q11, d16, d7 @mul (row 1) 332 vmlal.u8 q11, d17, d6 @mul (row 1) 333 334 vtbl.8 d15, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 2) 335 vadd.s8 d4, d4, d29 @ref_main_idx (row 3) 336 vadd.s8 d5, d5, d29 @ref_main_idx + 1 (row 3) 337 338 vst1.8 d24, [r2], r3 @st (row 0) 339 vrshrn.i16 d22, q11, #5 @round shft (row 1) 340 341 vtbl.8 d10, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 3) 342 vmull.u8 q10, d14, d7 @mul (row 2) 343 vmlal.u8 q10, d15, d6 @mul (row 2) 344 345 vtbl.8 d11, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 3) 346 vadd.s8 d8, d8, d29 @ref_main_idx (row 4) 347 vadd.s8 d9, d9, d29 @ref_main_idx + 1 (row 4) 348 349 vst1.8 d22, [r2], r3 @st (row 1) 350 vrshrn.i16 d20, q10, #5 @round shft (row 2) 351 352 vtbl.8 d12, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 4) 353 vmull.u8 q9, d10, d7 @mul (row 3) 354 vmlal.u8 q9, d11, d6 @mul (row 3) 355 356 vtbl.8 d13, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 4) 357 vadd.s8 d4, d4, d29 @ref_main_idx (row 5) 358 vadd.s8 d5, d5, d29 @ref_main_idx + 1 (row 5) 359 360 vst1.8 d20, [r2], r3 @st (row 2) 361 vrshrn.i16 d18, q9, #5 @round shft (row 3) 362 363 vtbl.8 d16, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 5) 364 vmull.u8 q12, d12, d7 @mul (row 4) 365 vmlal.u8 q12, d13, d6 @mul (row 4) 366 367 vtbl.8 d17, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 5) 368 vadd.s8 d8, d8, d29 @ref_main_idx (row 6) 369 vadd.s8 d9, d9, d29 @ref_main_idx + 1 (row 6) 370 371 vst1.8 d18, [r2], r3 @st (row 3) 372 cmp r4,#4 373 beq end_func 374 vrshrn.i16 d24, q12, #5 @round shft (row 4) 375 376 vtbl.8 d14, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 6) 377 vmull.u8 q11, d16, d7 @mul (row 5) 378 vmlal.u8 q11, d17, d6 @mul (row 5) 379 380 vtbl.8 d15, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 6) 381 vadd.s8 d4, d4, d29 @ref_main_idx (row 7) 382 vadd.s8 d5, d5, d29 @ref_main_idx + 1 (row 7) 383 384 vst1.8 d24, [r2], r3 @st (row 4) 385 vrshrn.i16 d22, q11, #5 @round shft (row 5) 386 387 vtbl.8 d10, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 7) 388 vmull.u8 q10, d14, d7 @mul (row 6) 389 vmlal.u8 q10, d15, d6 @mul (row 6) 390 391 vtbl.8 d11, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 7) 392 vmull.u8 q9, d10, d7 @mul (row 7) 393 vmlal.u8 q9, d11, d6 @mul (row 7) 394 395 vst1.8 d22, [r2], r3 @st (row 5) 396 vrshrn.i16 d20, q10, #5 @round shft (row 6) 397 vrshrn.i16 d18, q9, #5 @round shft (row 7) 398 399 vst1.8 d20, [r2], r3 @st (row 6) 400 401 subs r10, r10, #4 @subtract 8 and go to end if 8x8 402 403 vst1.8 d18, [r2], r3 @st (row 7) 404 405 beq end_func 406 407 subs r11, r11, #8 408 addgt r8, r8, #4 409 addgt r2, r2, r7 410 movle r8, r12 411 suble r2, r2, r4 412 addle r2, r2, #8 413 movle r11, r4, lsl #1 414 ldrle r14, col_for_intra_chroma_addr_2 415 ulbl5: 416 addle r14,r14,pc 417 addle r0, r0, #8 418 419 vld1.8 d31, [r14]! 