scottk/UltraGrid
1
0
Fork 0
mirror of https://github.com/outbackdingo/UltraGrid.git synced 2026-03-21 17:40:23 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
234348ebeec2a3f32148a1ddfe71aefa54ef6bd1
UltraGrid/src/video_codec.c

1724 lines
60 KiB
C
Raw Blame History

/**
* @file video_codec.c
* @author Martin Benes <martinbenesh@gmail.com>
* @author Lukas Hejtmanek <xhejtman@ics.muni.cz>
* @author Petr Holub <hopet@ics.muni.cz>
* @author Milos Liska <xliska@fi.muni.cz>
* @author Jiri Matela <matela@ics.muni.cz>
* @author Dalibor Matura <255899@mail.muni.cz>
* @author Ian Wesley-Smith <iwsmith@cct.lsu.edu>
*
* @brief This file contains video codec-related functions.
*
* This file contains video codecs' metadata and helper
* functions as well as pixelformat converting functions.
*/
/* Copyright (c) 2005-2013 CESNET z.s.p.o.
*
* Redistribution and use in source and binary forms, with or without
* modification, is permitted provided that the following conditions
* are met:
*
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
*
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
*
* This product includes software developed by CESNET z.s.p.o.
*
* 4. Neither the name of the CESNET nor the names of its contributors may be
* used to endorse or promote products derived from this software without
* specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING,
* BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
* AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
* EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
* INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
* SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#include "config_unix.h"
#include "config_win32.h"
#endif // HAVE_CONFIG_H
#include "debug.h"
#include <stdio.h>
#include <string.h>
#include "video_codec.h"
#include "utils/misc.h" // to_fourcc
#ifdef __SSSE3__
#include "tmmintrin.h"
// compat with older Clang compiler
#ifndef _mm_bslli_si128
#define _mm_bslli_si128 _mm_slli_si128
#endif
#ifndef _mm_bsrli_si128
#define _mm_bsrli_si128 _mm_srli_si128
#endif
#endif
#ifdef __cplusplus
#include <algorithm>
using std::max;
using std::min;
#else
#undef max
#undef min
#define max(a, b) (((a) > (b))? (a): (b))
#define min(a, b) (((a) < (b))? (a): (b))
#endif
#include "hwaccel_vdpau.h"
#ifdef __SSE2__
static void vc_deinterlace_aligned(unsigned char *src, long src_linesize, int lines);
static void vc_deinterlace_unaligned(unsigned char *src, long src_linesize, int lines);
#endif
static void vc_copylineToUYVY709(unsigned char *dst, const unsigned char *src, int dst_len,
int rshift, int gshift, int bshift, int pix_size) __attribute__((unused));
static void vc_copylineToUYVY601(unsigned char *dst, const unsigned char *src, int dst_len,
int rshift, int gshift, int bshift, int pix_size) __attribute__((unused));
/**
* Defines codec metadata
* @note
* Members that are not relevant for specified codec (eg. bpp, rgb for opaque
* and interframe for not opaque) should be zero.
*/
struct codec_info_t {
const char *name; ///< displayed name
const char *name_long; ///< more descriptive name
uint32_t fcc; ///< FourCC
int h_align; ///< Number of pixels each line is aligned to
double bpp; ///< Number of bytes per pixel
///< @note
///< Should be nonzero even for compressed codecs
///< when display gets requestes compressed codec
///< (otherwise division by zero occurs).
int bits_per_channel; ///< Number of bits per color channel
int block_size; ///< Bytes per pixel block (pixelformats only)
unsigned rgb:1; ///< Whether pixelformat is RGB
unsigned opaque:1; ///< If codec is opaque (= compressed)
unsigned interframe:1; ///< Indicates if compression is interframe
unsigned const_size:1; ///< Indicates if data length is constant for all resolutions (hw surfaces)
const char *file_extension; ///< Extension that should be added to name if frame is saved to file.
void (*free_extra_data)(void *); ///< Function to free extra data (used to free hw surfaces)
void *(*copy_data)(void *,
const void *,
size_t); ///< Function to copy data
};
static const struct codec_info_t codec_info[] = {
[VIDEO_CODEC_NONE] = {"(none)", "Undefined Codec",
0, 0, 0.0, 0, 0, FALSE, FALSE, FALSE, FALSE, NULL, NULL, NULL},
[RGBA] = {"RGBA", "Red Green Blue Alpha 32bit",
to_fourcc('R','G','B','A'), 1, 4.0, 8, 4, TRUE, FALSE, FALSE, FALSE, "rgba", NULL, NULL},
[UYVY] = {"UYVY", "YUV 4:2:2",
to_fourcc('U','Y','V','Y'), 2, 2, 8, 4, FALSE, FALSE, FALSE, FALSE, "yuv", NULL, NULL},
[YUYV] = {"YUYV", "YUV 4:2:2",
to_fourcc('Y','U','Y','V'), 2, 2, 8, 4, FALSE, FALSE, FALSE, FALSE, "yuv", NULL, NULL},
[R10k] = {"R10k", "RGB 4:4:4",
to_fourcc('R','1','0','k'), 64, 4, 10, 4, TRUE, FALSE, FALSE, FALSE, "r10k", NULL, NULL},
[v210] = {"v210", "YUV 4:2:2",
to_fourcc('v','2','1','0'), 48, 8.0 / 3.0, 10, 16, FALSE, FALSE, FALSE, FALSE, "v210", NULL, NULL},
[DVS10] = {"DVS10", "Centaurus 10bit YUV 4:2:2",
to_fourcc('D','S','1','0'), 48, 8.0 / 3.0, 10, 4, FALSE, FALSE, FALSE, FALSE, "dvs10", NULL, NULL},
[DXT1] = {"DXT1", "S3 Compressed Texture DXT1",
to_fourcc('D','X','T','1'), 0, 0.5, 2, 0, TRUE, TRUE, FALSE, FALSE, "dxt1", NULL, NULL},
[DXT1_YUV] = {"DXT1 YUV", "S3 Compressed Texture DXT1 YUV",
to_fourcc('D','X','T','Y'), 0, 0.5, 2, 0, FALSE, TRUE, FALSE, FALSE, "dxt1y", NULL, NULL}, /* packet YCbCr inside DXT1 channels */
[DXT5] = {"DXT5", "S3 Compressed Texture DXT5 YCoCg",
to_fourcc('D','X','T','5'), 0, 1.0, 4, 0, FALSE, TRUE, FALSE, FALSE, "yog", NULL, NULL},/* DXT5 YCoCg */
[RGB] = {"RGB", "Red Green Blue 24bit",
to_fourcc('R','G','B','2'), 1, 3.0, 8, 3, TRUE, FALSE, FALSE, FALSE, "rgb", NULL, NULL},
[DPX10] = {"DPX10", "DPX10",
to_fourcc('D','P','1','0'), 1, 4.0, 10, 4, TRUE, FALSE, FALSE, FALSE, "dpx", NULL, NULL},
[JPEG] = {"JPEG", "JPEG",
to_fourcc('J','P','E','G'), 0, 1.0, 8, 0, FALSE, TRUE, FALSE, FALSE, "jpg", NULL, NULL},
[RAW] = {"raw", "Raw SDI video",
