From 65d5d5865845f057cc6530a8d0f34db952d9009c Mon Sep 17 00:00:00 2001 From: Diego Biurrun Date: Mon, 23 Dec 2013 19:48:43 +0100 Subject: dsputil: Move SVQ1 encoding specific bits into svq1enc --- libavcodec/ppc/Makefile | 1 + libavcodec/ppc/int_altivec.c | 44 ---------------------- libavcodec/ppc/svq1enc_altivec.c | 80 ++++++++++++++++++++++++++++++++++++++++ 3 files changed, 81 insertions(+), 44 deletions(-) create mode 100644 libavcodec/ppc/svq1enc_altivec.c (limited to 'libavcodec/ppc') diff --git a/libavcodec/ppc/Makefile b/libavcodec/ppc/Makefile index 07806661f4..ec0674c817 100644 --- a/libavcodec/ppc/Makefile +++ b/libavcodec/ppc/Makefile @@ -12,6 +12,7 @@ OBJS-$(CONFIG_MPEGVIDEO) += ppc/mpegvideo_altivec.o OBJS-$(CONFIG_VIDEODSP) += ppc/videodsp_ppc.o OBJS-$(CONFIG_VP3DSP) += ppc/vp3dsp_altivec.o +OBJS-$(CONFIG_SVQ1_ENCODER) += ppc/svq1enc_altivec.o OBJS-$(CONFIG_VC1_DECODER) += ppc/vc1dsp_altivec.o OBJS-$(CONFIG_VORBIS_DECODER) += ppc/vorbisdsp_altivec.o OBJS-$(CONFIG_VP7_DECODER) += ppc/vp8dsp_altivec.o diff --git a/libavcodec/ppc/int_altivec.c b/libavcodec/ppc/int_altivec.c index cd1984a54c..fa3cb66095 100644 --- a/libavcodec/ppc/int_altivec.c +++ b/libavcodec/ppc/int_altivec.c @@ -34,48 +34,6 @@ #include "libavcodec/dsputil.h" #include "dsputil_altivec.h" -static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2, - int size) -{ - int i, size16 = size >> 4; - vector signed char vpix1; - vector signed short vpix2, vdiff, vpix1l, vpix1h; - union { - vector signed int vscore; - int32_t score[4]; - } u = { .vscore = vec_splat_s32(0) }; - -// XXX lazy way, fix it later - - while (size16) { - // score += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]); - // load pix1 and the first batch of pix2 - - vpix1 = vec_unaligned_load(pix1); - vpix2 = vec_unaligned_load(pix2); - pix2 += 8; - // unpack - vpix1h = vec_unpackh(vpix1); - vdiff = vec_sub(vpix1h, vpix2); - vpix1l = vec_unpackl(vpix1); - // load another batch from pix2 - vpix2 = vec_unaligned_load(pix2); - u.vscore = vec_msum(vdiff, vdiff, u.vscore); - vdiff = vec_sub(vpix1l, vpix2); - u.vscore = vec_msum(vdiff, vdiff, u.vscore); - pix1 += 16; - pix2 += 8; - size16--; - } - u.vscore = vec_sums(u.vscore, vec_splat_s32(0)); - - size %= 16; - for (i = 0; i < size; i++) - u.score[3] += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]); - - return u.score[3]; -} - static int32_t scalarproduct_int16_altivec(const int16_t *v1, const int16_t *v2, int order) { @@ -140,8 +98,6 @@ static int32_t scalarproduct_and_madd_int16_altivec(int16_t *v1, av_cold void ff_int_init_altivec(DSPContext *c, AVCodecContext *avctx) { - c->ssd_int8_vs_int16 = ssd_int8_vs_int16_altivec; - c->scalarproduct_int16 = scalarproduct_int16_altivec; c->scalarproduct_and_madd_int16 = scalarproduct_and_madd_int16_altivec; diff --git a/libavcodec/ppc/svq1enc_altivec.c b/libavcodec/ppc/svq1enc_altivec.c new file mode 100644 index 0000000000..005239f430 --- /dev/null +++ b/libavcodec/ppc/svq1enc_altivec.c @@ -0,0 +1,80 @@ +/* + * Copyright (c) 2007 Luca Barbato + * + * 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" +#if HAVE_ALTIVEC_H +#include +#endif + +#include "libavutil/attributes.h" +#include "libavutil/ppc/types_altivec.h" +#include "libavutil/ppc/util_altivec.h" +#include "libavcodec/svq1enc.h" + +#if HAVE_ALTIVEC +static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2, + int size) +{ + int i, size16 = size >> 4; + vector signed char vpix1; + vector signed short vpix2, vdiff, vpix1l, vpix1h; + union { + vector signed int vscore; + int32_t score[4]; + } u = { .vscore = vec_splat_s32(0) }; + + while (size16) { + // score += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]); + // load pix1 and the first batch of pix2 + + vpix1 = vec_unaligned_load(pix1); + vpix2 = vec_unaligned_load(pix2); + pix2 += 8; + // unpack + vpix1h = vec_unpackh(vpix1); + vdiff = vec_sub(vpix1h, vpix2); + vpix1l = vec_unpackl(vpix1); + // load another batch from pix2 + vpix2 = vec_unaligned_load(pix2); + u.vscore = vec_msum(vdiff, vdiff, u.vscore); + vdiff = vec_sub(vpix1l, vpix2); + u.vscore = vec_msum(vdiff, vdiff, u.vscore); + pix1 += 16; + pix2 += 8; + size16--; + } + u.vscore = vec_sums(u.vscore, vec_splat_s32(0)); + + size %= 16; + for (i = 0; i < size; i++) + u.score[3] += (pix1[i] - pix2[i]) * (pix1[i] - pix2[i]); + + return u.score[3]; +} +#endif /* HAVE_ALTIVEC */ + +av_cold void ff_svq1enc_init_ppc(SVQ1EncContext *c) +{ +#if HAVE_ALTIVEC + c->ssd_int8_vs_int16 = ssd_int8_vs_int16_altivec; +#endif /* HAVE_ALTIVEC */ +} -- cgit v1.2.3