1cabdff1aSopenharmony_ci/* 2cabdff1aSopenharmony_ci * Copyright (c) 2016 Clément Bœsch <clement stupeflix.com> 3cabdff1aSopenharmony_ci * Copyright (c) 2016 Matthieu Bouron <matthieu.bouron stupeflix.com> 4cabdff1aSopenharmony_ci * 5cabdff1aSopenharmony_ci * This file is part of FFmpeg. 6cabdff1aSopenharmony_ci * 7cabdff1aSopenharmony_ci * FFmpeg is free software; you can redistribute it and/or 8cabdff1aSopenharmony_ci * modify it under the terms of the GNU Lesser General Public 9cabdff1aSopenharmony_ci * License as published by the Free Software Foundation; either 10cabdff1aSopenharmony_ci * version 2.1 of the License, or (at your option) any later version. 11cabdff1aSopenharmony_ci * 12cabdff1aSopenharmony_ci * FFmpeg is distributed in the hope that it will be useful, 13cabdff1aSopenharmony_ci * but WITHOUT ANY WARRANTY; without even the implied warranty of 14cabdff1aSopenharmony_ci * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 15cabdff1aSopenharmony_ci * Lesser General Public License for more details. 16cabdff1aSopenharmony_ci * 17cabdff1aSopenharmony_ci * You should have received a copy of the GNU Lesser General Public 18cabdff1aSopenharmony_ci * License along with FFmpeg; if not, write to the Free Software 19cabdff1aSopenharmony_ci * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 20cabdff1aSopenharmony_ci */ 21cabdff1aSopenharmony_ci 22cabdff1aSopenharmony_ci#include "libavutil/arm/asm.S" 23cabdff1aSopenharmony_ci 24cabdff1aSopenharmony_cifunction ff_hscale_8_to_15_neon, export=1 25cabdff1aSopenharmony_ci push {r4-r12, lr} 26cabdff1aSopenharmony_ci vpush {q4-q7} 27cabdff1aSopenharmony_ci ldr r4, [sp, #104] @ filter 28cabdff1aSopenharmony_ci ldr r5, [sp, #108] @ filterPos 29cabdff1aSopenharmony_ci ldr r6, [sp, #112] @ filterSize 30cabdff1aSopenharmony_ci add r10, r4, r6, lsl #1 @ filter2 = filter + filterSize * 2 31cabdff1aSopenharmony_ci1: ldr r8, [r5], #4 @ filterPos[0] 32cabdff1aSopenharmony_ci ldr r9, [r5], #4 @ filterPos[1] 33cabdff1aSopenharmony_ci vmov.s32 q4, #0 @ val accumulator 34cabdff1aSopenharmony_ci vmov.s32 q5, #0 @ val accumulator 35cabdff1aSopenharmony_ci mov r7, r6 @ tmpfilterSize = filterSize 36cabdff1aSopenharmony_ci mov r0, r3 @ srcp 37cabdff1aSopenharmony_ci2: add r11, r0, r8 @ srcp + filterPos[0] 38cabdff1aSopenharmony_ci add r12, r0, r9 @ srcp + filterPos[1] 39cabdff1aSopenharmony_ci vld1.8 d0, [r11] @ srcp[filterPos[0] + {0..7}] 40cabdff1aSopenharmony_ci vld1.8 d2, [r12] @ srcp[filterPos[1] + {0..7}] 41cabdff1aSopenharmony_ci vld1.16 {q2}, [r4]! @ load 8x16-bit filter values 42cabdff1aSopenharmony_ci vld1.16 {q3}, [r10]! @ load 8x16-bit filter values 43cabdff1aSopenharmony_ci vmovl.u8 q0, d0 @ unpack src values to 16-bit 44cabdff1aSopenharmony_ci vmovl.u8 q1, d2 @ unpack src values to 16-bit 45cabdff1aSopenharmony_ci vmull.s16 q8, d0, d4 @ srcp[filterPos[0] + {0..7}] * filter[{0..7}] (part 1) 46cabdff1aSopenharmony_ci vmull.s16 q9, d1, d5 @ srcp[filterPos[0] + {0..7}] * filter[{0..7}] (part 2) 47cabdff1aSopenharmony_ci vmull.s16 q10, d2, d6 @ srcp[filterPos[1] + {0..7}] * filter[{0..7}] (part 1) 48cabdff1aSopenharmony_ci vmull.s16 q11, d3, d7 @ srcp[filterPos[1] + {0..7}] * filter[{0..7}] (part 2) 49cabdff1aSopenharmony_ci vpadd.s32 d16, d16, d17 @ horizontal pair adding of the 8x32-bit multiplied values into 4x32-bit (part 1) 50cabdff1aSopenharmony_ci vpadd.s32 d17, d18, d19 @ horizontal pair adding of the 8x32-bit multiplied values into 4x32-bit (part 2) 51cabdff1aSopenharmony_ci vpadd.s32 d20, d20, d21 @ horizontal pair adding of the 8x32-bit multiplied values into 4x32-bit (part 1) 52cabdff1aSopenharmony_ci vpadd.s32 d21, d22, d23 @ horizontal pair adding of the 8x32-bit multiplied values into 4x32-bit (part 2) 53cabdff1aSopenharmony_ci vadd.s32 q4, q8 @ update val accumulator 54cabdff1aSopenharmony_ci vadd.s32 q5, q10 @ update val accumulator 55cabdff1aSopenharmony_ci add r0, #8 @ srcp += 8 56cabdff1aSopenharmony_ci subs r7, #8 @ tmpfilterSize -= 8 57cabdff1aSopenharmony_ci bgt 2b @ loop until tmpfilterSize is consumed 58cabdff1aSopenharmony_ci mov r4, r10 @ filter = filter2 59cabdff1aSopenharmony_ci add r10, r10, r6, lsl #1 @ filter2 += filterSize * 2 60cabdff1aSopenharmony_ci vpadd.s32 d8, d8, d9 @ horizontal pair adding of the 8x32-bit sums into 4x32-bit (part 1) 61cabdff1aSopenharmony_ci vpadd.s32 d9, d10, d11 @ horizontal pair adding of the 8x32-bit sums into 4x32-bit (part 2) 62cabdff1aSopenharmony_ci vpadd.s32 d8, d8, d9 @ horizontal pair adding of the 4x32-bit sums into 2x32-bit 63cabdff1aSopenharmony_ci vqshrn.s32 d8, q4, #7 @ shift and clip the 2x16-bit final values 64cabdff1aSopenharmony_ci vst1.32 {d8[0]},[r1]! @ write destination 65cabdff1aSopenharmony_ci subs r2, #2 @ dstW -= 2 66cabdff1aSopenharmony_ci bgt 1b @ loop until end of line 67cabdff1aSopenharmony_ci vpop {q4-q7} 68cabdff1aSopenharmony_ci pop {r4-r12, lr} 69cabdff1aSopenharmony_ci mov pc, lr 70cabdff1aSopenharmony_ciendfunc 71