to_fourcc('r','a','w','s'), 0, 1.0, 0, 0, FALSE, TRUE, FALSE, FALSE, "raw", NULL, NULL}, /* raw SDI */
[H264] = {"H.264", "H.264/AVC",
to_fourcc('A','V','C','1'), 0, 1.0, 8, 0, FALSE, TRUE, TRUE, FALSE, "h264", NULL, NULL},
[H265] = {"H.265", "H.265/HEVC",
to_fourcc('H','E','V','C'), 0, 1.0, 8, 0, FALSE, TRUE, TRUE, FALSE, "h265", NULL, NULL},
[MJPG] = {"MJPEG", "MJPEG",
to_fourcc('M','J','P','G'), 0, 1.0, 8, 0, FALSE, TRUE, FALSE, FALSE, "jpg", NULL, NULL},
[VP8] = {"VP8", "Google VP8",
to_fourcc('V','P','8','0'), 0, 1.0, 8, 0, FALSE, TRUE, TRUE, FALSE, "vp8", NULL, NULL},
[VP9] = {"VP9", "Google VP9",
to_fourcc('V','P','9','0'), 0, 1.0, 8, 0, FALSE, TRUE, TRUE, FALSE, "vp9", NULL, NULL},
[BGR] = {"BGR", "Blue Green Red 24bit",
to_fourcc('B','G','R','2'), 1, 3.0, 8, 0, TRUE, FALSE, FALSE, FALSE, "bgr", NULL, NULL},
[J2K] = {"J2K", "JPEG 2000",
to_fourcc('M','J','2','C'), 0, 1.0, 8, 0, FALSE, TRUE, FALSE, FALSE, "j2k", NULL, NULL},
#ifdef HWACC_VDPAU
[HW_VDPAU] = {"HW_VDPAU", "VDPAU hardware surface",
to_fourcc('V', 'D', 'P', 'S'), 0, 1.0, 8, sizeof(hw_vdpau_frame), FALSE, TRUE, FALSE, TRUE, "vdpau", hw_vdpau_free_extra_data, hw_vdpau_frame_data_cpy}
#endif
};
/**
* This struct specifies alias FourCC used for another FourCC
*/
struct alternative_fourcc {
uint32_t alias;
uint32_t primary_fcc;
};
/**
* This array contains FourCC aliases mapping
*/
static const struct alternative_fourcc fourcc_aliases[] = {
// the following two are here because it was sent with wrong endiannes in past
{to_fourcc('A', 'B', 'G', 'R'), to_fourcc('R', 'G', 'B', 'A')},
{to_fourcc('2', 'B', 'G', 'R'), to_fourcc('R', 'G', 'B', '2')},
// following ones are rather for further compatibility (proposed codecs rename)
{to_fourcc('M', 'J', 'P', 'G'), to_fourcc('J', 'P', 'E', 'G')},
{to_fourcc('2', 'V', 'u', 'y'), to_fourcc('U', 'Y', 'V', 'Y')},
{to_fourcc('2', 'v', 'u', 'y'), to_fourcc('U', 'Y', 'V', 'Y')},
{to_fourcc('d', 'v', 's', '8'), to_fourcc('U', 'Y', 'V', 'Y')},
{to_fourcc('D', 'V', 'S', '8'), to_fourcc('U', 'Y', 'V', 'Y')},
{to_fourcc('y', 'u', 'v', '2'), to_fourcc('U', 'Y', 'V', 'Y')},
{to_fourcc('y', 'u', 'V', '2'), to_fourcc('U', 'Y', 'V', 'Y')},
};
struct alternative_codec_name {
const char *alias;
const char *primary_name;
};
static const struct alternative_codec_name codec_name_aliases[] = {
{"2vuy", "UYVY"},
{"AVC", "H.264"},
{"HEVC", "H.265"},
};
void show_codec_help(const char *module, const codec_t *codecs8, const codec_t *codecs10)
{
printf("\tSupported codecs (%s):\n", module);
printf("\t\t8bits\n");
while (*codecs8 != VIDEO_CODEC_NONE) {
printf("\t\t\t'%s' - %s\n", codec_info[*codecs8].name, codec_info[*codecs8].name_long);
codecs8++;
}
printf("\t\t10bits\n");
while (*codecs10 != VIDEO_CODEC_NONE) {
printf("\t\t\t'%s' - %s\n", codec_info[*codecs10].name, codec_info[*codecs10].name_long);
codecs10++;
}
}
int get_bits_per_component(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].bits_per_channel;
} else {
return 0;
}
}
double get_bpp(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].bpp;
} else {
return 0;
}
}
uint32_t get_fourcc(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].fcc;
} else {
return 0;
}
}
const char * get_codec_name(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].name;
} else {
return 0;
}
}
const char * get_codec_name_long(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].name_long;
} else {
return 0;
}
}
codec_t get_codec_from_fcc(uint32_t fourcc)
{
for (unsigned int i = 0; i < sizeof codec_info / sizeof(struct codec_info_t); ++i) {
if (fourcc == codec_info[i].fcc)
return i;
}
// try to look through aliases
for (size_t i = 0; i < sizeof(fourcc_aliases) / sizeof(struct alternative_fourcc); ++i) {
if (fourcc == fourcc_aliases[i].alias) {
for (unsigned int j = 0; j < sizeof codec_info / sizeof(struct codec_info_t); ++j) {
if (fourcc_aliases[i].primary_fcc == codec_info[j].fcc)
return j;
}
}
}
return VIDEO_CODEC_NONE;
}
/**
* Helper codec finding function
*
* Iterates through codec list and finds appropriate codec.
*
* @returns codec
*/
static codec_t get_codec_from_name_wo_alias(const char *name)
{
for (unsigned int i = 0; i < sizeof codec_info / sizeof(struct codec_info_t); ++i) {
if (codec_info[i].name && strcasecmp(codec_info[i].name, name) == 0) {
return i;
}
}
return VIDEO_CODEC_NONE;
}
codec_t get_codec_from_name(const char *name)
{
codec_t ret = get_codec_from_name_wo_alias(name);
if (ret != VIDEO_CODEC_NONE) {
return ret;
}
// try to find if this is not an alias
for (size_t i = 0; i < sizeof(codec_name_aliases) / sizeof(struct alternative_codec_name); ++i) {
if (strcasecmp(name, codec_name_aliases[i].alias) == 0) {
ret = get_codec_from_name_wo_alias(codec_name_aliases[i].primary_name);
if (ret != VIDEO_CODEC_NONE) {
return ret;
}
}
}
return VIDEO_CODEC_NONE;
}
const char *get_codec_file_extension(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].file_extension;
} else {
return 0;
}
}
/**
* @retval TRUE if codec is compressed
* @retval FALSE if codec is pixelformat
*/
int is_codec_opaque(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].opaque;
} else {
return 0;
}
}
/**
* Returns whether specified codec is an interframe compression.
* Not defined for pixelformats
* @retval TRUE if compression is interframe
* @retval FALSE if compression is not interframe
*/
int is_codec_interframe(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].interframe;
} else {
return 0;
}
}
/** @brief Returns TRUE if specified pixelformat is some form of RGB (not YUV).
*
* Unspecified for compressed codecs.
* @retval TRUE if pixelformat is RGB
* @retval FALSE if pixelformat is not a RGB */
int codec_is_a_rgb(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].rgb;
} else {
return 0;
}
}
/** @brief Returns TRUE if specified pixelformat has constant size regardles
* of resolution. If so the block_size value represents the size.
*
* Unspecified for compressed codecs.