420 vmull.s8 q6, d30, d31 @(col+1)*intra_pred_angle [0:7](col) 421 vmovn.s16 d10, q6 422 vshr.s16 q6, q6, #5 423 vqmovn.s16 d11, q6 424 vshl.s8 d11, d11, #1 425 orr r5,r0,r0, lsl#8 426 add r5,#0x002 427 add r5,#0x300 428 vdup.u16 d27, r5 @row value inc or reset accordingly 429 ldr r9, [r8] 430 mov r9,r9,lsl #1 431 add r9, r9, r0, lsl #1 432 @ sub r9, r9, #1 433 vdup.8 d26, r9 434 vadd.s8 d8, d27, d11 @ref_main_idx (add row) 435 mov r5,r2 436 437 @ sub r4,r4,#8 438 439 kernel_8_16_32: 440 vmov.i8 d29, #2 @contains #2 for adding to get ref_main_idx + 1 441 442 vsub.s8 d8, d8, d26 @ref_main_idx 443 vmov d26,d10 444 445 subs r11, r11, #8 446 add r6, r1, r9 447 vtbl.8 d10, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 7) 448 vadd.s8 d9, d29, d8 @ref_main_idx + 1 449 450 vmull.u8 q10, d14, d7 @mul (row 6) 451 vtbl.8 d11, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 7) 452 vmlal.u8 q10, d15, d6 @mul (row 6) 453 454 addle r0, r0, #8 455 addgt r8, r8, #4 456 vld1.8 {d0,d1,d2,d3}, [r6] @stores the 32 values reqd based on indices values (from least idx) 457 458 vst1.8 d24, [r5], r3 @st (row 4) 459 vrshrn.i16 d24, q11, #5 @round shft (row 5) 460 461 movle r8, r12 462 orr r9,r0,r0, lsl#8 463 mov r9,r9,lsl #1 464 add r9,#0x002 465 add r9,#0x300 466 vdup.u16 d27, r9 @row value inc or reset accordingly 467 468 ldrle r14, col_for_intra_chroma_addr_3 469 ulbl6: 470 addle r14,r14,pc 471 472 vadd.s8 d4, d29, d8 @ref_main_idx (row 1) 473 vtbl.8 d12, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 0) 474 vadd.s8 d5, d29, d9 @ref_main_idx + 1 (row 1) 475 476 vmov.i8 d29, #31 @contains #2 for adding to get ref_main_idx + 1 477 478 vmull.u8 q9, d10, d7 @mul (row 7) 479 vtbl.8 d13, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 0) 480 vmlal.u8 q9, d11, d6 @mul (row 7) 481 482 vld1.8 d31, [r14]! 483 vand d6, d29, d26 @fract values in d1/ idx values in d0 484 485 vmov.i8 d29, #4 @contains #2 for adding to get ref_main_idx + 1 486 487 vst1.8 d24, [r5], r3 @(from previous loop)st (row 5) 488 vrshrn.i16 d20, q10, #5 @(from previous loop)round shft (row 6) 489 490 vadd.s8 d8, d29, d8 @ref_main_idx (row 2) 491 vtbl.8 d16, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 1) 492 vadd.s8 d9, d29, d9 @ref_main_idx + 1 (row 2) 493 494 movle r11, r4,lsl #1 495 ldr r9, [r8] 496 mov r9,r9,lsl #1 497 vsub.s8 d7, d28, d6 @32-fract 498 499 vmull.u8 q12, d12, d7 @mul (row 0) 500 vtbl.8 d17, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 1) 501 vmlal.u8 q12, d13, d6 @mul (row 0) 502 503 vst1.8 d20, [r5], r3 @(from previous loop)st (row 6) 504 vrshrn.i16 d18, q9, #5 @(from previous loop)round shft (row 7) 505 506 vadd.s8 d4, d4, d29 @ref_main_idx (row 3) 507 vtbl.8 d14, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 2) 508 vadd.s8 d5, d5, d29 @ref_main_idx + 1 (row 3) 509 510 vmull.u8 q11, d16, d7 @mul (row 1) 511 vtbl.8 d15, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 2) 512 vmlal.u8 q11, d17, d6 @mul (row 1) 513 514 vrshrn.i16 d24, q12, #5 @round shft (row 0) 515 vst1.8 d18, [r5], r3 @(from previous loop)st (row 