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_sao_edge_offset_class2.s 22@* 23@* ,:brief 24@* Contains function definitions for inter prediction interpolation. 25@* Functions are coded using NEON intrinsics and can be compiled using@ ARM 26@* RVCT 27@* 28@* ,:author 29@* Parthiban V 30@* 31@* ,:par List of Functions: 32@* 33@* 34@* ,:remarks 35@* None 36@* 37@******************************************************************************* 38@*/ 39@void ihevc_sao_edge_offset_class2(UWORD8 *pu1_src, 40@ WORD32 src_strd, 41@ UWORD8 *pu1_src_left, 42@ UWORD8 *pu1_src_top, 43@ UWORD8 *pu1_src_top_left, 44@ UWORD8 *pu1_src_top_right, 45@ UWORD8 *pu1_src_bot_left, 46@ UWORD8 *pu1_avail, 47@ WORD8 *pi1_sao_offset, 48@ WORD32 wd, 49@ WORD32 ht) 50@**************Variables Vs Registers***************************************** 51@r0 => *pu1_src 52@r1 => src_strd 53@r2 => *pu1_src_left 54@r3 => *pu1_src_top 55@r4 => *pu1_src_top_left 56@r5 => *pu1_avail 57@r6 => *pi1_sao_offset 58@r7 => wd 59@r8=> ht 60 61.equ pu1_src_top_left_offset, 264 62.equ pu1_src_top_right_offset, 268 63.equ pu1_src_bot_left_offset, 272 64.equ pu1_avail_offset, 276 65.equ pi1_sao_offset, 280 66.equ wd_offset, 284 67.equ ht_offset, 288 68 69.text 70.syntax unified 71.p2align 2 72 73.extern gi1_table_edge_idx 74.globl ihevc_sao_edge_offset_class2_a9q 75 76gi1_table_edge_idx_addr_1: 77.long gi1_table_edge_idx - ulbl1 - 8 78 79gi1_table_edge_idx_addr_2: 80.long gi1_table_edge_idx - ulbl2 - 8 81 82gi1_table_edge_idx_addr_3: 83.long gi1_table_edge_idx - ulbl3 - 8 84 85ihevc_sao_edge_offset_class2_a9q: 86 87 88 STMFD sp!,{r4-r12,r14} @stack stores the values of the arguments 89 vpush {d8 - d15} 90 SUB sp,sp,#160 @Decrement the stack pointer to store some temp arr values 91 92 LDR r7,[sp,#wd_offset] @Loads wd 93 LDR r8,[sp,#ht_offset] @Loads ht 94 SUB r9,r7,#1 @wd - 1 95 96 LDR r4,[sp,#pu1_src_top_left_offset] @Loads pu1_src_top_left 97 LDRB r10,[r3,r9] @pu1_src_top[wd - 1] 98 99 STR r0,[sp,#152] @Store pu1_src in sp 100 MOV r9,r7 @Move width to r9 for loop count 101 102 STR r2,[sp,#156] @Store pu1_src_left in sp 103 LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail 104 LDR r6,[sp,#pi1_sao_offset] @Loads pi1_sao_offset 105 STR r3,[sp,#148] @Store pu1_src_top in sp 106 107 108 STRB r10,[sp] @u1_src_top_left_tmp = pu1_src_top[wd - 1] 109 SUB r10,r8,#1 @ht-1 110 MLA r11,r10,r1,r0 @pu1_src[(ht - 1) * src_strd + col] 111 ADD r12,sp,#2 @temp array 112 113AU1_SRC_TOP_LOOP: 114 VLD1.8 D0,[r11]! @pu1_src[(ht - 1) * src_strd + col] 115 SUBS r9,r9,#8 @Decrement the loop count by 8 116 VST1.8 D0,[r12]! @au1_src_top_tmp[col] = pu1_src[(ht - 1) * src_strd + col] 117 BNE AU1_SRC_TOP_LOOP 118 119PU1_AVAIL_4_LOOP: 120 LDRB r10,[r5,#4] @pu1_avail[4] 121 CMP r10,#0 122 LDRB r9,[r0] @u1_pos_0_0_tmp = pu1_src[0] 123 BEQ PU1_AVAIL_7_LOOP 124 125 LDRB r11,[r4] @pu1_src_top_left[0] 126 ADD r14,r0,r1 @pu1_src + src_strd 127 128 SUBS r12,r9,r11 @pu1_src[0] - pu1_src_top_left[0] 129 LDRB r4,[r14,#1] @pu1_src[1 + src_strd] 130 131 MVNLT r12,#0 132 MOVGT r12,#1 @SIGN(pu1_src[0] - pu1_src_top_left[0]) 133 134 LDR r14, gi1_table_edge_idx_addr_1 @table pointer 135ulbl1: 136 add r14,r14,pc 137 SUBS r11,r9,r4 @pu1_src[0] - pu1_src[1 + src_strd] 138 139 MVNLT r11,#0 140 MOVGT r11,#1 @SIGN(pu1_src[0] - pu1_src[1 + src_strd]) 141 ADD r4,r12,r11 @SIGN(pu1_src[0] - pu1_src_top_left[0]) + SIGN(pu1_src[0] - pu1_src[1 + src_strd]) 142 ADD r4,r4,#2 @edge_idx 143 144 LDRSB r12,[r14,r4] @edge_idx = gi1_table_edge_idx[edge_idx] 145 CMP r12,#0 @0 != edge_idx 146 BEQ PU1_AVAIL_7_LOOP 147 LDRSB r10,[r6,r12] @pi1_sao_offset[edge_idx] 148 ADD r9,r9,r10 @pu1_src[0] + pi1_sao_offset[edge_idx] 149 USAT r9,#8,r9 @u1_pos_0_0_tmp = CLIP3(pu1_src[0] + pi1_sao_offset[edge_idx], 0, (1 << bit_depth) - 1) 150 151PU1_AVAIL_7_LOOP: 152 LDRB r14,[r5,#7] @pu1_avail[7] 153 CMP r14,#0 154 SUB r10,r7,#1 @wd - 1 155 SUB r11,r8,#1 @ht - 1 156 MLA