From 16e6652286213014c28a9c3af70435e767d28638 Mon Sep 17 00:00:00 2001 From: James Yu Date: Sat, 17 May 2014 12:21:10 +0800 Subject: [PATCH] aarch64: NEON intrinsics dct_unquantize_h263. Intrinsics only used on aarch64 since the existing ARMv7 NEON asm is slightly faster (Cortex-A9, gcc-4.8, micro-benchmarks and full decoding time). Signed-off-by: James Yu Signed-off-by: Janne Grunau --- libavcodec/mpegvideo.c | 3 + libavcodec/mpegvideo.h | 1 + libavcodec/neon/Makefile | 1 + libavcodec/neon/mpegvideo.c | 134 ++++++++++++++++++++++++++++++++++++ 4 files changed, 139 insertions(+) create mode 100644 libavcodec/neon/Makefile create mode 100644 libavcodec/neon/mpegvideo.c diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c index e1daa2aee0..da425414b6 100644 --- a/libavcodec/mpegvideo.c +++ b/libavcodec/mpegvideo.c @@ -393,6 +393,9 @@ av_cold int ff_dct_common_init(MpegEncContext *s) s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_bitexact; s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_c; + if (HAVE_INTRINSICS_NEON) + ff_MPV_common_init_neon(s); + if (ARCH_ARM) ff_MPV_common_init_arm(s); if (ARCH_PPC) diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h index 33465091a5..3bd87dfb09 100644 --- a/libavcodec/mpegvideo.h +++ b/libavcodec/mpegvideo.h @@ -720,6 +720,7 @@ int ff_MPV_encode_picture(AVCodecContext *avctx, AVPacket *pkt, void ff_MPV_encode_init_x86(MpegEncContext *s); void ff_MPV_common_init_x86(MpegEncContext *s); void ff_MPV_common_init_arm(MpegEncContext *s); +void ff_MPV_common_init_neon(MpegEncContext *s); void ff_MPV_common_init_ppc(MpegEncContext *s); void ff_clean_intra_table_entries(MpegEncContext *s); void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h); diff --git a/libavcodec/neon/Makefile b/libavcodec/neon/Makefile new file mode 100644 index 0000000000..607f116a77 --- /dev/null +++ b/libavcodec/neon/Makefile @@ -0,0 +1 @@ +OBJS-$(CONFIG_MPEGVIDEO) += neon/mpegvideo.o diff --git a/libavcodec/neon/mpegvideo.c b/libavcodec/neon/mpegvideo.c new file mode 100644 index 0000000000..cb9bd66ed0 --- /dev/null +++ b/libavcodec/neon/mpegvideo.c @@ -0,0 +1,134 @@ +/* + * Copyright (c) 2010 Mans Rullgard + * Copyright (c) 2014 James Yu + * + * This file is part of Libav. + * + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * Libav is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include + +#include "config.h" + +#include "libavutil/cpu.h" +#if ARCH_AARCH64 +# include "libavutil/aarch64/cpu.h" +#elif ARCH_ARM +# include "libavutil/arm/cpu.h" +#endif + +#include "libavcodec/mpegvideo.h" + +static void inline ff_dct_unquantize_h263_neon(int qscale, int qadd, int nCoeffs, + int16_t *block) +{ + int16x8_t q0s16, q2s16, q3s16, q8s16, q10s16, q11s16, q13s16; + int16x8_t q14s16, q15s16, qzs16; + int16x4_t d0s16, d2s16, d3s16, dzs16; + uint16x8_t q1u16, q9u16; + uint16x4_t d1u16; + + dzs16 = vdup_n_s16(0); + qzs16 = vdupq_n_s16(0); + + q15s16 = vdupq_n_s16(qscale << 1); + q14s16 = vdupq_n_s16(qadd); + q13s16 = vnegq_s16(q14s16); + + if (nCoeffs > 4) { + for (; nCoeffs > 8; nCoeffs -= 16, block += 16) { + q0s16 = vld1q_s16(block); + q3s16 = vreinterpretq_s16_u16(vcltq_s16(q0s16, qzs16)); + q8s16 = vld1q_s16(block + 8); + q1u16 = vceqq_s16(q0s16, qzs16); + q2s16 = vmulq_s16(q0s16, q15s16); + q11s16 = vreinterpretq_s16_u16(vcltq_s16(q8s16, qzs16)); + q10s16 = vmulq_s16(q8s16, q15s16); + q3s16 = vbslq_s16(vreinterpretq_u16_s16(q3s16), q13s16, q14s16); + q11s16 = vbslq_s16(vreinterpretq_u16_s16(q11s16), q13s16, q14s16); + q2s16 = vaddq_s16(q2s16, q3s16); + q9u16 = vceqq_s16(q8s16, qzs16); + q10s16 = vaddq_s16(q10s16, q11s16); + q0s16 = vbslq_s16(q1u16, q0s16, q2s16); + q8s16 = vbslq_s16(q9u16, q8s16, q10s16); + vst1q_s16(block, q0s16); + vst1q_s16(block + 8, q8s16); + } + } + if (nCoeffs <= 0) + return; + + d0s16 = vld1_s16(block); + d3s16 = vreinterpret_s16_u16(vclt_s16(d0s16, dzs16)); + d1u16 = vceq_s16(d0s16, dzs16); + d2s16 = vmul_s16(d0s16, vget_high_s16(q15s16)); + d3s16 = vbsl_s16(vreinterpret_u16_s16(d3s16), + vget_high_s16(q13s16), vget_high_s16(q14s16)); + d2s16 = vadd_s16(d2s16, d3s16); + d0s16 = vbsl_s16(d1u16, d0s16, d2s16); + vst1_s16(block, d0s16); +} + +static void dct_unquantize_h263_inter_neon(MpegEncContext *s, int16_t *block, + int n, int qscale) +{ + int nCoeffs = s->inter_scantable.raster_end[s->block_last_index[n]]; + int qadd = (qscale - 1) | 1; + + ff_dct_unquantize_h263_neon(qscale, qadd, nCoeffs + 1, block); +} + +static void dct_unquantize_h263_intra_neon(MpegEncContext *s, int16_t *block, + int n, int qscale) +{ + int qadd; + int nCoeffs, blk0; + + if (!s->h263_aic) { + if (n < 4) + block[0] *= s->y_dc_scale; + else + block[0] *= s->c_dc_scale; + qadd = (qscale - 1) | 1; + } else { + qadd = 0; + } + + if (s->ac_pred) { + nCoeffs = 63; + } else { + nCoeffs = s->inter_scantable.raster_end[s->block_last_index[n]]; + if (nCoeffs <= 0) + return; + } + + blk0 = block[0]; + + ff_dct_unquantize_h263_neon(qscale, qadd, nCoeffs + 1, block); + + block[0] = blk0; +} + + +av_cold void ff_MPV_common_init_neon(MpegEncContext *s) +{ + int cpu_flags = av_get_cpu_flags(); + + if (have_neon(cpu_flags)) { + s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_neon; + s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_neon; + } +}