mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2024-10-03 01:02:19 +00:00
Code Issues 1 Projects Releases Wiki Activity

videomixer: Refactor processing functions

Browse source 
This allows easier plugging of optimized processing functions
in the future, like for SSE or AltiVec.
This commit is contained in:
Sebastian Dröge 2010-01-04 10:24:45 +01:00
parent 5975b01b01
commit 2950262186
10 changed files with 897 additions and 1239 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
gst/videomixer/Makefile.am
View file

@ -1,10 +1,10 @@
plugin_LTLIBRARIES = libgstvideomixer.la
libgstvideomixer_la_SOURCES = videomixer.c blend_ayuv.c blend_bgra.c blend_i420.c blend_rgb.c
libgstvideomixer_la_SOURCES = videomixer.c blend.c
libgstvideomixer_la_CFLAGS = $(GST_CFLAGS) $(GST_BASE_CFLAGS) $(GST_CONTROLLER_CFLAGS) $(GST_PLUGINS_BASE_CFLAGS) $(LIBOIL_CFLAGS)
libgstvideomixer_la_LIBADD = $(GST_LIBS) $(GST_BASE_LIBS) $(GST_CONTROLLER_LIBS) $(GST_PLUGINS_BASE_LIBS) -lgstvideo-@GST_MAJORMINOR@ $(LIBOIL_LIBS)
libgstvideomixer_la_LDFLAGS = $(GST_PLUGIN_LDFLAGS)
libgstvideomixer_la_LIBTOOLFLAGS = --tag=disable-static
# headers we need but don't want installed
noinst_HEADERS = videomixer.h videomixerpad.h
noinst_HEADERS = videomixer.h videomixerpad.h blend.h blend_mmx.h

621
gst/videomixer/blend.c Normal file
View file

@ -0,0 +1,621 @@
/*
* Copyright (C) 2004 Wim Taymans <wim@fluendo.com>
* Copyright (C) 2006 Mindfruit Bv.
* Author: Sjoerd Simons <sjoerd@luon.net>
* Author: Alex Ugarte <alexugarte@gmail.com>
* Copyright (C) 2009 Alex Ugarte <augarte@vicomtech.org>
* Copyright (C) 2009 Sebastian Dröge <sebastian.droege@collabora.co.uk>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "blend.h"
#include <liboil/liboil.h>
#include <liboil/liboilcpu.h>
#include <liboil/liboilfunction.h>
#include <string.h>
#ifdef HAVE_GCC_ASM
#if defined(HAVE_CPU_I386) || defined(HAVE_CPU_X86_64)
#define BUILD_X86_ASM
#define GENERIC
#include "blend_mmx.h"
#undef GENERIC
#endif
#endif
/* Below are the implementations of everything */
#define BLEND(D,S,alpha) (((D) * (255 - (alpha)) + (S) * (alpha)) >> 8)
inline static void
_blend_u8_c (guint8 * dest, const guint8 * src,
gint src_stride, gint dest_stride, gint src_width, gint src_height,
gint dest_width, gint b_alpha)
{
gint i, j;
gint src_add = src_stride - src_width;
gint dest_add = dest_stride - dest_width;
for (i = 0; i < src_height; i++) {
for (j = 0; j < src_width; j++) {
*dest = BLEND (*dest, *src, b_alpha);
dest++;
src++;
}
src += src_add;
dest += dest_add;
}
}
/* A32 is for AYUV, ARGB and BGRA */
#define BLEND_A32(name, LOOP) \
static void \
blend_##name (const guint8 * src, gint xpos, gint ypos, \
gint src_width, gint src_height, gdouble src_alpha, \
guint8 * dest, gint dest_width, gint dest_height) \
{ \
guint s_alpha; \
gint src_stride, dest_stride; \
\
src_stride = src_width * 4; \
dest_stride = dest_width * 4; \
\
s_alpha = CLAMP ((gint) (src_alpha * 256), 0, 256); \
\
/* If it's completely transparent... we just return */ \
if (G_UNLIKELY (s_alpha == 0)) \
return; \
\
/* adjust src pointers for negative sizes */ \
if (xpos < 0) { \
src += -xpos * 4; \
src_width -= -xpos; \
xpos = 0; \
} \
if (ypos < 0) { \
src += -ypos * src_stride; \
src_height -= -ypos; \
ypos = 0; \
} \
/* adjust width/height if the src is bigger than dest */ \
if (xpos + src_width > dest_width) { \
src_width = dest_width - xpos; \
} \
if (ypos + src_height > dest_height) { \
src_height = dest_height - ypos; \
} \
\
dest = dest + 4 * xpos + (ypos * dest_stride); \
\
LOOP (dest, src, src_height, src_width, src_stride, dest_stride, s_alpha); \
}
#define BLEND_A32_LOOP_C(name, A, C1, C2, C3) \
static inline void \
_blend_loop_##name##_c (guint8 *dest, const guint8 *src, gint src_height, gint src_width, gint src_stride, gint dest_stride, guint s_alpha) { \
gint i, j; \
gint alpha; \
gint src_add = src_stride - (4 * src_width); \
gint dest_add = dest_stride - (4 * src_width); \
\
for (i = 0; i < src_height; i++) { \
for (j = 0; j < src_width; j++) { \
alpha = (src[A] * s_alpha) >> 8; \
dest[A] = 0xff; \
dest[C1] = BLEND(dest[C1], src[C1], alpha); \
dest[C2] = BLEND(dest[C2], src[C2], alpha); \
dest[C3] = BLEND(dest[C3], src[C3], alpha); \
\
src += 4; \
dest += 4; \
} \
src += src_add; \
dest += dest_add; \
} \
}
BLEND_A32_LOOP_C (argb, 0, 1, 2, 3);
BLEND_A32_LOOP_C (bgra, 3, 2, 1, 0);
BLEND_A32 (argb_c, _blend_loop_argb_c);
BLEND_A32 (bgra_c, _blend_loop_bgra_c);
#define A32_CHECKER_C(name, RGB, A, C1, C2, C3) \
static void \
fill_checker_##name##_c (guint8 * dest, gint width, gint height) \
{ \
gint i, j; \
gint val; \
