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_horz_neon.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_luma_horz() 32 @* 33 @* @remarks 34 @* none 35 @* 36 @******************************************************************************* 37 @*/ 38 @ 39 @/** 40 @******************************************************************************* 41 @* 42 @* @brief 43 @* intra prediction interpolation filter for horizontal luma variable. 44 @* 45 @* @par description: 46 @* horizontal intraprediction(mode 10) with.extern samples location 47 @* pointed by 'pu1_ref' to the tu block location pointed by 'pu1_dst' refer 48 @* to section 8.4.4.2.6 in the standard (special case) 49 @* 50 @* @param[in] pu1_src 51 @* uword8 pointer to the source 52 @* 53 @* @param[out] 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 @void ihevc_intra_pred_chroma_horz(uword8 *pu1_ref, 76 @ word32 src_strd, 77 @ uword8 *pu1_dst, 78 @ word32 dst_strd, 79 @ word32 nt, 80 @ word32 mode) 81 @**************variables vs registers***************************************** 82 @r0 => *pu1_ref 83 @r1 => src_strd 84 @r2 => *pu1_dst 85 @r3 => dst_strd 86 87 .equ nt_offset, 104 88 89 .text 90 .align 4 91 92 93 94 95 .globl ihevc_intra_pred_chroma_horz_a9q 96 97 .type ihevc_intra_pred_chroma_horz_a9q, %function 98 99 ihevc_intra_pred_chroma_horz_a9q: 100 101 stmfd sp!, {r4-r12, r14} @stack stores the values of the arguments 102 vpush {d8 - d15} 103 104 ldr r4,[sp,#nt_offset] @loads nt 105 106 lsl r6,r4,#2 @four_nt 107 108 add r12,r0,r6 @*pu1_ref[four_nt] 109 cmp r4,#4 @if nt == 4 110 beq core_loop_4 111 112 cmp r4,#8 @if nt == 8 113 beq core_loop_8 114 115 @cmp r4,#16 @if nt == 16 116 @beq core_loop_16 117 118 sub r12,r12,#16 @move to 16th value pointer 119 add r9,r2,#16 120 121 core_loop_16: 122 vld1.16 {q0},[r12] @load 16 values. d1[7] will have the 1st value. 123 sub r12,r12,#16 124 vld1.16 {q5},[r12] @load 16 values. d1[7] will have the 1st value. 125 126 vdup.16 q1,d1[3] @duplicate the i value. 127 128 vdup.16 q2,d1[2] @duplicate the ii value. 129 vdup.16 q3,d1[1] @duplicate the iii value. 130 vst1.16 {q1},[r2],r3 @store in 1st row 0-16 columns 131 vst1.16 {q1},[r9],r3 @store in 1st row 16-32 columns 132 133 vdup.16 q4,d1[0] 134 vst1.16 {q2},[r2],r3 135 vst1.16 {q2},[r9],r3 136 137 vdup.16 q1,d0[3] 138 vst1.16 {q3},[r2],r3 139 vst1.16 {q3},[r9],r3 140 141 vdup.16 q2,d0[2] 142 vst1.16 {q4},[r2],r3 143 vst1.16 {q4},[r9],r3 144 145 vdup.16 q3,d0[1] 146 vst1.16 {q1},[r2],r3 147 vst1.16 {q1},[r9],r3 148 149 vdup.16 q4,d0[0] 150 vst1.16 {q2},[r2],r3 151 vst1.16 {q2},[r9],r3 152 153 vdup.16 q1,d11[3] 154 vst1.16 {q3},[r2],r3 155 vst1.16 {q3},[r9],r3 156 157 vdup.16 q2,d11[2] 158 vst1.16 {q4},[r2],r3 159 vst1.16 {q4},[r9],r3 160 161 vdup.16 q3,d11[1] 162 vst1.16 {q1},[r2],r3 163 vst1.16 {q1},[r9],r3 164 165 vdup.16 q4,d11[0] 166 vst1.16 {q2},[r2],r3 167 vst1.16 {q2},[r9],r3 168 169 vdup.16 q1,d10[3] 170 vst1.16 {q3},[r2],r3 171 vst1.16 {q3},[r9],r3 172 173 vdup.16 q2,d10[2] 174 vst1.16 {q4},[r2],r3 175 vst1.16 {q4},[r9],r3 176 177 vdup.16 q3,d10[1] 178 vst1.16 {q1},[r2],r3 179 vst1.16 {q1},[r9],r3 180 sub r12,r12,#16 @move to 16th value pointer 181 182 vdup.16 q4,d10[0] 183 vst1.16 {q2},[r2],r3 184 