* @retval TRUE if pixelformat is const size
* @retval FALSE if pixelformat is not const size */
int codec_is_const_size(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].const_size;
} else {
return 0;
}
}
int get_halign(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].h_align;
} else {
return 0;
}
}
/** @brief Returns aligned linesize according to pixelformat specification (in pixels) */
int get_aligned_length(int width_pixels, codec_t codec)
{
int h_align = get_halign(codec);
assert(h_align > 0);
return ((width_pixels + h_align - 1) / h_align) * h_align;
}
/** @brief Returns aligned linesize according to pixelformat specification (in bytes) */
int vc_get_linesize(unsigned int width, codec_t codec)
{
if (codec >= sizeof codec_info / sizeof(struct codec_info_t)) {
return 0;
}
if (codec_info[codec].h_align) {
width =
((width + codec_info[codec].h_align -
1) / codec_info[codec].h_align) *
codec_info[codec].h_align;
}
return width * codec_info[codec].bpp;
}
/// @brief returns @ref codec_info_t::block_size
int get_pf_block_size(codec_t codec)
{
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].block_size;
} else {
return 0;
}
}
void (*get_free_extra_data_fcn(codec_t codec))(void *){
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].free_extra_data;
} else {
return 0;
}
}
void *(*get_copy_data_fcn(codec_t codec))(void *, const void *, size_t){
unsigned int i = (unsigned int) codec;
if (i < sizeof codec_info / sizeof(struct codec_info_t)) {
return codec_info[i].copy_data;
} else {
return 0;
}
}
/** @brief Deinterlaces framebuffer.
*
* vc_deinterlace performs linear blend deinterlace on a framebuffer.
* @param[in,out] src framebuffer to be deinterlaced
* @param[in] src_linesize length of a line (bytes)
* @param[in] lines number of lines
* @see vc_deinterlace_aligned
* @see vc_deinterlace_unaligned
*/
void vc_deinterlace(unsigned char *src, long src_linesize, int lines)
{
#ifdef __SSE2__
if(((long int) src & 0x0F) == 0 && src_linesize % 16 == 0) {
vc_deinterlace_aligned(src, src_linesize, lines);
} else {
vc_deinterlace_unaligned(src, src_linesize, lines);
}
#else
for (int y = 0; y < lines; y += 2) {
for (int x = 0; x < src_linesize; ++x) {
int val = (*src + src[src_linesize] + 1) >> 1;
*src = src[src_linesize] = val;
src++;
}
src += src_linesize;
}
#endif
}
#ifdef __SSE2__
/**
* Aligned version of deinterlace filter
*
* @param src 16-byte aligned buffer
* @see vc_deinterlace
*/
static void vc_deinterlace_aligned(unsigned char *src, long src_linesize, int lines)
{
int i, j;
long pitch = src_linesize;
register long pitch2 = pitch * 2;
unsigned char *bline1, *bline2, *bline3;
register unsigned char *line1, *line2, *line3;
bline1 = src;
bline2 = src + pitch;
bline3 = src + 3 * pitch;
for (i = 0; i < src_linesize; i += 16) {
/* preload first two lines */
asm volatile ("movdqa (%0), %%xmm0\n"
"movdqa (%1), %%xmm1\n"::"r" ((unsigned long *)(void *)
bline1),
"r"((unsigned long *)(void *) bline2));
line1 = bline2;
line2 = bline2 + pitch;
line3 = bline3;
for (j = 0; j < lines - 4; j += 2) {
asm volatile ("movdqa (%1), %%xmm2\n"
"pavgb %%xmm2, %%xmm0\n"
"pavgb %%xmm1, %%xmm0\n"
"movdqa (%2), %%xmm1\n"
"movdqa %%xmm0, (%0)\n"
"pavgb %%xmm1, %%xmm0\n"
"pavgb %%xmm2, %%xmm0\n"
"movdqa %%xmm0, (%1)\n"::"r" ((unsigned
long *) (void *) line1),
"r"((unsigned long *) (void *) line2),
"r"((unsigned long *) (void *) line3)
);
line1 += pitch2;
line2 += pitch2;
line3 += pitch2;
}
bline1 += 16;
bline2 += 16;
bline3 += 16;
}
}
/**
* Unaligned version of deinterlace filter
*
* @param src 4-byte aligned buffer
* @see vc_deinterlace
*/
static void vc_deinterlace_unaligned(unsigned char *src, long src_linesize, int lines)
{
int i, j;
long pitch = src_linesize;
register long pitch2 = pitch * 2;
unsigned char *bline1, *bline2, *bline3;
register unsigned char *line1, *line2, *line3;
bline1 = src;
bline2 = src + pitch;
bline3 = src + 3 * pitch;
for (i = 0; i < src_linesize; i += 16) {
/* preload first two lines */
asm volatile ("movdqu (%0), %%xmm0\n"
"movdqu (%1), %%xmm1\n"::"r" (bline1),
"r" (bline2));
line1 = bline2;
line2 = bline2 + pitch;
line3 = bline3;
for (j = 0; j < lines - 4; j += 2) {
asm volatile ("movdqu (%1), %%xmm2\n"
"pavgb %%xmm2, %%xmm0\n"
"pavgb %%xmm1, %%xmm0\n"
"movdqu (%2), %%xmm1\n"
"movdqu %%xmm0, (%0)\n"
"pavgb %%xmm1, %%xmm0\n"
"pavgb %%xmm2, %%xmm0\n"
"movdqu %%xmm0, (%1)\n"::"r" (line1),
"r" (line2),
"r" (line3)
);
line1 += pitch2;
line2 += pitch2;
line3 += pitch2;
}
bline1 += 16;
bline2 += 16;
bline3 += 16;
}
}
#endif
/**
* @brief Converts v210 to UYVY
* @param[out] dst 4-byte aligned output buffer where UYVY will be stored
* @param[in] src 4-byte aligned input buffer containing v210 (by definition of v210
* should be even aligned to 16B boundary)
* @param[in] dst_len length of data that should be writen to dst buffer (in bytes)
*/
void vc_copylinev210(unsigned char *dst, const unsigned char *src, int dst_len)
{
struct {
unsigned a:10;
unsigned b:10;
unsigned c:10;
unsigned p1:2;
} const *s;
register uint32_t *d;
register uint32_t tmp;
d = (uint32_t *)(void *) dst;
s = (const void *)src;
while (dst_len >= 12) {
tmp = (s->a >> 2) | (s->b >> 2) << 8 | (((s)->c >> 2) << 16);
s++;
*(d++) = tmp | ((s->a >> 2) << 24);
tmp = (s->b >> 2) | (((s)->c >> 2) << 8);
s++;
*(d++) = tmp | ((s->a >> 2) << 16) | ((s->b >> 2) << 24);
tmp = (s->c >> 2);
s++;
*(d++) =
tmp | ((s->a >> 2) << 8) | ((s->b >> 2) << 16) |
((s->c >> 2) << 24);
s++;
dst_len -= 12;
}
if (dst_len >= 4) {
tmp = (s->a >> 2) | (s->b >> 2) << 8 | (((s)->c >> 2) << 16);
s++;
*(d++) = tmp | ((s->a >> 2) << 24);
}
if (dst_len >= 8) {
tmp = (s->b >> 2) | (((s)->c >> 2) << 8);
s++;
*(d++) = tmp | ((s->a >> 2) << 16) | ((s->b >> 2) << 24);
}
}
/**
* @brief Converts from YUYV to UYVY.