static const gint tab[] = { 80, 160, 80, 160 }; \
\
if (!RGB) { \
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[A] = 0xff; \
dest[C1] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; \
dest[C2] = 128; \
dest[C3] = 128; \
} \
} \
} else { \
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
val = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; \
dest[A] = 0xFF; \
dest[C1] = val; \
dest[C2] = val; \
dest[C3] = val; \
dest += 4; \
} \
} \
} \
}
A32_CHECKER_C (argb, TRUE, 0, 1, 2, 3);
A32_CHECKER_C (bgra, TRUE, 3, 2, 1, 0);
A32_CHECKER_C (ayuv, FALSE, 0, 1, 2, 3);
#define YUV_TO_R(Y,U,V) (CLAMP (1.164 * (Y - 16) + 1.596 * (V - 128), 0, 255))
#define YUV_TO_G(Y,U,V) (CLAMP (1.164 * (Y - 16) - 0.813 * (V - 128) - 0.391 * (U - 128), 0, 255))
#define YUV_TO_B(Y,U,V) (CLAMP (1.164 * (Y - 16) + 2.018 * (U - 128), 0, 255))
#define A32_COLOR(name, RGB, LOOP) \
static void \
fill_color_##name (guint8 * dest, gint width, gint height, gint Y, gint U, gint V) \
{ \
gint c1, c2, c3; \
\
if (RGB) { \
c1 = YUV_TO_R (Y, U, V); \
c2 = YUV_TO_G (Y, U, V); \
c3 = YUV_TO_B (Y, U, V); \
} else { \
c1 = Y; \
c2 = U; \
c3 = V; \
} \
LOOP (dest, height, width, c1, c2, c3); \
}
#define A32_COLOR_LOOP_C(name, A, C1, C2, C3) \
static inline void \
_fill_color_loop_##name##_c (guint8 *dest, gint height, gint width, gint c1, gint c2, gint c3) { \
gint i, j; \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[A] = 0xff; \
dest[C1] = c1; \
dest[C2] = c2; \
dest[C3] = c3; \
} \
} \
}
A32_COLOR_LOOP_C (ac1c2c3, 0, 1, 2, 3);
A32_COLOR_LOOP_C (c3c2c1a, 3, 2, 1, 0);
A32_COLOR (argb_c, TRUE, _fill_color_loop_ac1c2c3_c);
A32_COLOR (bgra_c, TRUE, _fill_color_loop_c3c2c1a_c);
A32_COLOR (ayuv_c, FALSE, _fill_color_loop_ac1c2c3_c);
/* I420 */
#define I420_Y_ROWSTRIDE(width) (GST_ROUND_UP_4(width))
#define I420_U_ROWSTRIDE(width) (GST_ROUND_UP_8(width)/2)
#define I420_V_ROWSTRIDE(width) ((GST_ROUND_UP_8(I420_Y_ROWSTRIDE(width)))/2)
#define I420_Y_OFFSET(w,h) (0)
#define I420_U_OFFSET(w,h) (I420_Y_OFFSET(w,h)+(I420_Y_ROWSTRIDE(w)*GST_ROUND_UP_2(h)))
#define I420_V_OFFSET(w,h) (I420_U_OFFSET(w,h)+(I420_U_ROWSTRIDE(w)*GST_ROUND_UP_2(h)/2))
#define I420_BLEND(name,MEMCPY,BLENDLOOP) \
inline static void \
_blend_i420_##name (const guint8 * src, guint8 * dest, \
gint src_stride, gint dest_stride, gint src_width, gint src_height, \
gint dest_width, gdouble src_alpha) \
{ \
gint i; \
gint b_alpha; \
\
/* If it's completely transparent... we just return */ \
if (G_UNLIKELY (src_alpha == 0.0)) { \
GST_INFO ("Fast copy (alpha == 0.0)"); \
return; \
} \
\
/* If it's completely opaque, we do a fast copy */ \
if (G_UNLIKELY (src_alpha == 1.0)) { \
GST_INFO ("Fast copy (alpha == 1.0)"); \
for (i = 0; i < src_height; i++) { \
MEMCPY (dest, src, src_width); \
src += src_stride; \
dest += dest_stride; \
} \
return; \
} \
\
b_alpha = CLAMP ((gint) (src_alpha * 256), 0, 256); \
\
BLENDLOOP(dest, src, src_stride, dest_stride, src_width, src_height, dest_width, b_alpha); \
} \
\
static void \
blend_i420_##name (const guint8 * src, gint xpos, gint ypos, \
gint src_width, gint src_height, gdouble src_alpha, \
guint8 * dest, gint dest_width, gint dest_height) \
{ \
const guint8 *b_src; \
guint8 *b_dest; \
gint b_src_width = src_width; \
gint b_src_height = src_height; \
gint xoffset = 0; \
gint yoffset = 0; \
\
xpos = GST_ROUND_UP_2 (xpos); \
ypos = GST_ROUND_UP_2 (ypos); \
\
/* adjust src pointers for negative sizes */ \
if (xpos < 0) { \
xoffset = -xpos; \
b_src_width -= -xpos; \
xpos = 0; \
} \
if (ypos < 0) { \
yoffset += -ypos; \
b_src_height -= -ypos; \
ypos = 0; \
} \
/* If x or y offset are larger then the source it's outside of the picture */ \
if (xoffset > src_width || yoffset > src_width) { \
return; \
} \
\
/* adjust width/height if the src is bigger than dest */ \
if (xpos + src_width > dest_width) { \
b_src_width = dest_width - xpos; \
} \
if (ypos + src_height > dest_height) { \
b_src_height = dest_height - ypos; \
} \
if (b_src_width < 0 || b_src_height < 0) { \
return; \
} \
\
/* First mix Y, then U, then V */ \
b_src = src + I420_Y_OFFSET (src_width, src_height); \
b_dest = dest + I420_Y_OFFSET (dest_width, dest_height); \
_blend_i420_##name (b_src + xoffset + yoffset * I420_Y_ROWSTRIDE (src_width), \
b_dest + xpos + ypos * I420_Y_ROWSTRIDE (dest_width), \
I420_Y_ROWSTRIDE (src_width), \
I420_Y_ROWSTRIDE (dest_width), b_src_width, b_src_height, \
dest_width, src_alpha); \
\
b_src = src + I420_U_OFFSET (src_width, src_height); \
b_dest = dest + I420_U_OFFSET (dest_width, dest_height); \
\
_blend_i420_##name (b_src + xoffset / 2 + \