7) 516 517 vadd.s8 d8, d8, d29 @ref_main_idx (row 4) 518 vtbl.8 d10, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 3) 519 vadd.s8 d9, d9, d29 @ref_main_idx + 1 (row 4) 520 521 vmull.u8 q10, d14, d7 @mul (row 2) 522 vtbl.8 d11, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 3) 523 vmlal.u8 q10, d15, d6 @mul (row 2) 524 525 vmull.s8 q7, d30, d31 @(col+1)*intra_pred_angle [0:7](col) 526 add r5,r2,r3,lsl#2 527 add r9, r9, r0, lsl #1 528 529 530 vst1.8 d24, [r2], r3 @st (row 0) 531 vrshrn.i16 d22, q11, #5 @round shft (row 1) 532 533 vadd.s8 d4, d4, d29 @ref_main_idx (row 5) 534 vtbl.8 d12, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 4) 535 vadd.s8 d5, d5, d29 @ref_main_idx + 1 (row 5) 536 537 vmull.u8 q9, d10, d7 @mul (row 3) 538 vtbl.8 d13, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 4) 539 vmlal.u8 q9, d11, d6 @mul (row 3) 540 541 vst1.8 d22, [r2], r3 @st (row 1) 542 vrshrn.i16 d20, q10, #5 @round shft (row 2) 543 544 vmovn.s16 d10, q7 545 vshr.s16 q7, q7, #5 546 547 vadd.s8 d8, d8, d29 @ref_main_idx (row 6) 548 vtbl.8 d16, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 5) 549 vadd.s8 d9, d9, d29 @ref_main_idx + 1 (row 6) 550 551 vmull.u8 q12, d12, d7 @mul (row 4) 552 vtbl.8 d17, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 5) 553 vmlal.u8 q12, d13, d6 @mul (row 4) 554 555 vst1.8 d20, [r2], r3 @st (row 2) 556 vrshrn.i16 d18, q9, #5 @round shft (row 3) 557 558 @ sub r9, r9, #1 559 vqmovn.s16 d11, q7 560 561 vadd.s8 d4, d4, d29 @ref_main_idx (row 7) 562 vtbl.8 d14, {d0,d1,d2,d3}, d8 @load from ref_main_idx (row 6) 563 vadd.s8 d5, d5, d29 @ref_main_idx + 1 (row 7) 564 565 vshl.u8 d11,#1 566 567 vmull.u8 q11, d16, d7 @mul (row 5) 568 vtbl.8 d15, {d0,d1,d2,d3}, d9 @load from ref_main_idx + 1 (row 6) 569 vmlal.u8 q11, d17, d6 @mul (row 5) 570 571 vadd.s8 d8, d27, d11 @ref_main_idx (add row) 572 vdup.8 d26, r9 573 574 vst1.8 d18, [r2], r3 @st (row 3) 575 vrshrn.i16 d24, q12, #5 @round shft (row 4) 576 577 578 add r2,r3, lsl #2 579 addgt r2, r7, r2 580 suble r2, r2, r4, lsl #1 581 addle r2,r2,#8 582 583 subs r10, r10, #4 @subtract 8 and go to end if 8x8 584 585 bne kernel_8_16_32 586 epil_8_16_32: 587 588 vtbl.8 d10, {d0,d1,d2,d3}, d4 @load from ref_main_idx (row 7) 589 590 vmull.u8 q10, d14, d7 @mul (row 6) 591 vtbl.8 d11, {d0,d1,d2,d3}, d5 @load from ref_main_idx + 1 (row 7) 592 vmlal.u8 q10, d15, d6 @mul (row 6) 593 594 vst1.8 d24, [r5], r3 @st (row 4) 595 vrshrn.i16 d24, q11, #5 @round shft (row 5) 596 597 vmull.u8 q9, d10, d7 @mul (row 7) 598 vmlal.u8 q9, d11, d6 @mul (row 7) 599 600 vst1.8 d24, [r5], r3 @(from previous loop)st (row 5) 601 vrshrn.i16 d20, q10, #5 @(from previous loop)round shft (row 6) 602 603 vst1.8 d20, [r5], r3 @(from previous loop)st (row 6) 604 vrshrn.i16 d18, q9, #5 @(from previous loop)round shft (row 7) 605 606 vst1.8 d18, [r5], r3 @st (row 7) 607 608 end_func: 609 add sp, sp, #132 610 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 611 612 613 614 615 616 617