vst1.16 {q2},[r9],r3 185 186 subs r4,r4,#16 @decrement the loop count by 16 187 vst1.16 {q3},[r2],r3 188 vst1.16 {q3},[r9],r3 189 190 vst1.16 {q4},[r2],r3 191 vst1.16 {q4},[r9],r3 192 bgt core_loop_16 193 vpop {d8 - d15} 194 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 195 b endloop 196 197 core_loop_8: 198 ldrb lr,[r12],#1 @pu1_ref[two_nt] 199 @vld1.8 {q15},[r12] @pu1_ref[two_nt + 1 + col] 200 201 vdup.8 d28,lr 202 sub r12,r12,#17 203 vld1.8 {q0},[r12] 204 205 sub r12,r12,#16 206 vld1.8 {q15},[r12] 207 vdup.16 q5,d1[3] 208 @vmovl.u8 q13,d26 209 210 vdup.16 q1,d1[2] 211 @vsubl.u8 q12,d30,d28 212 213 vdup.16 q2,d1[1] 214 @vshr.s16 q12,q12,#1 215 216 vdup.16 q3,d1[0] 217 @vqadd.s16 q11,q13,q12 218 219 vdup.16 q4,d0[3] 220 @vqmovun.s16 d22,q11 221 222 vst1.16 {q5},[r2],r3 223 224 vdup.16 q5,d0[2] 225 @vsubl.u8 q12,d31,d28 226 227 vdup.16 q6,d0[1] 228 @vshr.s16 q12,q12,#1 229 230 vdup.16 q7,d0[0] 231 @vqadd.s16 q11,q13,q12 232 233 vdup.16 q8,d0[3] 234 @vqmovun.s16 d22,q11 235 236 vst1.16 {q1},[r2],r3 237 @sub r2,r2,#8 238 239 vst1.16 {q2},[r2],r3 240 241 vst1.16 {q3},[r2],r3 242 vst1.16 {q4},[r2],r3 243 vst1.16 {q5},[r2],r3 244 245 @vdup.8 q1,d0[2] 246 vst1.16 {q6},[r2],r3 247 248 @vdup.8 q2,d0[1] 249 vst1.16 {q7},[r2],r3 250 251 @vdup.8 q3,d0[0] 252 @vst1.8 {q7},[r2],r3 253 254 @vdup.8 q4,d0[3] 255 @vst1.8 {q8},[r2],r3 256 257 @vdup.8 q5,d0[2] 258 @vst1.8 {q1},[r2],r3 259 260 @vdup.8 q6,d0[1] 261 @vst1.8 {q2},[r2],r3 262 263 @vdup.8 q7,d0[0] 264 @vst1.8 {q3},[r2],r3 265 266 @vst1.8 {q4},[r2],r3 267 @vst1.8 {q5},[r2],r3 268 @vst1.8 {q6},[r2],r3 269 @vst1.8 {q7},[r2],r3 270 vpop {d8 - d15} 271 272 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 273 b endloop 274 275 276 core_loop_4: 277 ldrb lr,[r12] @pu1_ref[two_nt] 278 add r12,r12,#1 @pu1_ref[two_nt + 1] 279 @vld1.8 {d30},[r12] @pu1_ref[two_nt + 1 + col] 280 281 sub r12,r12,#9 282 vld1.8 {d0},[r12] 283 sub r12,r12,#8 284 vld1.8 {d30},[r12] 285 vdup.16 d26,d0[3] 286 vdup.8 d28,lr 287 288 vdup.16 d3,d0[2] 289 vmovl.u8 q13,d26 290 291 vdup.16 d4,d0[1] 292 vsubl.u8 q12,d30,d28 293 294 vdup.16 d5,d0[0] 295 vshr.s16 q12,q12,#1 296 297 vdup.16 d6,d0[3] 298 vqadd.s16 q11,q13,q12 299 300 vdup.16 d7,d0[2] 301 vqmovun.s16 d22,q11 302 303 vst1.8 {d6},[r2],r3 304 vst1.8 {d3},[r2],r3 305 306 vdup.16 d8,d0[1] 307 vst1.8 {d4},[r2],r3 308 vst1.8 {d5},[r2],r3 309 310 vdup.16 d9,d0[0] 311 @vst1.8 {d6},[r2],r3 312 @vst1.8 {d7},[r2],r3 313 314 @vst1.8 {d8},[r2],r3 315 @vst1.8 {d9},[r2],r3 316 vpop {d8 - d15} 317 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 318 b endloop 319 320 321 @core_loop_4 322 ldrb lr,[r12] @pu1_ref[two_nt] 323 add r12,r12,#1 @pu1_ref[two_nt + 1] 324 vld1.8 {d30},[r12] @pu1_ref[two_nt + 1 + col] 325 326 sub r12,r12,#5 327 vld1.8 {d0},[r12] 328 vdup.8 d28,lr 329 vdup.8 d26,d0[3] 330 vmovl.u8 q13,d26 331 332 vdup.8 d3,d0[2] 333 vsubl.u8 q12,d30,d28 334 335 vdup.8 d4,d0[1] 336 vshr.s16 q12,q12,#1 337 338 vdup.8 d5,d0[0] 339 vqadd.s16 q11,q13,q12 340 341 vqmovun.s16 d22,q11 342 343 vst1.32 {d22[0]},[r2],r3 344 vst1.32 {d3[0]},[r2],r3 345 vst1.32 {d4[0]},[r2],r3 346 vst1.32 {d5[0]},[r2],r3 347 348 vpop {d8 - d15} 349 ldmfd sp!,{r4-r12,r15} @reload the registers from sp 350 351 endloop: 352 353 354