* @copydetails vc_copylinev210
*/
void vc_copylineYUYV(unsigned char *dst, const unsigned char *src, int dst_len)
{
#if defined __SSE2__
register uint32_t *d;
register const uint32_t *s;
const uint32_t * const end = (uint32_t *)(void *) dst + dst_len / 4;
uint32_t mask[4] = {
0xff00ff00ul,
0xff00ff00ul,
0xff00ff00ul,
0xff00ff00ul};
d = (uint32_t *)(void *) dst;
s = (const uint32_t *)(const void *) src;
assert(dst_len % 4 == 0);
if((dst_len % 16 == 0)) {
asm("movdqu (%0), %%xmm4\n"
"movdqa %%xmm4, %%xmm5\n"
"psrldq $1, %%xmm5\n"
: :"r"(mask));
while(d < end) {
asm volatile ("movdqu (%0), %%xmm0\n"
"movdqu %%xmm0, %%xmm1\n"
"pand %%xmm4, %%xmm0\n"
"psrldq $1, %%xmm0\n"
"pand %%xmm5, %%xmm1\n"
"pslldq $1, %%xmm1\n"
"por %%xmm0, %%xmm1\n"
"movdqu %%xmm1, (%1)\n"::"r" (s), "r"(d));
s += 4;
d += 4;
}
} else {
while(d < end) {
register uint32_t tmp = *s;
*d = ((tmp & 0x00ff0000) << 8) | ((tmp & 0xff000000) >> 8) |
((tmp & 0x000000ff) << 8) | ((tmp & 0x0000ff00) >> 8);
s++;
d++;
}
}
#else
char u, y1, v, y2;
while (dst_len > 0) {
y1 = *src++;
u = *src++;
y2 = *src++;
v = *src++;
*dst++ = u;
*dst++ = y1;
*dst++ = v;
*dst++ = y2;
dst_len -= 4;
}
#endif
}
/**
* @brief Converts from R10k to RGBA
*
* @param[out] dst 4B-aligned buffer that will contain result
* @param[in] src 4B-aligned buffer containing pixels in R10k
* @param[in] dst_len length of data that should be writen to dst buffer (in bytes)
* @param[in] rshift destination red shift
* @param[in] gshift destination green shift
* @param[in] bshift destination blue shift
*/
void
vc_copyliner10k(unsigned char *dst, const unsigned char *src, int len, int rshift,
int gshift, int bshift)
{
struct {
unsigned r:8;
unsigned gh:6;
unsigned p1:2;
unsigned bh:4;
unsigned p2:2;
unsigned gl:2;
unsigned p3:2;
unsigned p4:2;
unsigned bl:4;
} const *s;
register uint32_t *d;
register uint32_t tmp;
d = (uint32_t *)(void *) dst;
s = (const void *)(const void *) src;
while (len > 0) {
tmp =
(s->
r << rshift) | (((s->gh << 2) | s->
gl) << gshift) | (((s->bh << 4) | s->
bl) << bshift);
s++;
*(d++) = tmp;
tmp =
(s->
r << rshift) | (((s->gh << 2) | s->
gl) << gshift) | (((s->bh << 4) | s->
bl) << bshift);
s++;
*(d++) = tmp;
tmp =
(s->
r << rshift) | (((s->gh << 2) | s->
gl) << gshift) | (((s->bh << 4) | s->
bl) << bshift);
s++;
*(d++) = tmp;
tmp =
(s->
r << rshift) | (((s->gh << 2) | s->
gl) << gshift) | (((s->bh << 4) | s->
bl) << bshift);
s++;
*(d++) = tmp;
len -= 16;
}
}
/**
* @brief Changes color channels' order in RGBA
* @copydetails vc_copyliner10k
*/
void
vc_copylineRGBA(unsigned char *dst, const unsigned char *src, int len, int rshift,
int gshift, int bshift)
{
register uint32_t *d = (uint32_t *)(void *) dst;
register const uint32_t *s = (const uint32_t *)(const void *) src;
register uint32_t tmp;
if (rshift == 0 && gshift == 8 && bshift == 16) {
memcpy(dst, src, len);
} else {
while (len > 0) {
register unsigned int r, g, b;
tmp = *(s++);
r = tmp & 0xff;
g = (tmp >> 8) & 0xff;
b = (tmp >> 16) & 0xff;
tmp = (r << rshift) | (g << gshift) | (b << bshift);
*(d++) = tmp;
tmp = *(s++);
r = tmp & 0xff;
g = (tmp >> 8) & 0xff;
b = (tmp >> 16) & 0xff;
tmp = (r << rshift) | (g << gshift) | (b << bshift);
*(d++) = tmp;
tmp = *(s++);
r = tmp & 0xff;
g = (tmp >> 8) & 0xff;
b = (tmp >> 16) & 0xff;
tmp = (r << rshift) | (g << gshift) | (b << bshift);
*(d++) = tmp;
tmp = *(s++);
r = tmp & 0xff;
g = (tmp >> 8) & 0xff;
b = (tmp >> 16) & 0xff;
tmp = (r << rshift) | (g << gshift) | (b << bshift);
*(d++) = tmp;
len -= 16;
}
}
}
/**
* @brief Converts from DVS10 to v210
* @copydetails vc_copylinev210
*/
void vc_copylineDVS10toV210(unsigned char *dst, const unsigned char *src, int dst_len)
{
unsigned int *d;
const unsigned int *s1;
register unsigned int a,b;
d = (unsigned int *)(void *) dst;
s1 = (const unsigned int *)(const void *) src;
while(dst_len > 0) {
a = b = *s1++;
b = ((b >> 24) * 0x00010101) & 0x00300c03;
a <<= 2;
b |= a & (0xff<<2);
a <<= 2;
b |= a & (0xff00<<4);
a <<= 2;
b |= a & (0xff0000<<6);
*d++ = b;
dst_len -= 4;
}
}
/* convert 10bits Cb Y Cr A Y Cb Y A to 8bits Cb Y Cr Y Cb Y */
/* TODO: undo it - currently this decoder is broken */
#if 0 /* !(HAVE_MACOSX || HAVE_32B_LINUX) */
void vc_copylineDVS10(unsigned char *dst, unsigned char *src, int src_len)
{
register unsigned char *_d = dst, *_s = src;
while (src_len > 0) {
asm("movd %0, %%xmm4\n": :"r"(0xffffff));
asm volatile ("movdqa (%0), %%xmm0\n"
"movdqa 16(%0), %%xmm5\n"
"movdqa %%xmm0, %%xmm1\n"
"movdqa %%xmm0, %%xmm2\n"
"movdqa %%xmm0, %%xmm3\n"
"pand %%xmm4, %%xmm0\n"
"movdqa %%xmm5, %%xmm6\n"
"movdqa %%xmm5, %%xmm7\n"
"movdqa %%xmm5, %%xmm8\n"
"pand %%xmm4, %%xmm5\n"
"pslldq $4, %%xmm4\n"