yoffset / 2 * I420_U_ROWSTRIDE (src_width), \
b_dest + xpos / 2 + ypos / 2 * I420_U_ROWSTRIDE (dest_width), \
I420_U_ROWSTRIDE (src_width), I420_U_ROWSTRIDE (dest_width), \
b_src_width / 2, GST_ROUND_UP_2 (b_src_height) / 2, dest_width / 2, \
src_alpha); \
\
b_src = src + I420_V_OFFSET (src_width, src_height); \
b_dest = dest + I420_V_OFFSET (dest_width, dest_height); \
\
_blend_i420_##name (b_src + xoffset / 2 + \
yoffset / 2 * I420_V_ROWSTRIDE (src_width), \
b_dest + xpos / 2 + ypos / 2 * I420_V_ROWSTRIDE (dest_width), \
I420_V_ROWSTRIDE (src_width), I420_V_ROWSTRIDE (dest_width), \
b_src_width / 2, GST_ROUND_UP_2 (b_src_height) / 2, dest_width / 2, \
src_alpha); \
}
#define I420_FILL_CHECKER(name, MEMSET) \
static void \
fill_checker_i420_##name (guint8 * dest, gint width, gint height) \
{ \
gint size; \
gint i, j; \
static const int tab[] = { 80, 160, 80, 160 }; \
guint8 *p = dest; \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
*p++ = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; \
} \
p += I420_Y_ROWSTRIDE (width) - width; \
} \
\
size = (I420_U_ROWSTRIDE (width) * height) / 2; \
MEMSET (dest + I420_U_OFFSET (width, height), 0x80, size); \
\
size = (I420_V_ROWSTRIDE (width) * height) / 2; \
MEMSET (dest + I420_V_OFFSET (width, height), 0x80, size); \
}
#define I420_FILL_COLOR(name,MEMSET) \
static void \
fill_color_i420_##name (guint8 * dest, gint width, gint height, \
gint colY, gint colU, gint colV) \
{ \
gint size; \
\
size = I420_Y_ROWSTRIDE (width) * height; \
MEMSET (dest, colY, size); \
\
size = (I420_U_ROWSTRIDE (width) * height) / 2; \
MEMSET (dest + I420_U_OFFSET (width, height), colU, size); \
\
size = (I420_V_ROWSTRIDE (width) * height) / 2; \
MEMSET (dest + I420_V_OFFSET (width, height), colV, size); \
}
I420_BLEND (c, memcpy, _blend_u8_c);
I420_FILL_CHECKER (c, memset);
I420_FILL_COLOR (c, memset);
/* RGB, BGR, xRGB, xBGR, RGBx, BGRx */
#define RGB_BLEND(name, bpp, MEMCPY, BLENDLOOP) \
static void \
blend_##name (const guint8 * src, gint xpos, gint ypos, \
gint src_width, gint src_height, gdouble src_alpha, \
guint8 * dest, gint dest_width, gint dest_height) \
{ \
gint b_alpha; \
gint i; \
gint src_stride, dest_stride; \
\
src_stride = GST_ROUND_UP_4 (src_width * bpp); \
dest_stride = GST_ROUND_UP_4 (dest_width * bpp); \
\
b_alpha = CLAMP ((gint) (src_alpha * 256), 0, 256); \
\
/* adjust src pointers for negative sizes */ \
if (xpos < 0) { \
src += -xpos * bpp; \
src_width -= -xpos; \
xpos = 0; \
} \
if (ypos < 0) { \
src += -ypos * src_stride; \
src_height -= -ypos; \
ypos = 0; \
} \
/* adjust width/height if the src is bigger than dest */ \
if (xpos + src_width > dest_width) { \
src_width = dest_width - xpos; \
} \
if (ypos + src_height > dest_height) { \
src_height = dest_height - ypos; \
} \
\
dest = dest + bpp * xpos + (ypos * dest_stride); \
/* If it's completely transparent... we just return */ \
if (G_UNLIKELY (src_alpha == 0.0)) { \
GST_INFO ("Fast copy (alpha == 0.0)"); \
return; \
} \
\
/* If it's completely opaque, we do a fast copy */ \
if (G_UNLIKELY (src_alpha == 1.0)) { \
GST_INFO ("Fast copy (alpha == 1.0)"); \
for (i = 0; i < src_height; i++) { \
MEMCPY (dest, src, bpp * src_width); \
src += src_stride; \
dest += dest_stride; \
} \
return; \
} \
\
BLENDLOOP(dest, src, src_stride, dest_stride, bpp * src_width, src_height, bpp * dest_width, b_alpha); \
}
#define RGB_FILL_CHECKER_C(name, bpp, r, g, b) \
static void \
fill_checker_##name##_c (guint8 * dest, gint width, gint height) \
{ \
gint i, j; \
static const int tab[] = { 80, 160, 80, 160 }; \
gint dest_add = GST_ROUND_UP_4 (width * bpp) - width * bpp; \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[r] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* red */ \
dest[g] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* green */ \
dest[b] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* blue */ \
dest += bpp; \
} \
dest += dest_add; \
} \
}
#define RGB_FILL_COLOR(name, bpp, MEMSET_RGB) \
static void \
fill_color_##name (guint8 * dest, gint width, gint height, \
gint colY, gint colU, gint colV) \
{ \
gint red, green, blue; \
gint i; \
gint dest_stride = GST_ROUND_UP_4 (width * bpp); \
\
red = CLAMP (1.164 * (colY - 16) + 1.596 * (colV - 128), 0, 255); \
green = \
CLAMP (1.164 * (colY - 16) - 0.813 * (colV - 128) - 0.391 * (colU - 128), \
0, 255); \
blue = CLAMP (1.164 * (colY - 16) + 2.018 * (colU - 128), 0, 255); \
\
for (i = 0; i < height; i++) { \
MEMSET_RGB (dest, red, green, blue, width); \
dest += dest_stride; \
} \
}
#define MEMSET_RGB_C(name, bpp, r, g, b) \
static inline void \
