virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00
Code Issues Releases Activity

Move add/diff_int16 to lossless_videodsp

Browse Source 
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2014-01-20 20:32:51 +01:00
parent 78e39aa7ee
commit f70d7eb20c
16 changed files with 244 additions and 118 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
configure vendored
View File

@ -1683,6 +1683,7 @@ CONFIG_EXTRA="
huffman
intrax8
lgplv3
llviddsp
lpc
mpegaudio
mpegaudiodsp
@ -1904,8 +1905,8 @@ eatqi_decoder_select="aandcttables error_resilience mpegvideo"
exr_decoder_select="zlib"
ffv1_decoder_select="dsputil golomb rangecoder"
ffv1_encoder_select="dsputil rangecoder"
ffvhuff_decoder_select="dsputil"
ffvhuff_encoder_select="dsputil huffman"
ffvhuff_decoder_select="dsputil llviddsp"
ffvhuff_encoder_select="dsputil huffman llviddsp"
flac_decoder_select="golomb"
flac_encoder_select="dsputil golomb lpc"
flashsv_decoder_select="zlib"
@ -1927,8 +1928,8 @@ h263p_encoder_select="h263_encoder"
h264_decoder_select="golomb h264chroma h264dsp h264pred h264qpel videodsp"
h264_decoder_suggest="error_resilience"
hevc_decoder_select="dsputil golomb videodsp"
huffyuv_decoder_select="dsputil"
huffyuv_encoder_select="dsputil huffman"
huffyuv_decoder_select="dsputil llviddsp"
huffyuv_encoder_select="dsputil huffman llviddsp"
iac_decoder_select="dsputil fft mdct sinewin"
imc_decoder_select="dsputil fft mdct sinewin"
indeo3_decoder_select="hpeldsp"

1
libavcodec/Makefile
View File

@ -57,6 +57,7 @@ OBJS-$(CONFIG_HPELDSP) += hpeldsp.o
OBJS-$(CONFIG_HUFFMAN) += huffman.o
OBJS-$(CONFIG_INTRAX8) += intrax8.o intrax8dsp.o
OBJS-$(CONFIG_LIBXVID) += libxvid_rc.o
OBJS-$(CONFIG_LLVIDDSP) += lossless_videodsp.o
OBJS-$(CONFIG_LPC) += lpc.o
OBJS-$(CONFIG_LSP) += lsp.o
OBJS-$(CONFIG_MDCT) += mdct_fixed.o mdct_float.o mdct_fixed_32.o

41
libavcodec/dsputil.c
View File

@ -1888,45 +1888,6 @@ static void diff_bytes_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
dst[i+0] = src1[i+0]-src2[i+0];
}
static void add_int16_c(uint16_t *dst, const uint16_t *src, unsigned mask, int w){
long i;
unsigned long pw_lsb = (mask >> 1) * 0x0001000100010001ULL;
unsigned long pw_msb = pw_lsb + 0x0001000100010001ULL;
for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
long a = *(long*)(src+i);
long b = *(long*)(dst+i);
*(long*)(dst+i) = ((a&pw_lsb) + (b&pw_lsb)) ^ ((a^b)&pw_msb);
}
for(; i<w; i++)
dst[i] = (dst[i] + src[i]) & mask;
}
static void diff_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w){
long i;
#if !HAVE_FAST_UNALIGNED
if((long)src2 & (sizeof(long)-1)){
for(i=0; i+7<w; i+=8){
dst[i+0] = (src1[i+0]-src2[i+0]) & mask;
dst[i+1] = (src1[i+1]-src2[i+1]) & mask;
dst[i+2] = (src1[i+2]-src2[i+2]) & mask;
dst[i+3] = (src1[i+3]-src2[i+3]) & mask;
}
}else
#endif
{
unsigned long pw_lsb = (mask >> 1) * 0x0001000100010001ULL;
unsigned long pw_msb = pw_lsb + 0x0001000100010001ULL;
for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
long a = *(long*)(src1+i);
long b = *(long*)(src2+i);
*(long*)(dst+i) = ((a|pw_msb) - (b&pw_lsb)) ^ ((a^b^pw_msb)&pw_msb);
}
}
for (; i<w; i++)
dst[i] = (src1[i] - src2[i]) & mask;
}
static void add_hfyu_median_prediction_c(uint8_t *dst, const uint8_t *src1, const uint8_t *diff, int w, int *left, int *left_top){
int i;
uint8_t l, lt;
@ -2812,8 +2773,6 @@ av_cold void ff_dsputil_init(DSPContext* c, AVCodecContext *avctx)
c->add_bytes= add_bytes_c;
c->diff_bytes= diff_bytes_c;
c->add_int16 = add_int16_c;
c->diff_int16= diff_int16_c;
c->add_hfyu_median_prediction= add_hfyu_median_prediction_c;
c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
c->add_hfyu_left_prediction = add_hfyu_left_prediction_c;