"pand %%xmm4, %%xmm1\n"
"pand %%xmm4, %%xmm6\n"
"pslldq $4, %%xmm4\n"
"psrldq $1, %%xmm1\n"
"psrldq $1, %%xmm6\n"
"pand %%xmm4, %%xmm2\n"
"pand %%xmm4, %%xmm7\n"
"pslldq $4, %%xmm4\n"
"psrldq $2, %%xmm2\n"
"psrldq $2, %%xmm7\n"
"pand %%xmm4, %%xmm3\n"
"pand %%xmm4, %%xmm8\n"
"por %%xmm1, %%xmm0\n"
"psrldq $3, %%xmm3\n"
"psrldq $3, %%xmm8\n"
"por %%xmm2, %%xmm0\n"
"por %%xmm6, %%xmm5\n"
"por %%xmm3, %%xmm0\n"
"por %%xmm7, %%xmm5\n"
"movdq2q %%xmm0, %%mm0\n"
"por %%xmm8, %%xmm5\n"
"movdqa %%xmm5, %%xmm1\n"
"pslldq $12, %%xmm5\n"
"psrldq $4, %%xmm1\n"
"por %%xmm5, %%xmm0\n"
"psrldq $8, %%xmm0\n"
"movq %%mm0, (%1)\n"
"movdq2q %%xmm0, %%mm1\n"
"movdq2q %%xmm1, %%mm2\n"
"movq %%mm1, 8(%1)\n"
"movq %%mm2, 16(%1)\n"::"r" (_s), "r"(_d));
_s += 32;
_d += 24;
src_len -= 32;
}
}
#else
/**
* @brief Converts from DVS10 to UYVY
* @copydetails vc_copylinev210
*/
void vc_copylineDVS10(unsigned char *dst, const unsigned char *src, int dst_len)
{
int src_len = dst_len / 1.5; /* right units */
register const uint64_t *s;
register uint64_t *d;
register uint64_t a1, a2, a3, a4;
d = (uint64_t *)(void *) dst;
s = (const uint64_t *)(const void *) src;
while (src_len > 0) {
a1 = *(s++);
a2 = *(s++);
a3 = *(s++);
a4 = *(s++);
a1 = (a1 & 0xffffff) | ((a1 >> 8) & 0xffffff000000LL);
a2 = (a2 & 0xffffff) | ((a2 >> 8) & 0xffffff000000LL);
a3 = (a3 & 0xffffff) | ((a3 >> 8) & 0xffffff000000LL);
a4 = (a4 & 0xffffff) | ((a4 >> 8) & 0xffffff000000LL);
*(d++) = a1 | (a2 << 48); /* 0xa2|a2|a1|a1|a1|a1|a1|a1 */
*(d++) = (a2 >> 16) | (a3 << 32); /* 0xa3|a3|a3|a3|a2|a2|a2|a2 */
*(d++) = (a3 >> 32) | (a4 << 16); /* 0xa4|a4|a4|a4|a4|a4|a3|a3 */
src_len -= 16;
}
}
#endif /* !(HAVE_MACOSX || HAVE_32B_LINUX) */
/**
* @brief Changes color order of an RGB
* @copydetails vc_copyliner10k
*/
void vc_copylineRGB(unsigned char *dst, const unsigned char *src, int dst_len, int rshift, int gshift, int bshift)
{
register unsigned int r, g, b;
union {
unsigned int out;
unsigned char c[4];
} u;
if (rshift == 0 && gshift == 8 && bshift == 16) {
memcpy(dst, src, dst_len);
} else {
while (dst_len >= 3) {
r = *src++;
g = *src++;
b = *src++;
u.out = (r << rshift) | (g << gshift) | (b << bshift);
*dst++ = u.c[0];
*dst++ = u.c[1];
*dst++ = u.c[2];
dst_len -= 3;
}
}
}
/**
* @brief Converts from RGBA to RGB
* @copydetails vc_copylinev210
*/
void vc_copylineRGBAtoRGB(unsigned char *dst, const unsigned char *src, int dst_len, int rshift, int gshift, int bshift)
{
UNUSED(rshift);
UNUSED(gshift);
UNUSED(bshift);
vc_copylineRGBAtoRGBwithShift(dst, src, dst_len, 0, 8, 16);
}
/**
* @brief Converts from RGBA to RGB. Channels in RGBA can be differently ordered.
* @copydetails vc_copyliner10k
*/
void vc_copylineRGBAtoRGBwithShift(unsigned char *dst2, const unsigned char *src2, int dst_len, int rshift, int gshift, int bshift)
{
register const uint32_t * src = (const uint32_t *)(const void *) src2;
register uint32_t * dst = (uint32_t *)(void *) dst2;
while (dst_len >= 12) {
register uint32_t in1 = *src++;
register uint32_t in2 = *src++;
register uint32_t in3 = *src++;
register uint32_t in4 = *src++;
*dst++ = ((in2 >> rshift)) << 24 |
((in1 >> bshift) & 0xff) << 16 |
((in1 >> gshift) & 0xff) << 8 |
((in1 >> rshift) & 0xff);
*dst++ = ((in3 >> gshift)) << 24 |
((in3 >> rshift) & 0xff) << 16 |
((in2 >> bshift) & 0xff) << 8 |
((in2 >> gshift) & 0xff);
*dst++ = ((in4 >> bshift)) << 24 |
((in4 >> gshift) & 0xff) << 16 |
((in4 >> rshift) & 0xff) << 8 |
((in3 >> bshift) & 0xff);
dst_len -= 12;
}
uint8_t *dst_c = (uint8_t *) dst;
while (dst_len >= 4) {
register uint32_t in = *src++;
*dst_c++ = (in >> rshift) & 0xff;
*dst_c++ = (in >> gshift) & 0xff;
*dst_c++ = (in >> bshift) & 0xff;
dst_len -= 4;
}
}
/**
* @brief Converts from AGBR to RGB
* @copydetails vc_copylinev210
* @see vc_copylineRGBAtoRGBwithShift
* @see vc_copylineRGBAtoRGB
*/
void vc_copylineABGRtoRGB(unsigned char *dst2, const unsigned char *src2, int dst_len, int rshift, int gshift, int bshift)
{
UNUSED(rshift);
UNUSED(gshift);
UNUSED(bshift);
vc_copylineRGBAtoRGBwithShift(dst2, src2, dst_len, 16, 8, 0);
}
/**
* @brief Converts RGBA with different shifts to RGBA
*/
void vc_copylineToRGBA(unsigned char *dst, const unsigned char *src, int dst_len,
int src_rshift, int src_gshift, int src_bshift)
{
register const uint32_t * in = (const uint32_t *)(const void *) src;
register uint32_t * out = (uint32_t *)(void *) dst;
while (dst_len >= 4) {
register uint32_t in_val = *in++;
*out++ = ((in_val >> src_rshift) & 0xff) |
((in_val >> src_gshift) & 0xff) << 8 |
((in_val >> src_bshift) & 0xff) << 16;
dst_len -= 4;
}
}
/**
* @brief Converts UYVY to grayscale.
* @todo is this correct??
