| 1 | /* |
| 2 | * Copyright (c) 2010 Mans Rullgard |
| 3 | * Copyright (c) 2014 James Yu <james.yu@linaro.org> |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | #include <arm_neon.h> |
| 23 | |
| 24 | #include "config.h" |
| 25 | |
| 26 | #include "libavutil/cpu.h" |
| 27 | #if ARCH_AARCH64 |
| 28 | # include "libavutil/aarch64/cpu.h" |
| 29 | #elif ARCH_ARM |
| 30 | # include "libavutil/arm/cpu.h" |
| 31 | #endif |
| 32 | |
| 33 | #include "libavcodec/mpegvideo.h" |
| 34 | |
| 35 | static void inline ff_dct_unquantize_h263_neon(int qscale, int qadd, int nCoeffs, |
| 36 | int16_t *block) |
| 37 | { |
| 38 | int16x8_t q0s16, q2s16, q3s16, q8s16, q10s16, q11s16, q13s16; |
| 39 | int16x8_t q14s16, q15s16, qzs16; |
| 40 | int16x4_t d0s16, d2s16, d3s16, dzs16; |
| 41 | uint16x8_t q1u16, q9u16; |
| 42 | uint16x4_t d1u16; |
| 43 | |
| 44 | dzs16 = vdup_n_s16(0); |
| 45 | qzs16 = vdupq_n_s16(0); |
| 46 | |
| 47 | q15s16 = vdupq_n_s16(qscale << 1); |
| 48 | q14s16 = vdupq_n_s16(qadd); |
| 49 | q13s16 = vnegq_s16(q14s16); |
| 50 | |
| 51 | if (nCoeffs > 4) { |
| 52 | for (; nCoeffs > 8; nCoeffs -= 16, block += 16) { |
| 53 | q0s16 = vld1q_s16(block); |
| 54 | q3s16 = vreinterpretq_s16_u16(vcltq_s16(q0s16, qzs16)); |
| 55 | q8s16 = vld1q_s16(block + 8); |
| 56 | q1u16 = vceqq_s16(q0s16, qzs16); |
| 57 | q2s16 = vmulq_s16(q0s16, q15s16); |
| 58 | q11s16 = vreinterpretq_s16_u16(vcltq_s16(q8s16, qzs16)); |
| 59 | q10s16 = vmulq_s16(q8s16, q15s16); |
| 60 | q3s16 = vbslq_s16(vreinterpretq_u16_s16(q3s16), q13s16, q14s16); |
| 61 | q11s16 = vbslq_s16(vreinterpretq_u16_s16(q11s16), q13s16, q14s16); |
| 62 | q2s16 = vaddq_s16(q2s16, q3s16); |
| 63 | q9u16 = vceqq_s16(q8s16, qzs16); |
| 64 | q10s16 = vaddq_s16(q10s16, q11s16); |
| 65 | q0s16 = vbslq_s16(q1u16, q0s16, q2s16); |
| 66 | q8s16 = vbslq_s16(q9u16, q8s16, q10s16); |
| 67 | vst1q_s16(block, q0s16); |
| 68 | vst1q_s16(block + 8, q8s16); |
| 69 | } |
| 70 | } |
| 71 | if (nCoeffs <= 0) |
| 72 | return; |
| 73 | |
| 74 | d0s16 = vld1_s16(block); |
| 75 | d3s16 = vreinterpret_s16_u16(vclt_s16(d0s16, dzs16)); |
| 76 | d1u16 = vceq_s16(d0s16, dzs16); |
| 77 | d2s16 = vmul_s16(d0s16, vget_high_s16(q15s16)); |
| 78 | d3s16 = vbsl_s16(vreinterpret_u16_s16(d3s16), |
| 79 | vget_high_s16(q13s16), vget_high_s16(q14s16)); |
| 80 | d2s16 = vadd_s16(d2s16, d3s16); |
| 81 | d0s16 = vbsl_s16(d1u16, d0s16, d2s16); |
| 82 | vst1_s16(block, d0s16); |
| 83 | } |
| 84 | |
| 85 | static void dct_unquantize_h263_inter_neon(MpegEncContext *s, int16_t *block, |
| 86 | int n, int qscale) |
| 87 | { |
| 88 | int nCoeffs = s->inter_scantable.raster_end[s->block_last_index[n]]; |
| 89 | int qadd = (qscale - 1) | 1; |
| 90 | |
| 91 | ff_dct_unquantize_h263_neon(qscale, qadd, nCoeffs + 1, block); |
| 92 | } |
| 93 | |
| 94 | static void dct_unquantize_h263_intra_neon(MpegEncContext *s, int16_t *block, |
| 95 | int n, int qscale) |
| 96 | { |
| 97 | int qadd; |
| 98 | int nCoeffs, blk0; |
| 99 | |
| 100 | if (!s->h263_aic) { |
| 101 | if (n < 4) |
| 102 | block[0] *= s->y_dc_scale; |
| 103 | else |
| 104 | block[0] *= s->c_dc_scale; |
| 105 | qadd = (qscale - 1) | 1; |
| 106 | } else { |
| 107 | qadd = 0; |
| 108 | } |
| 109 | |
| 110 | if (s->ac_pred) { |
| 111 | nCoeffs = 63; |
| 112 | } else { |
| 113 | nCoeffs = s->inter_scantable.raster_end[s->block_last_index[n]]; |
| 114 | if (nCoeffs <= 0) |
| 115 | return; |
| 116 | } |
| 117 | |
| 118 | blk0 = block[0]; |
| 119 | |
| 120 | ff_dct_unquantize_h263_neon(qscale, qadd, nCoeffs + 1, block); |
| 121 | |
| 122 | block[0] = blk0; |
| 123 | } |
| 124 | |
| 125 | |
| 126 | av_cold void ff_mpv_common_init_neon(MpegEncContext *s) |
| 127 | { |
| 128 | int cpu_flags = av_get_cpu_flags(); |
| 129 | |
| 130 | if (have_neon(cpu_flags)) { |
| 131 | s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_neon; |
| 132 | s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_neon; |
| 133 | } |
| 134 | } |