_memset_##name##_c (guint8* dest, gint red, gint green, gint blue, gint width) { \
gint j; \
\
for (j = 0; j < width; j++) { \
dest[r] = red; \
dest[g] = green; \
dest[b] = blue; \
dest += bpp; \
} \
}
RGB_BLEND (rgb_c, 3, memcpy, _blend_u8_c);
RGB_FILL_CHECKER_C (rgb, 3, 0, 1, 2);
MEMSET_RGB_C (rgb, 3, 0, 1, 2);
RGB_FILL_COLOR (rgb_c, 3, _memset_rgb_c);
MEMSET_RGB_C (bgr, 3, 2, 1, 0);
RGB_FILL_COLOR (bgr_c, 3, _memset_bgr_c);
RGB_BLEND (xrgb_c, 4, memcpy, _blend_u8_c);
RGB_FILL_CHECKER_C (xrgb, 4, 1, 2, 3);
MEMSET_RGB_C (xrgb, 4, 1, 2, 3);
RGB_FILL_COLOR (xrgb_c, 4, _memset_xrgb_c);
MEMSET_RGB_C (xbgr, 4, 3, 2, 1);
RGB_FILL_COLOR (xbgr_c, 4, _memset_xbgr_c);
MEMSET_RGB_C (rgbx, 4, 0, 1, 2);
RGB_FILL_COLOR (rgbx_c, 4, _memset_rgbx_c);
MEMSET_RGB_C (bgrx, 4, 2, 1, 0);
RGB_FILL_COLOR (bgrx_c, 4, _memset_bgrx_c);
/* MMX Implementations */
#ifdef BUILD_X86_ASM
#define A32
#define NAME_BLEND _blend_loop_argb_mmx
#define NAME_FILL_COLOR _fill_color_loop_argb_mmx
#define A_OFF 0
#define C1_OFF 8
#define C2_OFF 16
#define C3_OFF 24
#include "blend_mmx.h"
#undef NAME_BLEND
#undef NAME_FILL_COLOR
#undef A_OFF
#undef C1_OFF
#undef C2_OFF
#undef C3_OFF
#define NAME_BLEND _blend_loop_bgra_mmx
#define NAME_FILL_COLOR _fill_color_loop_bgra_mmx
#define A_OFF 24
#define C1_OFF 16
#define C2_OFF 8
#define C3_OFF 0
#include "blend_mmx.h"
#undef NAME_BLEND
#undef NAME_FILL_COLOR
#undef A_OFF
#undef C1_OFF
#undef C2_OFF
#undef C3_OFF
#undef A32
BLEND_A32 (argb_mmx, _blend_loop_argb_mmx);
BLEND_A32 (bgra_mmx, _blend_loop_bgra_mmx);
A32_COLOR (argb_mmx, TRUE, _fill_color_loop_argb_mmx);
A32_COLOR (bgra_mmx, TRUE, _fill_color_loop_bgra_mmx);
A32_COLOR (ayuv_mmx, FALSE, _fill_color_loop_argb_mmx);
#endif
/* Init function */
BlendFunction gst_video_mixer_blend_argb;
BlendFunction gst_video_mixer_blend_bgra;
/* AYUV is equal to ARGB */
BlendFunction gst_video_mixer_blend_i420;
BlendFunction gst_video_mixer_blend_rgb;
/* BGR is equal to RGB */
BlendFunction gst_video_mixer_blend_rgbx;
/* BGRx, xRGB, xBGR are equal to RGBx */
FillCheckerFunction gst_video_mixer_fill_checker_argb;
FillCheckerFunction gst_video_mixer_fill_checker_bgra;
FillCheckerFunction gst_video_mixer_fill_checker_ayuv;
FillCheckerFunction gst_video_mixer_fill_checker_i420;
FillCheckerFunction gst_video_mixer_fill_checker_rgb;
/* BGR is equal to RGB */
FillCheckerFunction gst_video_mixer_fill_checker_xrgb;
/* BGRx, xRGB, xBGR are equal to RGBx */
FillColorFunction gst_video_mixer_fill_color_argb;
FillColorFunction gst_video_mixer_fill_color_bgra;
FillColorFunction gst_video_mixer_fill_color_ayuv;
FillColorFunction gst_video_mixer_fill_color_i420;
FillColorFunction gst_video_mixer_fill_color_rgb;
FillColorFunction gst_video_mixer_fill_color_bgr;
FillColorFunction gst_video_mixer_fill_color_xrgb;
FillColorFunction gst_video_mixer_fill_color_xbgr;
FillColorFunction gst_video_mixer_fill_color_rgbx;
FillColorFunction gst_video_mixer_fill_color_bgrx;
void
gst_video_mixer_init_blend (void)
{
guint cpu_flags;
oil_init ();
cpu_flags = oil_cpu_get_flags ();
gst_video_mixer_blend_argb = blend_argb_c;
gst_video_mixer_blend_bgra = blend_bgra_c;
gst_video_mixer_blend_i420 = blend_i420_c;
gst_video_mixer_blend_rgb = blend_rgb_c;
gst_video_mixer_blend_xrgb = blend_xrgb_c;
gst_video_mixer_fill_checker_argb = fill_checker_argb_c;
gst_video_mixer_fill_checker_bgra = fill_checker_bgra_c;
gst_video_mixer_fill_checker_ayuv = fill_checker_ayuv_c;
gst_video_mixer_fill_checker_i420 = fill_checker_i420_c;
gst_video_mixer_fill_checker_rgb = fill_checker_rgb_c;
gst_video_mixer_fill_checker_xrgb = fill_checker_xrgb_c;
gst_video_mixer_fill_color_argb = fill_color_argb_c;
gst_video_mixer_fill_color_bgra = fill_color_bgra_c;
gst_video_mixer_fill_color_ayuv = fill_color_ayuv_c;
gst_video_mixer_fill_color_i420 = fill_color_i420_c;
gst_video_mixer_fill_color_rgb = fill_color_rgb_c;
gst_video_mixer_fill_color_bgr = fill_color_bgr_c;
gst_video_mixer_fill_color_xrgb = fill_color_xrgb_c;
gst_video_mixer_fill_color_xbgr = fill_color_xbgr_c;
gst_video_mixer_fill_color_rgbx = fill_color_rgbx_c;
gst_video_mixer_fill_color_bgrx = fill_color_bgrx_c;
#ifdef BUILD_X86_ASM
if (cpu_flags & OIL_IMPL_FLAG_MMX) {
gst_video_mixer_blend_argb = blend_argb_mmx;
gst_video_mixer_blend_bgra = blend_bgra_mmx;
gst_video_mixer_fill_color_argb = fill_color_argb_mmx;
gst_video_mixer_fill_color_bgra = fill_color_bgra_mmx;
gst_video_mixer_fill_color_ayuv = fill_color_ayuv_mmx;
}
#endif
}

64
gst/videomixer/blend.h Normal file
View file

@ -0,0 +1,64 @@
/*
* Copyright (C) 2009 Sebastian Dröge <sebastian.droege@collabora.co.uk>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#ifndef __BLEND_H__