*/
void vc_copylineUYVYtoGrayscale(unsigned char *dst, const unsigned char *src, int dst_len) {
while(dst_len > 0) {
src++; // U
*dst++ = *src++; // Y
src++; // V
*dst++ = *src++; // Y
dst_len -= 2;
}
}
/**
* @brief Converts RGB to RGBA
* @copydetails vc_copyliner10k
*/
void vc_copylineRGBtoRGBA(unsigned char *dst, const unsigned char *src, int dst_len, int rshift, int gshift, int bshift)
{
register unsigned int r, g, b;
register uint32_t *d = (uint32_t *)(void *) dst;
while (dst_len >= 4) {
r = *src++;
g = *src++;
b = *src++;
*d++ = (r << rshift) | (g << gshift) | (b << bshift);
dst_len -= 4;
}
}
/**
* @brief Converts RGB(A) into UYVY
*
* Uses full scale Rec. 601 YUV (aka JPEG)
* @copydetails vc_copyliner10k
* @param[in] source pixel size (3 for RGB, 4 for RGBA)
*/
static void vc_copylineToUYVY601(unsigned char *dst, const unsigned char *src, int dst_len,
int rshift, int gshift, int bshift, int pix_size) {
register int r, g, b;
register int y1, y2, u ,v;
register uint32_t *d = (uint32_t *)(void *) dst;
while (dst_len >= 4) {
r = *(src + rshift);
g = *(src + gshift);
b = *(src + bshift);
src += pix_size;
y1 = 19595 * r + 38469 * g + 7471 * b;
u = -9642 * r -18931 * g + 28573 * b;
v = 40304 * r - 33750 * g - 6554 * b;
r = *(src + rshift);
g = *(src + gshift);
b = *(src + bshift);
src += pix_size;
y2 = 19595 * r + 38469 * g + 7471 * b;
u += -9642 * r -18931 * g + 28573 * b;
v += 40304 * r - 33750 * g - 6554 * b;
u = u / 2 + (1<<23);
v = v / 2 + (1<<23);
*d++ = (min(max(y2, 0), (1<<24)-1) >> 16) << 24 |
(min(max(v, 0), (1<<24)-1) >> 16) << 16 |
(min(max(y1, 0), (1<<24)-1) >> 16) << 8 |
(min(max(u, 0), (1<<24)-1) >> 16);
dst_len -= 4;
}
}
/**
* @brief Converts RGB(A) into UYVY
*
* Uses Rec. 709 with standard SDI ceiling and floor
* @copydetails vc_copyliner10k
* @param[in] source pixel size (3 for RGB, 4 for RGBA)
*/
static void vc_copylineToUYVY709(unsigned char *dst, const unsigned char *src, int dst_len,
int rshift, int gshift, int bshift, int pix_size) {
register int r, g, b;
register int y1, y2, u ,v;
register uint32_t *d = (uint32_t *)(void *) dst;
while (dst_len >= 4) {
r = *(src + rshift);
g = *(src + gshift);
b = *(src + bshift);
src += pix_size;
y1 = 11993 * r + 40239 * g + 4063 * b + (1<<20);
u = -6619 * r -22151 * g + 28770 * b;
v = 28770 * r - 26149 * g - 2621 * b;
r = *(src + rshift);
g = *(src + gshift);
b = *(src + bshift);
src += pix_size;
y2 = 11993 * r + 40239 * g + 4063 * b + (1<<20);
u += -6619 * r -22151 * g + 28770 * b;
v += 28770 * r - 26149 * g - 2621 * b;
u = u / 2 + (1<<23);
v = v / 2 + (1<<23);
*d++ = (min(max(y2, 0), (1<<24)-1) >> 16) << 24 |
(min(max(v, 0), (1<<24)-1) >> 16) << 16 |
(min(max(y1, 0), (1<<24)-1) >> 16) << 8 |
(min(max(u, 0), (1<<24)-1) >> 16);
dst_len -= 4;
}
}
/**
* @brief Converts UYVY to RGB.
* Uses Rec. 709 with standard SDI ceiling and floor
* @copydetails vc_copylinev210
* @todo make it faster if needed
*/
void vc_copylineUYVYtoRGB(unsigned char *dst, const unsigned char *src, int dst_len) {
while (dst_len >= 4) {
register int y1, y2, u ,v;
u = *src++;
y1 = *src++;
v = *src++;
y2 = *src++;
*dst++ = min(max(1.164*(y1 - 16) + 1.793*(v - 128), 0), 255);
*dst++ = min(max(1.164*(y1 - 16) - 0.534*(v - 128) - 0.213*(u - 128), 0), 255);
*dst++ = min(max(1.164*(y1 - 16) + 2.115*(u - 128), 0), 255);
*dst++ = min(max(1.164*(y2 - 16) + 1.793*(v - 128), 0), 255);
*dst++ = min(max(1.164*(y2 - 16) - 0.534*(v - 128) - 0.213*(u - 128), 0), 255);
*dst++ = min(max(1.164*(y2 - 16) + 2.115*(u - 128), 0), 255);
dst_len -= 6;
}
}
/**
* @brief Converts UYVY to RGB using SSE.
* Uses Rec. 709 with standard SDI ceiling and floor
* There can be some inaccuracies due to the use of integer arithmetic
* @copydetails vc_copylinev210
* @todo make it faster if needed
*/
void vc_copylineUYVYtoRGB_SSE(unsigned char *dst, const unsigned char *src, int dst_len){
#ifdef __SSSE3__
__m128i yfactor = _mm_set1_epi16(74); //1.164 << 6
__m128i rvfactor = _mm_set1_epi16(115); //1.793 << 6
__m128i gvfactor = _mm_set1_epi16(34); //0.534 << 6
__m128i gufactor = _mm_set1_epi16(14); //0.213 << 6
__m128i bufactor = _mm_set1_epi16(135); //2.115 << 6
__m128i ysub = _mm_set1_epi16(16);
__m128i uvsub = _mm_set1_epi16(128);
__m128i zero128 = _mm_set1_epi32(0);
__m128i max = _mm_set1_epi16(255);
//YYVVYYUU
__m128i ymask = _mm_set1_epi32(0x00FF00FF);
__m128i umask = _mm_set1_epi32(0x000000FF);
__m128i rgbshuffle = _mm_setr_epi8(0, 2, 1, 4, 6, 5, 8, 10, 9, 12, 14, 13, 15, 11, 7, 3);
__m128i yuv;
__m128i rgb;
__m128i y;
__m128i u;
__m128i v;
__m128i r;
__m128i g;
__m128i b;
while(dst_len >= 28){
yuv = _mm_lddqu_si128((__m128i const*) src);
src += 16;
u = _mm_and_si128(yuv, umask);
u = _mm_or_si128(u, _mm_bslli_si128(u, 2));
yuv = _mm_bsrli_si128(yuv, 1);
y = _mm_and_si128(yuv, ymask);
yuv = _mm_bsrli_si128(yuv, 1);
v = _mm_and_si128(yuv, umask);
v = _mm_or_si128(v, _mm_bslli_si128(v, 2));
y = _mm_subs_epi16(y, ysub);
y = _mm_mullo_epi16(y, yfactor);
u = _mm_subs_epi16(u, uvsub);
v = _mm_subs_epi16(v, uvsub);
r = _mm_adds_epi16(y, _mm_mullo_epi16(v, rvfactor));
g = _mm_subs_epi16(y, _mm_mullo_epi16(v, gvfactor));
g = _mm_subs_epi16(g, _mm_mullo_epi16(u, gufactor));
b = _mm_adds_epi16(y, _mm_mullo_epi16(u, bufactor));
//Make sure that the result is in the interval 0..255
r = _mm_max_epi16(zero128, r);
g = _mm_max_epi16(zero128, g);
b = _mm_max_epi16(zero128, b);
r = _mm_srli_epi16(r, 6);
g = _mm_srli_epi16(g, 6);
b = _mm_srli_epi16(b, 6);
r = _mm_min_epi16(max, r);
g = _mm_min_epi16(max, g);
b = _mm_min_epi16(max, b);
rgb = _mm_or_si128(_mm_bslli_si128(g, 1), r);
rgb = _mm_unpacklo_epi8(rgb, b);
rgb = _mm_shuffle_epi8(rgb, rgbshuffle);
_mm_storeu_si128((__m128i *) dst, rgb);
dst += 12;
rgb = _mm_or_si128(_mm_bslli_si128(g, 1), r);
rgb = _mm_unpackhi_epi8(rgb, b);
rgb = _mm_shuffle_epi8(rgb, rgbshuffle);
_mm_storeu_si128((__m128i *) dst, rgb);
dst += 12;
dst_len -= 24;
}
#endif
//copy last few pixels
vc_copylineUYVYtoRGB(dst, src, dst_len);
}
/**
* @brief Converts RGB to UYVY.