r12,r11,r1,r10 @wd - 1 + (ht - 1) * src_strd 157 ADD r12,r12,r0 @pu1_src[wd - 1 + (ht - 1) * src_strd] 158 LDRB r10,[r12] @u1_pos_wd_ht_tmp = pu1_src[wd - 1 + (ht - 1) * src_strd] 159 BEQ PU1_AVAIL 160 161 SUB r4,r12,r1 @pu1_src[(wd - 1 + (ht - 1) * src_strd) - src_strd] 162 LDRB r11,[r4,#-1] @Load pu1_src[wd - 1 + (ht - 1) * src_strd - 1 - src_strd] 163 ADD r14,r12,r1 @pu1_src[(wd - 1 + (ht - 1) * src_strd) + src_strd] 164 165 SUBS r11,r10,r11 @pu1_src[wd - 1 + (ht - 1) * src_strd] - pu1_src[wd - 1 + (ht - 1) * src_strd- 1 - src_strd] 166 LDRB r4,[r14,#1] @Load pu1_src[wd - 1 + (ht - 1) * src_strd + 1 + src_strd] 167 168 MVNLT r11,#0 169 MOVGT r11,#1 @SIGN(pu1_src[wd - 1 + (ht - 1) * src_strd] - pu1_src[wd - 1 + (ht - 1) * src_strd- 1 - src_strd]) 170 171 SUBS r4,r10,r4 @pu1_src[wd - 1 + (ht - 1) * src_strd] - pu1_src[wd - 1 + (ht - 1) * src_strd + 1 + src_strd] 172 MVNLT r4,#0 173 MOVGT r4,#1 @SIGN(pu1_src[wd - 1 + (ht - 1) * src_strd] - pu1_src[wd - 1 + (ht - 1) * src_strd + 1 + src_strd]) 174 175 ADD r11,r11,r4 @Add 2 sign value 176 ADD r11,r11,#2 @edge_idx 177 LDR r14, gi1_table_edge_idx_addr_2 @table pointer 178ulbl2: 179 add r14,r14,pc 180 181 LDRSB r12,[r14,r11] @edge_idx = gi1_table_edge_idx[edge_idx] 182 CMP r12,#0 183 BEQ PU1_AVAIL 184 LDRSB r11,[r6,r12] @pi1_sao_offset[edge_idx] 185 ADD r10,r10,r11 @pu1_src[wd - 1 + (ht - 1) * src_strd] + pi1_sao_offset[edge_idx] 186 USAT r10,#8,r10 @u1_pos_wd_ht_tmp = CLIP3(pu1_src[wd - 1 + (ht - 1) * src_strd] + pi1_sao_offset[edge_idx], 0, (1 << bit_depth) - 1) 187 188PU1_AVAIL: 189 MOV r12,r8 @Move ht 190 VMOV.I8 Q0,#2 @const_2 = vdupq_n_s8(2) 191 LDRB r11,[r5,#3] @pu1_avail[3] 192 193 MOV r14,r2 @Move pu1_src_left to pu1_src_left_cpy 194 VMOV.I16 Q1,#0 @const_min_clip = vdupq_n_s16(0) 195 CMP r11,#0 196 197 LDRB r5,[r5,#2] @pu1_avail[2] 198 VMOV.I16 Q2,#255 @const_max_clip = vdupq_n_u16((1 << bit_depth) - 1) 199 SUBEQ r12,r12,#1 @ht_tmp-- 200 201 CMP r5,#0 202 VLD1.8 D7,[r6] @offset_tbl = vld1_s8(pi1_sao_offset) 203 LDR r11, gi1_table_edge_idx_addr_3 @table pointer 204ulbl3: 205 add r11,r11,pc 206 207 ADDEQ r0,r0,r1 @pu1_src += src_strd 208 VLD1.8 D6,[r11] @edge_idx_tbl = vld1_s8(gi1_table_edge_idx) 209 SUBEQ r12,r12,#1 @ht_tmp-- 210 211 MOV r6,r7 @move wd to r6 loop_count 212 VMOV.S8 Q4,#0xFF @au1_mask = vdupq_n_s8(-1) 213 ADDEQ r14,r14,#1 @pu1_src_left_cpy += 1 214 215 STR r0,[sp,#144] @Store pu1_src in sp 216 CMP r7,#16 @Compare wd with 16 217 218 BLT WIDTH_RESIDUE @If not jump to WIDTH_RESIDUE where loop is unrolled for 8 case 219 CMP r8,#4 @Compare ht with 4 220 BLE WD_16_HT_4_LOOP @If jump to WD_16_HT_4_LOOP 221 222WIDTH_LOOP_16: 223 LDR r7,[sp,#wd_offset] @Loads wd 224 225 LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail 226 CMP r6,r7 @col == wd 227 LDRBEQ r8,[r5] @pu1_avail[0] 228 MOVNE r8,#-1 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0) 229 230 VMOV.8 d8[0],r8 @au1_mask = vsetq_lane_s8((-1||pu1_avail[0]), au1_mask, 0) 231 CMP r6,#16 @if(col == 16) 232 BNE SKIP_AU1_MASK_VAL 233 LDRB r8,[r5,#1] @pu1_avail[1] 234 VMOV.8 d9[7],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15) 235 236SKIP_AU1_MASK_VAL: 237 LDRB r11,[r5,#2] @pu1_avail[2] 238 CMP r11,#0 239 240 SUBEQ r8,r0,r1 @pu1_src - src_strd 241 MOVNE r8,r3 @pu1_src_top_cpy 242 SUB r8,r8,#1 @pu1_src_top_cpy - 1 || pu1_src - src_strd - 1 243 244 LDR r7,[sp,#wd_offset] @Loads wd 245 VLD1.8 D10,[r8]! @pu1_top_row = vld1q_u8(pu1_src - src_strd - 1) || vld1q_u8(pu1_src_top_cpy - 1) 246 VLD1.8 D11,[r8] @pu1_top_row = vld1q_u8(pu1_src - src_strd - 1) || vld1q_u8(pu1_src_top_cpy - 1) 247 SUB r8,#8 248 ADD r3,r3,#16 249 250 ADD r5,sp,#66 @*au1_src_left_tmp 251 VLD1.8 D12,[r0]! @pu1_cur_row = vld1q_u8(pu1_src) 252 VLD1.8 D13,[r0] @pu1_cur_row = vld1q_u8(pu1_src) 253 SUB r0,#8 254 LDR r4,[sp,#ht_offset] @Loads ht 255 256 SUB r7,r7,r6 @(wd - col) 257 VCGT.U8 Q7,Q6,Q5 @vcgtq_u8(pu1_cur_row, pu1_top_row) 258 LDR r8,[sp,#152] @Loads *pu1_src 259 260 ADD r7,r7,#15 @15 + (wd - col) 261 VCLT.U8 Q8,Q6,Q5 @vcltq_u8(pu1_cur_row, pu1_top_row) 262 ADD r7,r8,r7 @pu1_src[0 * src_strd + 15 + (wd - col)] 263 264 SUB r5,r5,#1 265 VSUB.U8 Q7,Q8,Q7 @sign_up = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 266 267AU1_SRC_LEFT_LOOP: 268 LDRB r8,[r7],r1 @load the value and increment by src_strd 269 STRB r8,[r5,#1]! @store it in the stack pointer 270 SUBS r4,r4,#1 @decrement the loop count 271 BNE AU1_SRC_LEFT_LOOP 272 273 ADD r8,r0,r1 @I Iteration *pu1_src + src_strd 274 VMOV.I8 Q9,#0 275 LDR r4,[sp,#pu1_avail_offset] @I Loads pu1_avail 276 277 MOV r7,r12 @row count, move ht_tmp to r7 278 VLD1.8 D16,[r8]! @I pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 279 VLD1.8 D17,[r8] @I pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 280 SUB r8,#8 281 LDRB r4,[r4,#2] @I pu1_avail[2] 282 283 LDRB r5,[r8,#16] @I pu1_src_cpy[src_strd + 16] 284 VMOV.8 D18[0],r5 @I pu1_next_row_tmp = vsetq_lane_u8(pu1_src_cpy[src_strd + 16], pu1_next_row_tmp, 0) 285 286 VEXT.8 Q9,Q8,Q9,#1 @I pu1_next_row_tmp = vextq_u8(pu1_next_row, pu1_next_row_tmp, 1) 287 CMP r4,#0 @I 288 BNE SIGN_UP_CHANGE_DONE @I 289 290SIGN_UP_CHANGE: 291 SUB r2,r12,r7 @I ht_tmp - row 292 LDRB r11,[r0] @I pu1_src_cpy[0] 293 ADD r2,r14,r2 @I pu1_src_left_cpy[ht_tmp - row] 294 295 LDRB r5,[r2,#-1] @I load the value 296 SUBS r4,r11,r5 @I pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row] 297 MVNLT r4,#0 @I 298 MOVGT r4,#1 @I SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]) 299 VMOV.8 D14[0],r4 @I sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]), sign_up, 0) 300 301SIGN_UP_CHANGE_DONE: 302 VCGT.U8 Q5,Q6,Q9 @I vcgtq_u8(pu1_cur_row, pu1_next_row_tmp) 303 VADD.I8 Q12,Q0,Q7 @I edge_idx = vaddq_s8(const_2, sign_up) 304 305 VCLT.U8 Q9,Q6,Q9 @I vcltq_u8(pu1_cur_row, pu1_next_row_tmp) 306 VSUB.U8 Q5,Q9,Q5 @I sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 307 308 VADD.I8 Q12,Q12,Q5 @I edge_idx = vaddq_s8(edge_idx, sign_down) 309 VTBL.8 D18,{D6},D24 @I vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx)) 310 VTBL.8 D19,{D6},D25 @I vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx)) 311 312 VAND Q9,Q9,Q4 @I edge_idx = vandq_s8(edge_idx, au1_mask) 313 314 VNEG.S8 Q7,Q5 @I sign_up = vnegq_s8(sign_down) 315 VTBL.8 D10,{D7},D18 @I offset = vtbl1_s8(offset_tbl, vget_low_s8(edge_idx)) 316 VEXT.8 Q7,Q7,Q7,#15 @I sign_up = vextq_s8(sign_up, sign_up, 15) 317 318 VMOVL.U8 Q10,D12 @I pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row))) 319 VTBL.8 D11,{D7},D19 @I offset = vtbl1_s8(offset_tbl, vget_high_s8(edge_idx)) 320 VADDW.S8 Q10,Q10,D10 @I pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset) 321 322 VMAX.S16 Q10,Q10,Q1 @I pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip) 323 VMOVL.U8 Q11,D13 @I pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row))) 324 325 VMIN.U16 Q10,Q10,Q2 @I pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip)) 326 VMOV Q6,Q8 @I pu1_cur_row = pu1_next_row 327 328 VADDW.S8 Q11,Q11,D11 @I pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset) 329 VMOVN.I16 D20,Q10 @I vmovn_s16(pi2_tmp_cur_row.val[0]) 330 331 VMAX.S16 Q11,Q11,Q1 @I pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip) 332 SUB r7,r7,#1 @I Decrement the ht_tmp loop count by 1 333 334 VMIN.U16 Q11,Q11,Q2 @I pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip)) 335 336 VMOVN.I16 D21,Q11 @I vmovn_s16(pi2_tmp_cur_row.val[1]) 337 338PU1_SRC_LOOP: 339 340 VST1.8 {Q10},[r0],r1 @I vst1q_u8(pu1_src_cpy, pu1_cur_row) 341 ADD r8,r0,r1 @II iteration *pu1_src + src_strd 342 343 VLD1.8 D16,[r8]! @II pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 344 VLD1.8 D17,[r8] @II pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 345 SUB r8,#8 346 ADD r11,r8,r1 @III iteration *pu1_src + src_strd 347 348 LDRB r5,[r8,#16] @II pu1_src_cpy[src_strd + 16] 349 VLD1.8 D30,[r11]! @III pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 350 VLD1.8 D31,[r11] @III pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 351 SUB r11,#8 352 LDRB r4,[r0] @II pu1_src_cpy[0] 353 354 LDRB r8,[r11,#16] @III pu1_src_cpy[src_strd + 16] 355 VMOV.8 D28[0],r5 @II pu1_next_row_tmp = vsetq_lane_u8(pu1_src_cpy[src_strd + 16], pu1_next_row_tmp, 0) 356 357 SUB r5,r12,r7 @II ht_tmp - row 358 VEXT.8 Q11,Q8,Q14,#1 @II pu1_next_row_tmp = vextq_u8(pu1_next_row, pu1_next_row_tmp, 1) 359 ADD r5,r14,r5 @II pu1_src_left_cpy[ht_tmp - row] 360 361 LDRB r5,[r5,#-1] @II load the value 362 VMOV.8 D18[0],r8 @III pu1_next_row_tmp = vsetq_lane_u8(pu1_src_cpy[src_strd + 16], pu1_next_row_tmp, 0) 363 SUB r7,r7,#1 @II Decrement the ht_tmp loop count by 1 364 365 SUBS r4,r4,r5 @II pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row] 366 VEXT.8 Q9,Q15,Q9,#1 @III pu1_next_row_tmp = vextq_u8(pu1_next_row, pu1_next_row_tmp, 1) 367 LDRB r2,[r0,r1] @III pu1_src_cpy[0] 368 369 VCGT.U8 Q12,Q6,Q11 @II vcgtq_u8(pu1_cur_row, pu1_next_row_tmp) 370 SUB r5,r12,r7 @III ht_tmp - row 371 372 MVNLT r4,#0 @II 373 VCLT.U8 Q11,Q6,Q11 @II vcltq_u8(pu1_cur_row, pu1_next_row_tmp) 374 ADD r5,r14,r5 @III pu1_src_left_cpy[ht_tmp - row] 375 376 MOVGT r4,#1 @II SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]) 377 VSUB.U8 Q12,Q11,Q12 @II sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 378 LDRB r5,[r5,#-1] @III load the value 379 380 SUBS r2,r2,r5 @III pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row] 381 VMOV.8 D14[0],r4 @II sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]), sign_up, 0) 382 383 MVNLT r2,#0 @III 384 VCGT.U8 Q5,Q8,Q9 @III vcgtq_u8(pu1_cur_row, pu1_next_row_tmp) 385 MOVGT r2,#1 @III SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]) 386 387 VADD.I8 Q11,Q0,Q7 @II edge_idx = vaddq_s8(const_2, sign_up) 388 VADD.I8 Q11,Q11,Q12 @II edge_idx = vaddq_s8(edge_idx, sign_down) 389 390 VCLT.U8 Q9,Q8,Q9 @III vcltq_u8(pu1_cur_row, pu1_next_row_tmp) 391 VTBL.8 D22,{D6},D22 @II vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx)) 392 VNEG.S8 Q7,Q12 @II sign_up = vnegq_s8(sign_down) 393 394 VSUB.U8 Q5,Q9,Q5 @III sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 395 VTBL.8 D23,{D6},D23 @II vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx)) 396 VEXT.8 Q7,Q7,Q7,#15 @II sign_up = vextq_s8(sign_up, sign_up, 15) 397 398 VAND Q11,Q11,Q4 @II edge_idx = vandq_s8(edge_idx, au1_mask) 399 VMOV.8 D14[0],r2 @III sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]), sign_up, 0) 400 401 VADD.I8 Q9,Q0,Q7 @III edge_idx = vaddq_s8(const_2, sign_up) 402 VTBL.8 D24,{D7},D22 @II offset = vtbl1_s8(offset_tbl, vget_low_s8(edge_idx)) 403 VADD.I8 Q9,Q9,Q5 @III edge_idx = vaddq_s8(edge_idx, sign_down) 404 405 VMOVL.U8 Q13,D12 @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row))) 406 VTBL.8 D18,{D6},D18 @III vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx)) 407 VNEG.S8 Q7,Q5 @III sign_up = vnegq_s8(sign_down) 408 409 VADDW.S8 Q13,Q13,D24 @II pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset) 410 VTBL.8 D19,{D6},D19 @III vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx)) 411 VEXT.8 Q7,Q7,Q7,#15 @III sign_up = vextq_s8(sign_up, sign_up, 15) 412 413 VAND Q9,Q9,Q4 @III edge_idx = vandq_s8(edge_idx, au1_mask) 414 VMOVL.U8 Q10,D16 @III pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row))) 415 416 VMAX.S16 Q13,Q13,Q1 @II pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip) 417 VTBL.8 D10,{D7},D18 @III offset = vtbl1_s8(offset_tbl, vget_low_s8(edge_idx)) 418 VADDW.S8 Q10,Q10,D10 @III pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset) 419 420 VMIN.U16 Q13,Q13,Q2 @II pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip)) 421 VTBL.8 D25,{D7},D23 @II offset = vtbl1_s8(offset_tbl, vget_high_s8(edge_idx)) 422 VMAX.S16 Q10,Q10,Q1 @III pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip) 423 424 VMOVL.U8 Q14,D13 @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row))) 425 VMIN.U16 Q10,Q10,Q2 @III pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip)) 426 427 VADDW.S8 Q14,Q14,D25 @II pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset) 428 VTBL.8 D11,{D7},D19 @III offset = vtbl1_s8(offset_tbl, vget_high_s8(edge_idx)) 429 VMAX.S16 Q14,Q14,Q1 @II pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip) 430 431 VMIN.U16 Q14,Q14,Q2 @II pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip)) 432 VMOVL.U8 Q9,D17 @III pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row))) 433 434 VMOV Q6,Q15 @III pu1_cur_row = pu1_next_row 435 VMOVN.I16 D26,Q13 @II vmovn_s16(pi2_tmp_cur_row.val[0]) 436 437 VMOVN.I16 D27,Q14 @II vmovn_s16(pi2_tmp_cur_row.val[1]) 438 VADDW.S8 Q9,Q9,D11 @III pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset) 439 440 VMAX.S16 Q9,Q9,Q1 @III pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip) 441 VMOVN.I16 D20,Q10 @III vmovn_s16(pi2_tmp_cur_row.val[0]) 442 443 SUB r7,r7,#1 @III Decrement the ht_tmp loop count by 1 444 VMIN.U16 Q9,Q9,Q2 @III pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip)) 445 CMP r7,#1 @III 446 447 VST1.8 {Q13},[r0],r1 @II vst1q_u8(pu1_src_cpy, pu1_cur_row) 448 VMOVN.I16 D21,Q9 @III vmovn_s16(pi2_tmp_cur_row.val[1]) 449 450 BGT PU1_SRC_LOOP @III If not equal jump to PU1_SRC_LOOP 451 BLT INNER_LOOP_DONE 452 453 VST1.8 {Q10},[r0],r1 @III vst1q_u8(pu1_src_cpy, pu1_cur_row) 454 ADD r8,r0,r1 @*pu1_src + src_strd 455 456 LDRB r2,[r0] @pu1_src_cpy[0] 457 VLD1.8 D16,[r8]! @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 458 VLD1.8 D17,[r8] @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 459 SUB r8,#8 460 LDRB r5,[r8,#16] @pu1_src_cpy[src_strd + 16] 461 462 SUB r11,r12,r7 @ht_tmp - row 463 VMOV.8 D18[0],r5 @pu1_next_row_tmp = vsetq_lane_u8(pu1_src_cpy[src_strd + 16], pu1_next_row_tmp, 0) 464 ADD r11,r14,r11 @pu1_src_left_cpy[ht_tmp - row] 465 466 LDRB r5,[r11,#-1] @load the value 467 VEXT.8 Q9,Q8,Q9,#1 @pu1_next_row_tmp = vextq_u8(pu1_next_row, pu1_next_row_tmp, 1) 468 SUBS r4,r2,r5 @pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row] 469 470 VCGT.U8 Q5,Q6,Q9 @vcgtq_u8(pu1_cur_row, pu1_next_row_tmp) 471 MVNLT r4,#0 472 473 MOVGT r4,#1 @SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]) 474 VCLT.U8 Q9,Q6,Q9 @vcltq_u8(pu1_cur_row, pu1_next_row_tmp) 475 476 VMOV.8 D14[0],r4 @sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]), sign_up, 0) 477 VSUB.U8 Q5,Q9,Q5 @sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 478 479 VADD.I8 Q9,Q0,Q7 @edge_idx = vaddq_s8(const_2, sign_up) 480 VADD.I8 Q9,Q9,Q5 @edge_idx = vaddq_s8(edge_idx, sign_down) 481 482 VTBL.8 D18,{D6},D18 @vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx)) 483 VNEG.S8 Q7,Q5 @sign_up = vnegq_s8(sign_down) 484 485 VTBL.8 D19,{D6},D19 @vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx)) 486 VEXT.8 Q7,Q7,Q7,#15 @sign_up = vextq_s8(sign_up, sign_up, 15) 487 488 VAND Q9,Q9,Q4 @edge_idx = vandq_s8(edge_idx, au1_mask) 489 490 VTBL.8 D10,{D7},D18 @offset = vtbl1_s8(offset_tbl, vget_low_s8(edge_idx)) 491 492 VMOVL.U8 Q10,D12 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row))) 493 VTBL.8 D11,{D7},D19 @offset = vtbl1_s8(offset_tbl, vget_high_s8(edge_idx)) 494 VADDW.S8 Q10,Q10,D10 @pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset) 495 496 VMAX.S16 Q10,Q10,Q1 @pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip) 497 VMOVL.U8 Q6,D13 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row))) 498 499 VMIN.U16 Q10,Q10,Q2 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip)) 500 VADDW.S8 Q6,Q6,D11 @pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset) 501 502 VMAX.S16 Q6,Q6,Q1 @pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip) 503 VMOVN.I16 D20,Q10 @vmovn_s16(pi2_tmp_cur_row.val[0]) 504 505 VMIN.U16 Q6,Q6,Q2 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip)) 506 VMOVN.I16 D21,Q6 @vmovn_s16(pi2_tmp_cur_row.val[1]) 507 508 509INNER_LOOP_DONE: 510 ADD r5,sp,#66 @*au1_src_left_tmp 511 VST1.8 {Q10},[r0],r1 @vst1q_u8(pu1_src_cpy, pu1_cur_row) 512 LDR r2,[sp,#156] @Loads *pu1_src_left 513 514 LDR r8,[sp,#ht_offset] @Loads ht 515 SUB r5,r5,#1 516 517 SUB r2,r2,#1 518SRC_LEFT_LOOP: 519 LDRB r7,[r5,#1]! @au1_src_left_tmp[row] 520 SUBS r8,r8,#1 521 STRB r7,[r2,#1]! @pu1_src_left[row] = au1_src_left_tmp[row] 522 BNE SRC_LEFT_LOOP 523 524 SUB r6,r6,#16 @Decrement the wd loop count by 16 525 CMP r6,#8 @Check whether residue remains 526 BLT RE_ASSINING_LOOP @Jump to re-assigning loop 527 LDR r7,[sp,#wd_offset] @Loads wd 528 LDR r0,[sp,#144] @Loads *pu1_src 529 SUB r7,r7,r6 530 ADD r0,r0,r7 531 BGT WIDTH_LOOP_16 @If not equal jump to width_loop 532 BEQ WIDTH_RESIDUE @If residue remains jump to residue loop 533 534 535WD_16_HT_4_LOOP: 536 LDR r7,[sp,#wd_offset] @Loads wd 537 LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail 538 CMP r6,r7 @col == wd 539 LDRBEQ r8,[r5] @pu1_avail[0] 540 MOVNE r8,#-1 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0) 541 542 VMOV.8 d8[0],r8 @au1_mask = vsetq_lane_s8((-1||pu1_avail[0]), au1_mask, 0) 543 CMP r6,#16 @if(col == 16) 544 BNE SKIP_AU1_MASK_VAL_WD_16_HT_4 545 LDRB r8,[r5,#1] @pu1_avail[1] 546 VMOV.8 d9[7],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15) 547 548SKIP_AU1_MASK_VAL_WD_16_HT_4: 549 LDRB r8,[r5,#2] @pu1_avail[2] 550 CMP r8,#0 551 552 SUBEQ r8,r0,r1 @pu1_src - src_strd 553 MOVNE r8,r3 554 SUB r8,r8,#1 @pu1_src_top_cpy - 1 || pu1_src - src_strd - 1 555 556 LDR r7,[sp,#wd_offset] @Loads wd 557 VLD1.8 D10,[r8]! @pu1_top_row = vld1q_u8(pu1_src - src_strd - 1) || vld1q_u8(pu1_src_top_cpy - 1) 558 VLD1.8 D11,[r8] @pu1_top_row = vld1q_u8(pu1_src - src_strd - 1) || vld1q_u8(pu1_src_top_cpy - 1) 559 SUB r8,#8 560 ADD r3,r3,#16 561 562 ADD r5,sp,#66 @*au1_src_left_tmp 563 VLD1.8 D12,[r0]! @pu1_cur_row = vld1q_u8(pu1_src) 564 VLD1.8 D13,[r0] @pu1_cur_row = vld1q_u8(pu1_src) 565 SUB r0,#8 566 LDR r4,[sp,#ht_offset] @Loads ht 567 568 SUB r7,r7,r6 @(wd - col) 569 VCGT.U8 Q7,Q6,Q5 @vcgtq_u8(pu1_cur_row, pu1_top_row) 570 LDR r8,[sp,#152] @Loads *pu1_src 571 572 ADD r7,r7,#15 @15 + (wd - col) 573 VCLT.U8 Q8,Q6,Q5 @vcltq_u8(pu1_cur_row, pu1_top_row) 574 ADD r7,r8,r7 @pu1_src[0 * src_strd + 15 + (wd - col)] 575 576 SUB r5,r5,#1 577 VSUB.U8 Q7,Q8,Q7 @sign_up = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 578 579AU1_SRC_LEFT_LOOP_WD_16_HT_4: 580 LDRB r8,[r7],r1 @load the value and increment by src_strd 581 SUBS r4,r4,#1 @decrement the loop count 582 STRB r8,[r5,#1]! @store it in the stack pointer 583 BNE AU1_SRC_LEFT_LOOP_WD_16_HT_4 584 585 VMOV.I8 Q9,#0 586 MOV r7,r12 @row count, move ht_tmp to r7 587 588PU1_SRC_LOOP_WD_16_HT_4: 589 ADD r8,r0,r1 @*pu1_src + src_strd 590 VLD1.8 D16,[r8]! @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 591 VLD1.8 D17,[r8] @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 592 SUB r8,#8 593 594 LDRB r5,[r8,#16] @pu1_src_cpy[src_strd + 16] 595 VMOV.8 D18[0],r5 @pu1_next_row_tmp = vsetq_lane_u8(pu1_src_cpy[src_strd + 16], pu1_next_row_tmp, 0) 596 VEXT.8 Q9,Q8,Q9,#1 @pu1_next_row_tmp = vextq_u8(pu1_next_row, pu1_next_row_tmp, 1) 597 598 CMP r7,r12 599 BLT SIGN_UP_CHANGE_WD_16_HT_4 600 LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail 601 LDRB r5,[r5,#2] @pu1_avail[2] 602 CMP r5,#0 603 BNE SIGN_UP_CHANGE_DONE_WD_16_HT_4 604 605SIGN_UP_CHANGE_WD_16_HT_4: 606 LDRB r8,[r0] @pu1_src_cpy[0] 607 SUB r5,r12,r7 @ht_tmp - row 608 ADD r5,r14,r5 @pu1_src_left_cpy[ht_tmp - row] 609 LDRB r5,[r5,#-1] @load the value 610 SUBS r8,r8,r5 @pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row] 611 MVNLT r8,#0 612 MOVGT r8,#1 @SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]) 613 VMOV.8 d14[0],r8 @sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]), sign_up, 0) 614 615SIGN_UP_CHANGE_DONE_WD_16_HT_4: 616 VCGT.U8 Q10,Q6,Q9 @vcgtq_u8(pu1_cur_row, pu1_next_row_tmp) 