2
libavcodec/dsputil.h
View File

@ -193,8 +193,6 @@ typedef struct DSPContext {
/* huffyuv specific */
void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
void (*diff_bytes)(uint8_t *dst/*align 16*/, const uint8_t *src1/*align 16*/, const uint8_t *src2/*align 1*/,int w);
void (*add_int16)(uint16_t *dst/*align 16*/, const uint16_t *src/*align 16*/, unsigned mask, int w);
void (*diff_int16)(uint16_t *dst/*align 16*/, const uint16_t *src1/*align 16*/, const uint16_t *src2/*align 1*/, unsigned mask, int w);
/**
* subtract huffyuv's variant of median prediction
* note, this might read from src1[-1], src2[-1]

1
libavcodec/huffyuv.c
View File

@ -81,6 +81,7 @@ av_cold void ff_huffyuv_common_init(AVCodecContext *avctx)
s->flags = avctx->flags;
ff_dsputil_init(&s->dsp, avctx);
ff_llviddsp_init(&s->llviddsp);
s->width = avctx->width;
s->height = avctx->height;

2
libavcodec/huffyuv.h
View File

@ -35,6 +35,7 @@
#include "dsputil.h"
#include "get_bits.h"
#include "put_bits.h"
#include "lossless_videodsp.h"
#define VLC_BITS 11
@ -92,6 +93,7 @@ typedef struct HYuvContext {
uint8_t *bitstream_buffer;
unsigned int bitstream_buffer_size;
DSPContext dsp;
LLVidDSPContext llviddsp;
} HYuvContext;
void ff_huffyuv_common_init(AVCodecContext *s);

2
libavcodec/huffyuvdec.c
View File

@ -708,7 +708,7 @@ static void add_bytes(HYuvContext *s, uint8_t *dst, uint8_t *src, int w)
if (s->bps <= 8) {
s->dsp.add_bytes(dst, src, w);
} else {
s->dsp.add_int16((uint16_t*)dst, (const uint16_t*)src, s->n - 1, w);
s->llviddsp.add_int16((uint16_t*)dst, (const uint16_t*)src, s->n - 1, w);
}
}

2
libavcodec/huffyuvenc.c
View File

@ -41,7 +41,7 @@ static inline void diff_bytes(HYuvContext *s, uint8_t *dst,
if (s->bps <= 8) {
s->dsp.diff_bytes(dst, src0, src1, w);
} else {
s->dsp.diff_int16((uint16_t *)dst, (const uint16_t *)src0, (const uint16_t *)src1, s->n - 1, w);
s->llviddsp.diff_int16((uint16_t *)dst, (const uint16_t *)src0, (const uint16_t *)src1, s->n - 1, w);
}
}

69
libavcodec/lossless_videodsp.c Normal file
View File

@ -0,0 +1,69 @@
/*
* Lossless video DSP utils
*
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "avcodec.h"
#include "lossless_videodsp.h"
static void add_int16_c(uint16_t *dst, const uint16_t *src, unsigned mask, int w){
long i;
unsigned long pw_lsb = (mask >> 1) * 0x0001000100010001ULL;
unsigned long pw_msb = pw_lsb + 0x0001000100010001ULL;
for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
long a = *(long*)(src+i);
long b = *(long*)(dst+i);
*(long*)(dst+i) = ((a&pw_lsb) + (b&pw_lsb)) ^ ((a^b)&pw_msb);
}
for(; i<w; i++)
dst[i] = (dst[i] + src[i]) & mask;
}
static void diff_int16_c(uint16_t *dst, const uint16_t *src1, const uint16_t *src2, unsigned mask, int w){
long i;
#if !HAVE_FAST_UNALIGNED
if((long)src2 & (sizeof(long)-1)){
for(i=0; i+7<w; i+=8){
dst[i+0] = (src1[i+0]-src2[i+0]) & mask;
dst[i+1] = (src1[i+1]-src2[i+1]) & mask;
dst[i+2] = (src1[i+2]-src2[i+2]) & mask;
dst[i+3] = (src1[i+3]-src2[i+3]) & mask;
}
}else
#endif
{
unsigned long pw_lsb = (mask >> 1) * 0x0001000100010001ULL;
unsigned long pw_msb = pw_lsb + 0x0001000100010001ULL;
for (i = 0; i <= w - (int)sizeof(long)/2; i += sizeof(long)/2) {
long a = *(long*)(src1+i);
long b = *(long*)(src2+i);
*(long*)(dst+i) = ((a|pw_msb) - (b&pw_lsb)) ^ ((a^b^pw_msb)&pw_msb);
}
}
for (; i<w; i++)
dst[i] = (src1[i] - src2[i]) & mask;
}
void ff_llviddsp_init(LLVidDSPContext *c)
{
c->add_int16 = add_int16_c;
c->diff_int16= diff_int16_c;
if (ARCH_X86)
ff_llviddsp_init_x86(c);
}