* Uses full scale Rec. 601 YUV (aka JPEG)
* @copydetails vc_copylinev210
*/
void vc_copylineRGBtoUYVY(unsigned char *dst, const unsigned char *src, int dst_len)
{
vc_copylineToUYVY709(dst, src, dst_len, 0, 1, 2, 3);
}
/**
* @brief Converts RGB to UYVY using SSE.
* Uses full scale Rec. 601 YUV (aka JPEG)
* There can be some inaccuracies due to the use of integer arithmetic
* @copydetails vc_copylinev210
*/
void vc_copylineRGBtoUYVY_SSE(unsigned char *dst, const unsigned char *src, int dst_len){
#ifdef __SSSE3__
__m128i rgb;
__m128i yuv;
__m128i r;
__m128i g;
__m128i b;
__m128i y;
__m128i u;
__m128i v;
//BB BB BB BB GR GR GR GR
__m128i shuffle = _mm_setr_epi8(0, 1, 3, 4, 6, 7, 9, 10, 2, 2, 5, 5, 8, 8, 11, 11);
__m128i mask = _mm_setr_epi16(0x00FF, 0x00FF, 0x00FF, 0x00FF, 0, 0, 0, 0);
__m128i lowmask = _mm_setr_epi16(0, 0, 0, 0, 0x00FF, 0x00FF, 0x00FF, 0x00FF);
__m128i zero = _mm_set1_epi16(0);
__m128i yrf = _mm_set1_epi16(23);
__m128i ygf = _mm_set1_epi16(79);
__m128i ybf = _mm_set1_epi16(8);
__m128i urf = _mm_set1_epi16(13);
__m128i ugf = _mm_set1_epi16(43);
__m128i ubf = _mm_set1_epi16(56);
__m128i vrf = _mm_set1_epi16(56);
__m128i vgf = _mm_set1_epi16(51);
__m128i vbf = _mm_set1_epi16(5);
__m128i yadd = _mm_set1_epi16(2048);
__m128i uvadd = _mm_set1_epi16(16384);
while(dst_len >= 16){
//Load first 4 pixels
rgb = _mm_lddqu_si128((__m128i const*) src);
src += 12;
rgb = _mm_shuffle_epi8(rgb, shuffle);
r = _mm_and_si128(rgb, mask);
rgb = _mm_bsrli_si128(rgb, 1);
//0B BB BB BB BG RG RG RG
g = _mm_and_si128(rgb, mask);
rgb = _mm_bsrli_si128(rgb, 7);
//00 00 00 00 BB BB BB BB
b = _mm_and_si128(rgb, mask);
//Load next 4 pixels
rgb = _mm_lddqu_si128((__m128i const*) src);
src += 12;
rgb = _mm_shuffle_epi8(rgb, shuffle);
b = _mm_or_si128(b, _mm_and_si128(rgb, lowmask));
rgb = _mm_bslli_si128(rgb, 7);
g = _mm_or_si128(g, _mm_and_si128(rgb, lowmask));
//0B BB BB BB BG RG RG RG
rgb = _mm_bslli_si128(rgb, 1);
r = _mm_or_si128(r, _mm_and_si128(rgb, lowmask));
//00 00 00 00 BB BB BB BB
//Compute YUV values
y = _mm_adds_epi16(yadd, _mm_mullo_epi16(r, yrf));
y = _mm_adds_epi16(y, _mm_mullo_epi16(g, ygf));
y = _mm_adds_epi16(y, _mm_mullo_epi16(b, ybf));
u = _mm_subs_epi16(uvadd, _mm_mullo_epi16(r, urf));
u = _mm_subs_epi16(u, _mm_mullo_epi16(g, ugf));
u = _mm_adds_epi16(u, _mm_mullo_epi16(b, ubf));
v = _mm_adds_epi16(uvadd, _mm_mullo_epi16(r, vrf));
v = _mm_subs_epi16(v, _mm_mullo_epi16(g, vgf));
v = _mm_subs_epi16(v, _mm_mullo_epi16(b, vbf));
y = _mm_srli_epi16(y, 7);
u = _mm_srli_epi16(u, 7);
v = _mm_srli_epi16(v, 7);
u = _mm_hadd_epi16(u, v);
u = _mm_srli_epi16(u, 1);
v = _mm_unpackhi_epi16(zero, u);
u = _mm_unpacklo_epi16(u, zero);
y = _mm_bslli_si128(y, 1);
yuv = _mm_or_si128(y, u);
yuv = _mm_or_si128(yuv, v);
_mm_storeu_si128((__m128i *) dst, yuv);
dst += 16;
dst_len -= 16;
}
#endif
//copy last few pixels
vc_copylineRGBtoUYVY(dst, src, dst_len);
}
/**
* @brief Converts RGB to Grayscale using SSE.
* There can be some inaccuracies due to the use of integer arithmetic
* @copydetails vc_copylinev210
*/
void vc_copylineRGBtoGrayscale_SSE(unsigned char *dst, const unsigned char *src, int dst_len){
#ifdef __SSSE3__
__m128i rgb;
__m128i r;
__m128i g;
__m128i b;
__m128i y;
//BB BB BB BB GR GR GR GR
__m128i inshuffle = _mm_setr_epi8(0, 1, 3, 4, 6, 7, 9, 10, 2, 2, 5, 5, 8, 8, 11, 11);
__m128i outshuffle = _mm_setr_epi8(0, 2, 4, 6, 8, 10, 12, 14, 1, 1, 1, 1, 1, 1, 1, 1);
__m128i mask = _mm_setr_epi16(0x00FF, 0x00FF, 0x00FF, 0x00FF, 0, 0, 0, 0);
__m128i lowmask = _mm_setr_epi16(0, 0, 0, 0, 0x00FF, 0x00FF, 0x00FF, 0x00FF);
__m128i yrf = _mm_set1_epi16(23);
__m128i ygf = _mm_set1_epi16(79);
__m128i ybf = _mm_set1_epi16(8);
__m128i yadd = _mm_set1_epi16(2048);
while(dst_len >= 16){
//Load first 4 pixels
rgb = _mm_lddqu_si128((__m128i const*) src);
src += 12;
rgb = _mm_shuffle_epi8(rgb, inshuffle);
//BB BB BB BB GR GR GR GR
r = _mm_and_si128(rgb, mask);
rgb = _mm_bsrli_si128(rgb, 1);
//0B BB BB BB BG RG RG RG
g = _mm_and_si128(rgb, mask);
rgb = _mm_bsrli_si128(rgb, 7);
//00 00 00 00 BB BB BB BB
b = _mm_and_si128(rgb, mask);
//Load next 4 pixels
rgb = _mm_lddqu_si128((__m128i const*) src);
src += 12;
rgb = _mm_shuffle_epi8(rgb, inshuffle);
//BB BB BB BB GR GR GR GR
b = _mm_or_si128(b, _mm_and_si128(rgb, lowmask));
rgb = _mm_bslli_si128(rgb, 7);
//BG RG RG RG R0 00 00 00
g = _mm_or_si128(g, _mm_and_si128(rgb, lowmask));
rgb = _mm_bslli_si128(rgb, 1);
//GR GR GR GR 00 00 00 00
r = _mm_or_si128(r, _mm_and_si128(rgb, lowmask));
//Compute Y values
y = _mm_adds_epi16(yadd, _mm_mullo_epi16(r, yrf));
y = _mm_adds_epi16(y, _mm_mullo_epi16(g, ygf));
y = _mm_adds_epi16(y, _mm_mullo_epi16(b, ybf));
y = _mm_srli_epi16(y, 7);
y = _mm_shuffle_epi8(y, outshuffle);
_mm_storeu_si128((__m128i *) dst, y);
dst += 8;
dst_len -= 8;
}
#endif
//copy last few pixels
register int ri, gi, bi;
register int y1, y2;
register uint16_t *d = (uint16_t *)(void *) dst;
while (dst_len >= 2) {
ri = *(src++);
gi = *(src++);
bi = *(src++);
y1 = 11993 * ri + 40239 * gi + 4063 * bi + (1<<20);
ri = *(src++);
gi = *(src++);
bi = *(src++);
y2 = 11993 * ri + 40239 * gi + 4063 * bi + (1<<20);
*d++ = (min(max(y2, 0), (1<<24)-1) >> 16) << 8 |
(min(max(y1, 0), (1<<24)-1) >> 16);
dst_len -= 2;
}
}
/**
* @brief Converts BGR to UYVY.