#define __BLEND_H__
#include <gst/gst.h>
typedef void (*BlendFunction) (const guint8 * src, gint xpos, gint ypos, gint src_width, gint src_height, gdouble src_alpha, guint8 * dest, gint dest_width, gint dest_height);
typedef void (*FillCheckerFunction) (guint8 * dest, gint width, gint height);
typedef void (*FillColorFunction) (guint8 * dest, gint width, gint height, gint c1, gint c2, gint c3);
extern BlendFunction gst_video_mixer_blend_argb;
extern BlendFunction gst_video_mixer_blend_bgra;
#define gst_video_mixer_blend_ayuv gst_video_mixer_blend_argb
extern BlendFunction gst_video_mixer_blend_i420;
extern BlendFunction gst_video_mixer_blend_rgb;
#define gst_video_mixer_blend_bgr gst_video_mixer_blend_rgb
extern BlendFunction gst_video_mixer_blend_rgbx;
#define gst_video_mixer_blend_bgrx gst_video_mixer_blend_rgbx
#define gst_video_mixer_blend_xrgb gst_video_mixer_blend_rgbx
#define gst_video_mixer_blend_xbgr gst_video_mixer_blend_rgbx
extern FillCheckerFunction gst_video_mixer_fill_checker_argb;
extern FillCheckerFunction gst_video_mixer_fill_checker_bgra;
extern FillCheckerFunction gst_video_mixer_fill_checker_ayuv;
extern FillCheckerFunction gst_video_mixer_fill_checker_i420;
extern FillCheckerFunction gst_video_mixer_fill_checker_rgb;
#define gst_video_mixer_fill_checker_bgr gst_video_mixer_fill_checker_rgb
extern FillCheckerFunction gst_video_mixer_fill_checker_rgbx;
#define gst_video_mixer_fill_checker_bgrx gst_video_mixer_fill_checker_rgbx
#define gst_video_mixer_fill_checker_xrgb gst_video_mixer_fill_checker_rgbx
#define gst_video_mixer_fill_checker_xbgr gst_video_mixer_fill_checker_rgbx
extern FillColorFunction gst_video_mixer_fill_color_argb;
extern FillColorFunction gst_video_mixer_fill_color_bgra;
extern FillColorFunction gst_video_mixer_fill_color_ayuv;
extern FillColorFunction gst_video_mixer_fill_color_i420;
extern FillColorFunction gst_video_mixer_fill_color_rgb;
extern FillColorFunction gst_video_mixer_fill_color_bgr;
extern FillColorFunction gst_video_mixer_fill_color_xrgb;
extern FillColorFunction gst_video_mixer_fill_color_xbgr;
extern FillColorFunction gst_video_mixer_fill_color_rgbx;
extern FillColorFunction gst_video_mixer_fill_color_bgrx;
void gst_video_mixer_init_blend (void);
#endif /* __BLEND_H__ */

477
gst/videomixer/blend_ayuv.c
View file

@ -1,477 +0,0 @@
/*
* Copyright (C) 2004 Wim Taymans <wim@fluendo.com>
* Copyright (C) 2009 Sebastian Dröge <sebastian.droege@collabora.co.uk>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <gst/gst.h>
#ifdef HAVE_GCC_ASM
#if defined(HAVE_CPU_I386) || defined(HAVE_CPU_X86_64)
#define BUILD_X86_ASM
#endif
#endif
#define BLEND_NORMAL(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = ((Y1*(255-alpha))+(Y2*alpha))>>8; \
U = ((U1*(255-alpha))+(U2*alpha))>>8; \
V = ((V1*(255-alpha))+(V2*alpha))>>8;
#define BLEND_ADD(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = Y1+((Y2*alpha)>>8); \
U = U1+(((127*(255-alpha)+(U2*alpha)))>>8)-127; \
V = V1+(((127*(255-alpha)+(V2*alpha)))>>8)-127; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} \
U = MIN (U,255); \
V = MIN (V,255);
#define BLEND_SUBTRACT(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = Y1-((Y2*alpha)>>8); \
U = U1+(((127*(255-alpha)+(U2*alpha)))>>8)-127; \
V = V1+(((127*(255-alpha)+(V2*alpha)))>>8)-127; \
if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
}
#define BLEND_DARKEN(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
if (Y1 < Y2) { \
Y = Y1; U = U1; V = V1; \
} \
else { \
Y = ((Y1*(255-alpha))+(Y2*alpha))>>8; \
U = ((U1*(255-alpha))+(U2*alpha))>>8; \
V = ((V1*(255-alpha))+(V2*alpha))>>8; \
}
#define BLEND_LIGHTEN(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
if (Y1 > Y2) { \
Y = Y1; U = U1; V = V1; \
} \
else { \
Y = ((Y1*(255-alpha))+(Y2*alpha))>>8; \
U = ((U1*(255-alpha))+(U2*alpha))>>8; \
V = ((V1*(255-alpha))+(V2*alpha))>>8; \
}
#define BLEND_MULTIPLY(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = (Y1*(256*(255-alpha) +(Y2*alpha)))>>16; \
U = ((U1*(255-alpha)*256)+(alpha*(U1*Y2+128*(256-Y2))))>>16; \
V = ((V1*(255-alpha)*256)+(alpha*(V1*Y2+128*(256-Y2))))>>16;
#define BLEND_DIFFERENCE(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = ABS((gint)Y1-(gint)Y2)+127; \
U = ABS((gint)U1-(gint)U2)+127; \
V = ABS((gint)V1-(gint)V2)+127; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
U = CLAMP(U, 0, 255); \
V = CLAMP(V, 0, 255);
#define BLEND_EXCLUSION(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = ((gint)(Y1^0xff)*Y2+(gint)(Y2^0xff)*Y1)>>8; \
U = ((gint)(U1^0xff)*Y2+(gint)(Y2^0xff)*U1)>>8; \