617 VCLT.U8 Q11,Q6,Q9 @vcltq_u8(pu1_cur_row, pu1_next_row_tmp) 618 VSUB.U8 Q12,Q11,Q10 @sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 619 620 VADD.I8 Q13,Q0,Q7 @edge_idx = vaddq_s8(const_2, sign_up) 621 VADD.I8 Q13,Q13,Q12 @edge_idx = vaddq_s8(edge_idx, sign_down) 622 VTBL.8 D26,{D6},D26 @vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx)) 623 VTBL.8 D27,{D6},D27 @vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx)) 624 625 VAND Q13,Q13,Q4 @edge_idx = vandq_s8(edge_idx, au1_mask) 626 627 VNEG.S8 Q7,Q12 @sign_up = vnegq_s8(sign_down) 628 VEXT.8 Q7,Q7,Q7,#15 @sign_up = vextq_s8(sign_up, sign_up, 15) 629 630 VTBL.8 D24,{D7},D26 @offset = vtbl1_s8(offset_tbl, vget_low_s8(edge_idx)) 631 VMOVL.U8 Q14,D12 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row))) 632 VADDW.S8 Q14,Q14,D24 @pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset) 633 VMAX.S16 Q14,Q14,Q1 @pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip) 634 VMIN.U16 Q14,Q14,Q2 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip)) 635 636 VTBL.8 D25,{D7},D27 @offset = vtbl1_s8(offset_tbl, vget_high_s8(edge_idx)) 637 VMOVL.U8 Q15,D13 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vmovl_u8(vget_high_u8(pu1_cur_row))) 638 VADDW.S8 Q15,Q15,D25 @pi2_tmp_cur_row.val[1] = vaddw_s8(pi2_tmp_cur_row.val[1], offset) 639 VMAX.S16 Q15,Q15,Q1 @pi2_tmp_cur_row.val[1] = vmaxq_s16(pi2_tmp_cur_row.val[1], const_min_clip) 640 VMIN.U16 Q15,Q15,Q2 @pi2_tmp_cur_row.val[1] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[1]), const_max_clip)) 641 642 VMOVN.I16 D28,Q14 @vmovn_s16(pi2_tmp_cur_row.val[0]) 643 VMOVN.I16 D29,Q15 @vmovn_s16(pi2_tmp_cur_row.val[1]) 644 645 VST1.8 {Q14},[r0],r1 @vst1q_u8(pu1_src_cpy, pu1_cur_row) 646 647 VMOV Q6,Q8 @pu1_cur_row = pu1_next_row 648 SUBS r7,r7,#1 @Decrement the ht_tmp loop count by 1 649 BNE PU1_SRC_LOOP_WD_16_HT_4 @If not equal jump to PU1_SRC_LOOP_WD_16_HT_4 650 651 LDR r8,[sp,#ht_offset] @Loads ht 652 ADD r5,sp,#66 @*au1_src_left_tmp 653 LDR r2,[sp,#156] @Loads *pu1_src_left 654 SUB r5,r5,#1 655 SUB r2,r2,#1 656 657SRC_LEFT_LOOP_WD_16_HT_4: 658 LDRB r7,[r5,#1]! @au1_src_left_tmp[row] 659 STRB r7,[r2,#1]! @pu1_src_left[row] = au1_src_left_tmp[row] 660 SUBS r8,r8,#1 661 BNE SRC_LEFT_LOOP_WD_16_HT_4 662 663 SUBS r6,r6,#16 @Decrement the wd loop count by 16 664 BLE RE_ASSINING_LOOP @Jump to re-assigning loop 665 666 667WIDTH_RESIDUE: 668 LDR r7,[sp,#wd_offset] @Loads wd 669 LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail 670 CMP r6,r7 @wd_residue == wd 671 LDRBEQ r8,[r5] @pu1_avail[0] 672 673 MOVNE r8,#-1 674 VMOV.8 d8[0],r8 @au1_mask = vsetq_lane_s8(-1, au1_mask, 0) 675 676 LDRB r8,[r5,#1] @pu1_avail[1] 677 VMOV.8 d8[7],r8 @au1_mask = vsetq_lane_s8(pu1_avail[1], au1_mask, 15) 678 679PU1_AVAIL_2_RESIDUE: 680 LDRB r11,[r5,#2] @pu1_avail[2] 681 VLD1.8 D12,[r0]! @pu1_cur_row = vld1q_u8(pu1_src) 682 VLD1.8 D13,[r0] @pu1_cur_row = vld1q_u8(pu1_src) 683 SUB r0,#8 684 CMP r11,#0 685 686 SUBEQ r8,r0,r1 @pu1_src - src_strd 687 MOVNE r8,r3 688 689 SUB r8,r8,#1 690 691 ADD r5,sp,#66 @*au1_src_left_tmp 692 VLD1.8 D10,[r8]! @pu1_top_row = vld1q_u8(pu1_src_top_cpy - 1) 693 VLD1.8 D11,[r8]! @pu1_top_row = vld1q_u8(pu1_src_top_cpy - 1) 694 LDR r7,[sp,#wd_offset] @Loads wd 695 696 LDR r4,[sp,#ht_offset] @Loads ht 697 VCGT.U8 Q7,Q6,Q5 @vcgtq_u8(pu1_cur_row, pu1_top_row) 698 SUB r7,r7,#1 @(wd - 1) 699 700 LDR r8,[sp,#152] @Loads *pu1_src 701 VCLT.U8 Q8,Q6,Q5 @vcltq_u8(pu1_cur_row, pu1_top_row) 702 SUB r5,r5,#1 703 704 ADD r7,r8,r7 @pu1_src[0 * src_strd + (wd - 1)] 705 VSUB.U8 Q7,Q8,Q7 @sign_up = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 706 707 708AU1_SRC_LEFT_LOOP_RESIDUE: 709 LDRB r8,[r7],r1 @load the value and increment by src_strd 710 SUBS r4,r4,#1 @decrement the loop count 711 STRB r8,[r5,#1]! @store it in the stack pointer 712 BNE AU1_SRC_LEFT_LOOP_RESIDUE 713 714 715 MOV r7,r12 @row count, move ht_tmp to r7 716 