36
libavcodec/lossless_videodsp.h Normal file
View File

@ -0,0 +1,36 @@
/*
* Lossless video DSP utils
*
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_LOSSLESS_VIDEODSP_H
#define AVCODEC_LOSSLESS_VIDEODSP_H
#include "avcodec.h"
#include "libavutil/cpu.h"
typedef struct LLVidDSPContext {
void (*add_int16)(uint16_t *dst/*align 16*/, const uint16_t *src/*align 16*/, unsigned mask, int w);
void (*diff_int16)(uint16_t *dst/*align 16*/, const uint16_t *src1/*align 16*/, const uint16_t *src2/*align 1*/, unsigned mask, int w);
} LLVidDSPContext;
void ff_llviddsp_init(LLVidDSPContext *llviddsp);
void ff_llviddsp_init_x86(LLVidDSPContext *llviddsp);
#endif //AVCODEC_LOSSLESS_VIDEODSP_H

2
libavcodec/x86/Makefile
View File

@ -18,6 +18,7 @@ OBJS-$(CONFIG_H264DSP) += x86/h264dsp_init.o
OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred_init.o
OBJS-$(CONFIG_H264QPEL) += x86/h264_qpel.o
OBJS-$(CONFIG_HPELDSP) += x86/hpeldsp_init.o
OBJS-$(CONFIG_LLVIDDSP) += x86/lossless_videodsp_init.o
OBJS-$(CONFIG_LPC) += x86/lpc.o
OBJS-$(CONFIG_MLP_DECODER) += x86/mlpdsp.o
OBJS-$(CONFIG_MPEGAUDIODSP) += x86/mpegaudiodsp.o
@ -86,6 +87,7 @@ YASM-OBJS-$(CONFIG_H264QPEL) += x86/h264_qpel_8bit.o \
x86/qpel.o
YASM-OBJS-$(CONFIG_HPELDSP) += x86/fpel.o \
x86/hpeldsp.o
YASM-OBJS-$(CONFIG_LLVIDDSP) += x86/lossless_videodsp.o
YASM-OBJS-$(CONFIG_MPEGAUDIODSP) += x86/imdct36.o
YASM-OBJS-$(CONFIG_PNG_DECODER) += x86/pngdsp.o
YASM-OBJS-$(CONFIG_PRORES_DECODER) += x86/proresdsp.o

64
libavcodec/x86/dsputil.asm
View File

@ -466,70 +466,6 @@ cglobal add_hfyu_left_prediction, 3,3,7, dst, src, w, left
ADD_HFYU_LEFT_LOOP 0, 0
%macro ADD_INT16_LOOP 1 ; %1 = is_aligned
movd m4, maskq
punpcklwd m4, m4
punpcklwd m4, m4
punpcklwd m4, m4
add wq, wq
test wq, 2*mmsize - 1
jz %%.tomainloop
%%.wordloop:
sub wq, 2
mov ax, [srcq+wq]
add ax, [dstq+wq]
and ax, maskw
mov [dstq+wq], ax
test wq, 2*mmsize - 1
jnz %%.wordloop
%%.tomainloop:
add srcq, wq
add dstq, wq
neg wq
jz %%.end
%%.loop:
%if %1
mova m0, [srcq+wq]
mova m1, [dstq+wq]
mova m2, [srcq+wq+mmsize]
mova m3, [dstq+wq+mmsize]
%else
movu m0, [srcq+wq]
movu m1, [dstq+wq]
movu m2, [srcq+wq+mmsize]
movu m3, [dstq+wq+mmsize]
%endif
paddw m0, m1
paddw m2, m3
pand m0, m4
pand m2, m4
%if %1
mova [dstq+wq] , m0
mova [dstq+wq+mmsize], m2
%else
movu [dstq+wq] , m0
movu [dstq+wq+mmsize], m2
%endif
add wq, 2*mmsize
jl %%.loop
%%.end:
RET
%endmacro
INIT_MMX mmx
cglobal add_int16, 4,4,5, dst, src, mask, w
ADD_INT16_LOOP 1
INIT_XMM sse2
cglobal add_int16, 4,4,5, dst, src, mask, w
test srcq, mmsize-1
jnz .unaligned
test dstq, mmsize-1
jnz .unaligned
ADD_INT16_LOOP 1
.unaligned:
ADD_INT16_LOOP 0
;-----------------------------------------------------------------------------
; void ff_vector_clip_int32(int32_t *dst, const int32_t *src, int32_t min,
; int32_t max, unsigned int len)