V = ((gint)(V1^0xff)*Y2+(gint)(Y2^0xff)*V1)>>8; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
U = CLAMP(U, 0, 255); \
V = CLAMP(V, 0, 255);
#define BLEND_SOFTLIGHT(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = (gint)Y1+(gint)Y2 - 127; \
U = (gint)U1+(gint)U2 - 127; \
V = (gint)V1+(gint)V2 - 127; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
#define BLEND_HARDLIGHT(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = (gint)Y1+(gint)Y2*2 - 255; \
U = (gint)U1+(gint)U2 - 127; \
V = (gint)V1+(gint)V2 - 127; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
#define BLEND_MODE BLEND_NORMAL
#if 0
#define BLEND_MODE BLEND_NORMAL
#define BLEND_MODE BLEND_ADD
#define BLEND_MODE BLEND_SUBTRACT
#define BLEND_MODE BLEND_LIGHTEN
#define BLEND_MODE BLEND_DARKEN
#define BLEND_MODE BLEND_MULTIPLY
#define BLEND_MODE BLEND_DIFFERENCE
#define BLEND_MODE BLEND_EXCLUSION
#define BLEND_MODE BLEND_SOFTLIGHT
#define BLEND_MODE BLEND_HARDLIGHT
#endif
/* note that this function does packing conversion and blending at the
* same time */
void
gst_videomixer_blend_ayuv_ayuv (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height)
{
guint s_alpha, alpha;
gint i, j;
gint src_stride, dest_stride;
gint src_add, dest_add;
src_stride = src_width * 4;
dest_stride = dest_width * 4;
s_alpha = CLAMP ((gint) (src_alpha * 256), 0, 256);
//g_print ("%f %d\n", src_alpha, s_alpha);
/* adjust src pointers for negative sizes */
if (xpos < 0) {
src += -xpos * 4;
src_width -= -xpos;
xpos = 0;
}
if (ypos < 0) {
src += -ypos * src_stride;
src_height -= -ypos;
ypos = 0;
}
/* adjust width/height if the src is bigger than dest */
if (xpos + src_width > dest_width) {
src_width = dest_width - xpos;
}
if (ypos + src_height > dest_height) {
src_height = dest_height - ypos;
}
src_add = src_stride - (4 * src_width);
dest_add = dest_stride - (4 * src_width);
dest = dest + 4 * xpos + (ypos * dest_stride);
for (i = 0; i < src_height; i++) {
for (j = 0; j < src_width; j++) {
#if 0
gint Y, U, V;
alpha = (src[0] * s_alpha) >> 8;
Y = dest[1];
U = dest[2];
V = dest[3];
dest[0] = 0xff;
dest[1] = (((src[1] - Y) * alpha) >> 8) + Y;
dest[2] = (((src[2] - U) * alpha) >> 8) + U;
dest[3] = (((src[3] - V) * alpha) >> 8) + V;
#else
gint Y, U, V;
alpha = (src[0] * s_alpha) >> 8;
BLEND_MODE (dest[1], dest[2], dest[3], src[1], src[2], src[3],
alpha, Y, U, V);
dest[0] = 0xff;
dest[1] = Y;
dest[2] = U;
dest[3] = V;
#endif
src += 4;
dest += 4;
}
src += src_add;
dest += dest_add;
}
}
#undef BLEND_MODE
#ifdef BUILD_X86_ASM
void
gst_videomixer_blend_ayuv_ayuv_mmx (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height)
{
gint s_alpha;
gint i;
gint src_stride, dest_stride;
gint src_add, dest_add;
src_stride = src_width * 4;
dest_stride = dest_width * 4;
s_alpha = CLAMP ((gint) (src_alpha * 256), 0, 256);
/* adjust src pointers for negative sizes */
if (xpos < 0) {
src += -xpos * 4;
src_width -= -xpos;
xpos = 0;
}
if (ypos < 0) {
src += -ypos * src_stride;
src_height -= -ypos;
ypos = 0;
}
/* adjust width/height if the src is bigger than dest */
if (xpos + src_width > dest_width) {
src_width = dest_width - xpos;
}
if (ypos + src_height > dest_height) {
src_height = dest_height - ypos;
}
src_add = src_stride - (4 * src_width);
dest_add = dest_stride - (4 * src_width);
dest = dest + 4 * xpos + (ypos * dest_stride);
for (i = 0; i < src_height; i++) {
gulong old_ebx;
/* (P1 * (256 - A) + (P2 * A)) / 256
* => (P1 * 256 - P1 * A + P2 * A) / 256
* => (P1 * 256 + A * (P2 - P1) / 256
* => P1 + (A * (P2 - P1)) / 256
*/
/* *INDENT-OFF* */
__asm__ __volatile__ (
" movl %%ebx , %6 \n\t"
" pcmpeqd %%mm5 , %%mm5 \n\t" /* mm5 = 0xffff... */
" psrld $24 , %%mm5 \n\t" /* mm5 = 00 00 00 ff 00 00 00 0ff, selector for alpha */
" mov %4 , %%eax \n\t" /* eax = s_alpha */
" movd %%eax , %%mm6 \n\t" /* mm6 = s_alpha */
" punpckldq %%mm6 , %%mm6 \n\t" /* mm6 = 00 00 00 aa 00 00 00 aa, alpha scale factor */
" movl %5 , %%ebx \n\t" /* ebx = src_width */
" test $1 , %%ebx \n\t" /* check odd pixel */
" je 1f \n\t"
/* do odd pixel */
" movd (%2) , %%mm2 \n\t" /* mm2 = src, 00 00 00 00 sv su sy sa */
" movd (%3) , %%mm1 \n\t" /* mm1 = dest, 00 00 00 00 dv du dy da */
" movq %%mm2 , %%mm0 \n\t"
" punpcklbw %%mm7 , %%mm2 \n\t" /* mm2 = 00 sv 00 su 00 sy 00 sa */
" pand %%mm5 , %%mm0 \n\t" /* mm0 = 00 00 00 00 00 00 00 sa, get alpha component */
" punpcklbw %%mm7 , %%mm1 \n\t" /* mm1 = 00 dv 00 du 00 dy 00 da */
" pmullw %%mm6 , %%mm0 \n\t" /* mult with scale */
" psubw %%mm1 , %%mm2 \n\t" /* mm2 = mm2 - mm1 */
" punpcklwd %%mm0 , %%mm0 \n\t"
" punpckldq %%mm0 , %%mm0 \n\t" /* mm0 == 00 aa 00 aa 00 aa 00 aa */