717PU1_SRC_LOOP_RESIDUE: 718 VMOV.I8 Q9,#0 719 ADD r8,r0,r1 @*pu1_src + src_strd 720 VLD1.8 D16,[r8]! @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 721 VLD1.8 D17,[r8] @pu1_next_row = vld1q_u8(pu1_src_cpy + src_strd) 722 SUB r8,#8 723 724 LDRB r8,[r8,#16] @pu1_src_cpy[src_strd + 16] 725 VMOV.8 d18[0],r8 @pu1_next_row_tmp = vsetq_lane_u8(pu1_src_cpy[src_strd + 16], pu1_next_row_tmp, 0) 726 VEXT.8 Q9,Q8,Q9,#1 @pu1_next_row_tmp = vextq_u8(pu1_next_row, pu1_next_row_tmp, 1) 727 728 CMP r7,r12 729 BLT SIGN_UP_CHANGE_RESIDUE 730 LDR r5,[sp,#pu1_avail_offset] @Loads pu1_avail 731 LDRB r5,[r5,#2] @pu1_avail[2] 732 CMP r5,#0 733 BNE SIGN_UP_CHANGE_DONE_RESIDUE 734 735SIGN_UP_CHANGE_RESIDUE: 736 LDRB r8,[r0] @pu1_src_cpy[0] 737 SUB r5,r12,r7 @ht_tmp - row 738 739 ADD r5,r14,r5 740 LDRB r5,[r5,#-1] @load the value 741 SUBS r8,r8,r5 @pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row] 742 MVNLT r8,#0 743 MOVGT r8,#1 @SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]) 744 VMOV.8 d14[0],r8 @sign_up = sign_up = vsetq_lane_s8(SIGN(pu1_src_cpy[0] - pu1_src_left_cpy[ht_tmp - 1 - row]), sign_up, 0) 745 746SIGN_UP_CHANGE_DONE_RESIDUE: 747 VCGT.U8 Q10,Q6,Q9 @vcgtq_u8(pu1_cur_row, pu1_next_row_tmp) 748 VCLT.U8 Q11,Q6,Q9 @vcltq_u8(pu1_cur_row, pu1_next_row_tmp) 749 VSUB.U8 Q12,Q11,Q10 @sign_down = vreinterpretq_s8_u8(vsubq_u8(cmp_lt, cmp_gt)) 750 751 VADD.I8 Q13,Q0,Q7 @edge_idx = vaddq_s8(const_2, sign_up) 752 VADD.I8 Q13,Q13,Q12 @edge_idx = vaddq_s8(edge_idx, sign_down) 753 VTBL.8 D26,{D6},D26 @vtbl1_s8(edge_idx_tbl, vget_low_s8(edge_idx)) 754 VTBL.8 D27,{D6},D27 @vtbl1_s8(edge_idx_tbl, vget_high_s8(edge_idx)) 755 756 VAND Q13,Q13,Q4 @edge_idx = vandq_s8(edge_idx, au1_mask) 757 758 VNEG.S8 Q7,Q12 @sign_up = vnegq_s8(sign_down) 759 VEXT.8 Q7,Q7,Q7,#15 @sign_up = vextq_s8(sign_up, sign_up, 15) 760 761 VTBL.8 D24,{D7},D26 @offset = vtbl1_s8(offset_tbl, vget_low_s8(edge_idx)) 762 VMOVL.U8 Q14,D12 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vmovl_u8(vget_low_u8(pu1_cur_row))) 763 VADDW.S8 Q14,Q14,D24 @pi2_tmp_cur_row.val[0] = vaddw_s8(pi2_tmp_cur_row.val[0], offset) 764 VMAX.S16 Q14,Q14,Q1 @pi2_tmp_cur_row.val[0] = vmaxq_s16(pi2_tmp_cur_row.val[0], const_min_clip) 765 VMIN.U16 Q14,Q14,Q2 @pi2_tmp_cur_row.val[0] = vreinterpretq_s16_u16(vminq_u16(vreinterpretq_u16_s16(pi2_tmp_cur_row.val[0]), const_max_clip)) 766 767 VMOVN.I16 D30,Q14 @vmovn_s16(pi2_tmp_cur_row.val[0]) 768 769 VST1.8 {D30},[r0],r1 @vst1q_u8(pu1_src_cpy, pu1_cur_row) 770 VMOV Q6,Q8 @pu1_cur_row = pu1_next_row 771 SUBS r7,r7,#1 772 BNE PU1_SRC_LOOP_RESIDUE 773 774 LDR r8,[sp,#ht_offset] @Loads ht 775 ADD r5,sp,#66 @*au1_src_left_tmp 776 777 LDR r2,[sp,#156] @Loads *pu1_src_left 778 SUB r5,r5,#1 779 780 SUB r2,r2,#1 781 782SRC_LEFT_LOOP_RESIDUE: 783 LDRB r7,[r5,#1]! @au1_src_left_tmp[row] 784 SUBS r8,r8,#1 785 STRB r7,[r2,#1]! @pu1_src_left[row] = au1_src_left_tmp[row] 786 BNE SRC_LEFT_LOOP_RESIDUE 787 788 789RE_ASSINING_LOOP: 790 LDR r8,[sp,#ht_offset] @Loads ht 791 LDR r7,[sp,#wd_offset] @Loads wd 792 793 LDR r0,[sp,#152] @Loads *pu1_src 794 SUB r8,r8,#1 @ht - 1 795 796 MLA r6,r8,r1,r7 @wd - 1 + (ht - 1) * src_strd 797 STRB r9,[r0] @pu1_src_org[0] = u1_pos_0_0_tmp 798 799 LDR r4,[sp,#pu1_src_top_left_offset] @Loads pu1_src_top_left 800 ADD r6,r0,r6 @pu1_src[wd - 1 + (ht - 1) * src_strd] 801 802 ADD r12,sp,#2 803 STRB r10,[r6,#-1] @pu1_src_org[wd - 1 + (ht - 1) * src_strd] = u1_pos_wd_ht_tmp 804 805 LDRB r11,[sp] @load u1_src_top_left_tmp from stack pointer 806 LDR r3,[sp,#148] @Loads pu1_src_top 807 808 STRB r11,[r4] @*pu1_src_top_left = u1_src_top_left_tmp 809 810SRC_TOP_LOOP: 811 VLD1.8 D0,[r12]! @pu1_src_top[col] = au1_src_top_tmp[col] 812 SUBS r7,r7,#8 @Decrement the width 813 VST1.8 D0,[r3]! @pu1_src_top[col] = au1_src_top_tmp[col] 814 BNE SRC_TOP_LOOP 815 816END_LOOPS: 817 ADD sp,sp,#160 818 vpop {d8 - d15} 819 LDMFD sp!,{r4-r12,r15} @Reload the registers from SP 820 821 822 823