* Uses full scale Rec. 601 YUV (aka JPEG)
* @copydetails vc_copylinev210
*/
void vc_copylineBGRtoUYVY(unsigned char *dst, const unsigned char *src, int dst_len)
{
vc_copylineToUYVY709(dst, src, dst_len, 2, 1, 0, 3);
}
/**
* @brief Converts RGBA to UYVY.
* Uses full scale Rec. 601 YUV (aka JPEG)
* @copydetails vc_copylinev210
*/
void vc_copylineRGBAtoUYVY(unsigned char *dst, const unsigned char *src, int dst_len)
{
vc_copylineToUYVY709(dst, src, dst_len, 0, 1, 2, 4);
}
/**
* Converts BGR to RGB.
* @copydetails vc_copylinev210
*/
void vc_copylineBGRtoRGB(unsigned char *dst, const unsigned char *src, int dst_len, int rshift, int gshift, int bshift)
{
UNUSED(rshift);
UNUSED(gshift);
UNUSED(bshift);
vc_copylineRGB(dst, src, dst_len, 16, 8, 0);
}
/**
* @brief Converts DPX10 to RGBA
* @copydetails vc_copyliner10k
*/
void
vc_copylineDPX10toRGBA(unsigned char *dst, const unsigned char *src, int dst_len, int rshift, int gshift, int bshift)
{
register const unsigned int *in = (const unsigned int *)(const void *) src;
register unsigned int *out = (unsigned int *)(void *) dst;
register int r,g,b;
while (dst_len >= 4) {
register unsigned int val = *in;
r = val >> 24;
g = 0xff & (val >> 14);
b = 0xff & (val >> 4);
*out++ = (r << rshift) | (g << gshift) | (b << bshift);
++in;
dst_len -= 4;
}
}
/**
* @brief Converts DPX10 to RGB.
* @copydetails vc_copylinev210
*/
void
vc_copylineDPX10toRGB(unsigned char *dst, const unsigned char *src, int dst_len)
{
register const unsigned int *in = (const unsigned int *)(const void *) src;
register unsigned int *out = (unsigned int *)(void *) dst;
register int r1,g1,b1,r2,g2,b2;
while (dst_len >= 12) {
register unsigned int val;
val = *in++;
r1 = val >> 24;
g1 = 0xff & (val >> 14);
b1 = 0xff & (val >> 4);
val = *in++;
r2 = val >> 24;
g2 = 0xff & (val >> 14);
b2 = 0xff & (val >> 4);
*out++ = r1 | g1 << 8 | b1 << 16 | r2 << 24;
val = *in++;
r1 = val >> 24;
g1 = 0xff & (val >> 14);
b1 = 0xff & (val >> 4);
*out++ = g2 | b2 << 8 | r1 << 16 | g1 << 24;
val = *in++;
r2 = val >> 24;
g2 = 0xff & (val >> 14);
b2 = 0xff & (val >> 4);
*out++ = b1 | r2 << 8 | g2 << 16 | b2 << 24;
dst_len -= 12;
}
if (dst_len != 0) {
/// @todo
log_msg(LOG_LEVEL_WARNING, "TODO: incomplete %s implementation!\n", __func__);
}
}
struct decoder_item {
decoder_t decoder;
codec_t in;
codec_t out;
bool slow;
};
static const struct decoder_item decoders[] = {
{ (decoder_t) vc_copylineDVS10, DVS10, UYVY, false },
{ (decoder_t) vc_copylinev210, v210, UYVY, false },
{ (decoder_t) vc_copylineYUYV, YUYV, UYVY, false },
{ (decoder_t) vc_copyliner10k, R10k, RGBA, false },
{ vc_copylineRGBA, RGBA, RGBA, false },
{ (decoder_t) vc_copylineDVS10toV210, DVS10, v210, false },
{ (decoder_t) vc_copylineRGBAtoRGB, RGBA, RGB, false },
{ (decoder_t) vc_copylineRGBtoRGBA, RGB, RGBA, false },
{ (decoder_t) vc_copylineRGBtoUYVY, RGB, UYVY, true },
{ (decoder_t) vc_copylineUYVYtoRGB, UYVY, RGB, true },
{ (decoder_t) vc_copylineBGRtoUYVY, BGR, UYVY, true },
{ (decoder_t) vc_copylineRGBAtoUYVY, RGBA, UYVY, true },
{ (decoder_t) vc_copylineBGRtoRGB, BGR, RGB, false },
{ (decoder_t) vc_copylineDPX10toRGBA, DPX10, RGBA, false },
{ (decoder_t) vc_copylineDPX10toRGB, DPX10, RGB, false },
{ vc_copylineRGB, RGB, RGB, false },
};
/**
* Returns line decoder for specifiedn input and output codec.
*/
decoder_t get_decoder_from_to(codec_t in, codec_t out, bool slow)
{
for (unsigned int i = 0; i < sizeof(decoders)/sizeof(struct decoder_item); ++i) {
if (decoders[i].in == in && decoders[i].out == out &&
(decoders[i].slow == false || slow == true)) {
return decoders[i].decoder;
}
}
if (in == out)
return (decoder_t) memcpy;
return NULL;
}
/**
* Tries to find specified codec in set of video codecs.
* The set must by ended by VIDEO_CODEC_NONE.
*/
bool codec_is_in_set(codec_t codec, codec_t *set)
{
assert (codec != VIDEO_CODEC_NONE);
assert (set != NULL);
while (*set != VIDEO_CODEC_NONE) {
if (*(set++) == codec)
return true;
}
return false;
}
bool clear_video_buffer(unsigned char *data, size_t linesize, size_t pitch, size_t height, codec_t color_spec)
{
uint32_t pattern[4];
switch (color_spec) {
case BGR:
case RGB:
case RGBA:
memset(pattern, 0, sizeof(pattern));
break;
case UYVY:
for (int i = 0; i < 4; i++) {
pattern[i] = 0x00800080;
}
break;
case v210:
pattern[0] = 0x20000200;
pattern[1] = 0x00080000;
pattern[2] = 0x20000200;
pattern[3] = 0x00080000;
break;
#ifdef HWACC_VDPAU
case HW_VDPAU:
hw_vdpau_frame_init((hw_vdpau_frame *) data);
return true;
#endif
default:
return false;
}
for (size_t y = 0; y < height; ++y) {
uintptr_t i;
for( i = 0; i < (linesize & (~15)); i+=16)
{
memcpy(data + i, pattern, 16);
}
for( ; i < linesize; i++ )
{
((char*)data)[i] = ((char*)pattern)[i&15];
}
data += pitch;
}
return true;
}
Reference in New Issue View Git Blame Copy Permalink