" psrlw $8 , %%mm0 \n\t"
" pmullw %%mm0 , %%mm2 \n\t" /* mm2 == a * mm2 */
" psllw $8 , %%mm1 \n\t" /* scale up */
" paddw %%mm1 , %%mm2 \n\t" /* mm2 == mm2 + mm1 */
" psrlw $8 , %%mm2 \n\t" /* scale down */
" por %%mm5 , %%mm2 \n\t" /* set alpha to ff */
" packuswb %%mm2 , %%mm2 \n\t"
" movd %%mm2 , (%3) \n\t" /* dest = mm1 */
" add $4 , %1 \n\t"
" add $4 , %0 \n\t"
"1: \n\t"
" sar $1 , %%ebx \n\t" /* prepare for 2 pixel per loop */
" cmp $0 , %%ebx \n\t"
" je 3f \n\t"
"2: \n\t"
/* do even pixels */
" movq (%2) , %%mm2 \n\t" /* mm2 = src, sv1 su1 sy1 sa1 sv0 su0 sy0 sa0 */
" movq (%3) , %%mm1 \n\t" /* mm1 = dest, dv1 du1 dy1 da1 dv0 du0 dy0 da0 */
" movq %%mm2 , %%mm4 \n\t"
" movq %%mm1 , %%mm3 \n\t"
" movq %%mm2 , %%mm0 \n\t" /* copy for doing the alpha */
" pxor %%mm7 , %%mm7 \n\t"
" punpcklbw %%mm7 , %%mm2 \n\t" /* mm2 = 00 sv0 00 su0 00 sy0 00 sa0 */
" punpckhbw %%mm7 , %%mm4 \n\t" /* mm4 = 00 sv1 00 su1 00 sy1 00 sa1 */
" punpcklbw %%mm7 , %%mm1 \n\t" /* mm1 = 00 dv0 00 du0 00 dy0 00 da0 */
" punpckhbw %%mm7 , %%mm3 \n\t" /* mm2 = 00 dv1 00 du1 00 dy1 00 da1 */
" pand %%mm5 , %%mm0 \n\t" /* mm0 = 00 00 00 sa1 00 00 00 sa0 */
" psubw %%mm1 , %%mm2 \n\t" /* mm2 = mm2 - mm1 */
" pmullw %%mm6 , %%mm0 \n\t" /* mult with scale */
" psubw %%mm3 , %%mm4 \n\t" /* mm4 = mm4 - mm3 */
" psrlw $8 , %%mm0 \n\t" /* scale back */
" movq %%mm0 , %%mm7 \n\t" /* save copy */
" punpcklwd %%mm0 , %%mm0 \n\t" /* mm0 = 00 00 00 00 00 aa0 00 aa0 */
" punpckhwd %%mm7 , %%mm7 \n\t" /* mm7 = 00 00 00 00 00 aa1 00 aa1 */
" punpckldq %%mm0 , %%mm0 \n\t" /* mm0 = 00 aa0 00 aa0 00 aa0 00 aa0 */
" punpckldq %%mm7 , %%mm7 \n\t" /* mm7 = 00 aa1 00 aa1 00 aa1 00 aa1 */
" pmullw %%mm0 , %%mm2 \n\t" /* mm2 == aa * mm2 */
" pmullw %%mm7 , %%mm4 \n\t" /* mm2 == aa * mm2 */
" psllw $8 , %%mm1 \n\t"
" psllw $8 , %%mm3 \n\t"
" paddw %%mm1 , %%mm2 \n\t" /* mm2 == mm2 + mm1 */
" paddw %%mm3 , %%mm4 \n\t" /* mm2 == mm2 + mm1 */
" psrlw $8 , %%mm2 \n\t"
" psrlw $8 , %%mm4 \n\t"
" packuswb %%mm4 , %%mm2 \n\t"
" por %%mm5 , %%mm2 \n\t" /* set alpha to ff */
" movq %%mm2 , (%3) \n\t"
" add $8 , %1 \n\t"
" add $8 , %0 \n\t"
" dec %%ebx \n\t"
" jne 2b \n\t"
"3: \n\t"
" movl %6 , %%ebx \n\t"
:"=r" (src), "=r" (dest)
:"0" (src), "1" (dest), "m" (s_alpha), "m" (src_width), "m" (old_ebx)
:"%eax", "memory"
#ifdef __MMX__
, "mm0", "mm1", "mm2", "mm5", "mm6", "mm7"
#endif
);
/* *INDENT-ON* */
src += src_add;
dest += dest_add;
}
__asm__ __volatile__ ("emms");
}
#endif
/* fill a buffer with a checkerboard pattern */
void
gst_videomixer_fill_ayuv_checker (guint8 * dest, gint width, gint height)
{
gint i, j;
static const int tab[] = { 80, 160, 80, 160 };
for (i = 0; i < height; i++) {
for (j = 0; j < width; j++) {
*dest++ = 0xff;
*dest++ = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)];
*dest++ = 128;
*dest++ = 128;
}
}
}
void
gst_videomixer_fill_ayuv_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV)
{
gint i, j;
for (i = 0; i < height; i++) {
for (j = 0; j < width; j++) {
*dest++ = 0xff;
*dest++ = colY;
*dest++ = colU;
*dest++ = colV;
}
}
}
#ifdef BUILD_X86_ASM
void
gst_videomixer_fill_ayuv_color_mmx (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV)
{
guint64 val;
guint nvals = width * height;
val = (((guint64) 0xff)) | (((guint64) colY) << 8) |
(((guint64) colU) << 16) | (((guint64) colV) << 24);
val = (val << 32) | val;
/* *INDENT-OFF* */
__asm__ __volatile__ (
"cmp $2 , %2 \n\t"
"jb 2f \n\t"
"movq %4 , %%mm0 \n\t"
"1: \n\t"
"movq %%mm0 , (%1) \n\t"
"sub $2 , %0 \n\t"
"add $8 , %1 \n\t"
"cmp $2 , %2 \n\t"
"jae 1b \n\t"
"emms \n\t"
"2: \n\t"
: "=r" (nvals), "=r" (dest)
: "0" (nvals), "1" (dest), "m" (val)
: "memory"
#ifdef __MMX__
, "mm0"
#endif
);
/* *INDENT-ON* */
if (nvals)
GST_WRITE_UINT32_LE (&dest[-4], (guint32) (val & 0xffffffff));
}
#endif

133
gst/videomixer/blend_bgra.c
View file

@ -1,133 +0,0 @@
/*
* Copyright (C) 2009 Alex Ugarte <augarte@vicomtech.org>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#include <gst/gst.h>
#define BLEND_NORMAL(B1,G1,R1,B2,G2,R2,B,G,R,alpha) \
B = ((B1*(255-alpha))+(B2*alpha))>>8; \
G = ((G1*(255-alpha))+(G2*alpha))>>8; \
R = ((R1*(255-alpha))+(R2*alpha))>>8;
#define BLEND_MODE BLEND_NORMAL
#define CREATE_FUNCTIONS(name, a, r, g, b) \
void \
gst_videomixer_blend_##name##_##name (guint8 * src, gint xpos, gint ypos, \
gint src_width, gint src_height, gdouble src_alpha, \
guint8 * dest, gint dest_width, gint dest_height) \
{ \
gint alpha, s_alpha; \
gint i, j; \
gint src_stride, dest_stride; \
gint src_add, dest_add; \
gint B, G, R; \
\
src_stride = src_width * 4; \
dest_stride = dest_width * 4; \
\