3
libavcodec/x86/dsputil_init.c
View File

@ -542,7 +542,6 @@ static av_cold void dsputil_init_mmx(DSPContext *c, AVCodecContext *avctx,
#endif /* HAVE_MMX_INLINE */
#if HAVE_MMX_EXTERNAL
c->add_int16 = ff_add_int16_mmx;
c->vector_clip_int32 = ff_vector_clip_int32_mmx;
#endif /* HAVE_MMX_EXTERNAL */
}
@ -626,8 +625,6 @@ static av_cold void dsputil_init_sse2(DSPContext *c, AVCodecContext *avctx,
c->vector_clip_int32 = ff_vector_clip_int32_sse2;
}
c->bswap_buf = ff_bswap32_buf_sse2;
c->add_int16 = ff_add_int16_sse2;
#endif /* HAVE_SSE2_EXTERNAL */
}

2
libavcodec/x86/dsputil_x86.h
View File

@ -116,8 +116,6 @@ void ff_clear_blocks_mmx(int16_t *blocks);
void ff_clear_blocks_sse(int16_t *blocks);
void ff_add_bytes_mmx(uint8_t *dst, uint8_t *src, int w);
void ff_add_int16_mmx(uint16_t *dst, const uint16_t *src, unsigned mask, int w);
void ff_add_int16_sse2(uint16_t *dst, const uint16_t *src, unsigned mask, int w);
void ff_add_hfyu_median_prediction_cmov(uint8_t *dst, const uint8_t *top,
const uint8_t *diff, int w,

88
libavcodec/x86/lossless_videodsp.asm Normal file
View File

@ -0,0 +1,88 @@
;******************************************************************************
;* SIMD lossless video DSP utils
;* Copyright (c) 2014 Michael Niedermayer
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg 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.
;*
;* FFmpeg 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 FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "libavutil/x86/x86util.asm"
SECTION_TEXT
%macro ADD_INT16_LOOP 1 ; %1 = is_aligned
movd m4, maskq
punpcklwd m4, m4
punpcklwd m4, m4
punpcklwd m4, m4
add wq, wq
test wq, 2*mmsize - 1
jz %%.tomainloop
%%.wordloop:
sub wq, 2
mov ax, [srcq+wq]
add ax, [dstq+wq]
and ax, maskw
mov [dstq+wq], ax
test wq, 2*mmsize - 1
jnz %%.wordloop
%%.tomainloop:
add srcq, wq
add dstq, wq
neg wq
jz %%.end
%%.loop:
%if %1
mova m0, [srcq+wq]
mova m1, [dstq+wq]
mova m2, [srcq+wq+mmsize]
mova m3, [dstq+wq+mmsize]
%else
movu m0, [srcq+wq]
movu m1, [dstq+wq]
movu m2, [srcq+wq+mmsize]
movu m3, [dstq+wq+mmsize]
%endif
paddw m0, m1
paddw m2, m3
pand m0, m4
pand m2, m4
%if %1
mova [dstq+wq] , m0
mova [dstq+wq+mmsize], m2
%else
movu [dstq+wq] , m0
movu [dstq+wq+mmsize], m2
%endif
add wq, 2*mmsize
jl %%.loop
%%.end:
RET
%endmacro
INIT_MMX mmx
cglobal add_int16, 4,4,5, dst, src, mask, w
ADD_INT16_LOOP 1
INIT_XMM sse2
cglobal add_int16, 4,4,5, dst, src, mask, w
test srcq, mmsize-1
jnz .unaligned
test dstq, mmsize-1
jnz .unaligned
ADD_INT16_LOOP 1
.unaligned:
ADD_INT16_LOOP 0

38
libavcodec/x86/lossless_videodsp_init.c Normal file
View File

@ -0,0 +1,38 @@
/*
* Lossless video DSP utils
*
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "../lossless_videodsp.h"
#include "libavutil/x86/cpu.h"
void ff_add_int16_mmx(uint16_t *dst, const uint16_t *src, unsigned mask, int w);
void ff_add_int16_sse2(uint16_t *dst, const uint16_t *src, unsigned mask, int w);
void ff_llviddsp_init_x86(LLVidDSPContext *c)
{
int cpu_flags = av_get_cpu_flags();
if (EXTERNAL_MMX(cpu_flags)) {
c->add_int16 = ff_add_int16_mmx;
}
if (EXTERNAL_SSE2(cpu_flags)) {
c->add_int16 = ff_add_int16_sse2;
}
}