s_alpha = CLAMP ((gint) (src_alpha * 256), 0, 256); \
\
/* adjust src pointers for negative sizes */ \
if (xpos < 0) { \
src += -xpos * 4; \
src_width -= -xpos; \
xpos = 0; \
} \
if (ypos < 0) { \
src += -ypos * src_stride; \
src_height -= -ypos; \
ypos = 0; \
} \
/* adjust width/height if the src is bigger than dest */ \
if (xpos + src_width > dest_width) { \
src_width = dest_width - xpos; \
} \
if (ypos + src_height > dest_height) { \
src_height = dest_height - ypos; \
} \
\
src_add = src_stride - (4 * src_width); \
dest_add = dest_stride - (4 * src_width); \
\
dest = dest + 4 * xpos + (ypos * dest_stride); \
\
for (i = 0; i < src_height; i++) { \
for (j = 0; j < src_width; j++) { \
alpha = (src[a] * s_alpha) >> 8; \
BLEND_MODE (dest[b], dest[g], dest[r], src[b], src[g], src[r], \
B, G, R, alpha); \
dest[b] = B; \
dest[g] = G; \
dest[r] = R; \
dest[a] = 0xff; \
\
src += 4; \
dest += 4; \
} \
src += src_add; \
dest += dest_add; \
} \
} \
\
/* fill a buffer with a checkerboard pattern */ \
void \
gst_videomixer_fill_##name##_checker (guint8 * dest, gint width, gint height) \
{ \
gint i, j; \
static const int tab[] = { 80, 160, 80, 160 }; \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[b] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* blue */ \
dest[g] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* green */ \
dest[r] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* red */ \
dest[a] = 0xFF; /* alpha */ \
dest += 4; \
} \
} \
} \
\
void \
gst_videomixer_fill_##name##_color (guint8 * dest, gint width, gint height, \
gint colY, gint colU, gint colV) \
{ \
gint red, green, blue; \
gint i, j; \
\
red = CLAMP (1.164 * (colY - 16) + 1.596 * (colV - 128), 0, 255); \
green = \
CLAMP (1.164 * (colY - 16) - 0.813 * (colV - 128) - 0.391 * (colU - 128), \
0, 255); \
blue = CLAMP (1.164 * (colY - 16) + 2.018 * (colU - 128), 0, 255); \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[b] = blue; \
dest[g] = green; \
dest[r] = red; \
dest[a] = 0xff; \
dest += 4; \
} \
} \
}
CREATE_FUNCTIONS (argb, 0, 1, 2, 3);
CREATE_FUNCTIONS (bgra, 3, 2, 1, 0);
#undef BLEND_MODE

339
gst/videomixer/blend_i420.c
View file

@ -1,339 +0,0 @@
/*
* Copyright (C) 2006 Mindfruit Bv.
* Author: Sjoerd Simons <sjoerd@luon.net>
* Author: Alex Ugarte <alexugarte@gmail.com>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#include <gst/gst.h>
#include <string.h>
#include "videomixer.h"
#define BLEND_NORMAL(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = ((Y1*(255-alpha))+(Y2*alpha))>>8; \
U = ((U1*(255-alpha))+(U2*alpha))>>8; \
V = ((V1*(255-alpha))+(V2*alpha))>>8;
#define BLEND_ADD(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = Y1+((Y2*alpha)>>8); \
U = U1+(((127*(255-alpha)+(U2*alpha)))>>8)-127; \
V = V1+(((127*(255-alpha)+(V2*alpha)))>>8)-127; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} \
U = MIN (U,255); \
V = MIN (V,255);
#define BLEND_SUBTRACT(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = Y1-((Y2*alpha)>>8); \
U = U1+(((127*(255-alpha)+(U2*alpha)))>>8)-127; \
V = V1+(((127*(255-alpha)+(V2*alpha)))>>8)-127; \
if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
}
#define BLEND_DARKEN(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
if (Y1 < Y2) { \
Y = Y1; U = U1; V = V1; \
} \
else { \
Y = ((Y1*(255-alpha))+(Y2*alpha))>>8; \
U = ((U1*(255-alpha))+(U2*alpha))>>8; \
V = ((V1*(255-alpha))+(V2*alpha))>>8; \
}
#define BLEND_LIGHTEN(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
if (Y1 > Y2) { \
Y = Y1; U = U1; V = V1; \
} \
else { \
Y = ((Y1*(255-alpha))+(Y2*alpha))>>8; \
U = ((U1*(255-alpha))+(U2*alpha))>>8; \
V = ((V1*(255-alpha))+(V2*alpha))>>8; \
}
#define BLEND_MULTIPLY(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = (Y1*(256*(255-alpha) +(Y2*alpha)))>>16; \
U = ((U1*(255-alpha)*256)+(alpha*(U1*Y2+128*(256-Y2))))>>16; \
V = ((V1*(255-alpha)*256)+(alpha*(V1*Y2+128*(256-Y2))))>>16;
#define BLEND_DIFFERENCE(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = ABS((gint)Y1-(gint)Y2)+127; \
U = ABS((gint)U1-(gint)U2)+127; \
V = ABS((gint)V1-(gint)V2)+127; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
U = CLAMP(U, 0, 255); \
V = CLAMP(V, 0, 255);
#define BLEND_EXCLUSION(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = ((gint)(Y1^0xff)*Y2+(gint)(Y2^0xff)*Y1)>>8; \
U = ((gint)(U1^0xff)*Y2+(gint)(Y2^0xff)*U1)>>8; \
V = ((gint)(V1^0xff)*Y2+(gint)(Y2^0xff)*V1)>>8; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
U = CLAMP(U, 0, 255); \
V = CLAMP(V, 0, 255);
#define BLEND_SOFTLIGHT(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = (gint)Y1+(gint)Y2 - 127; \
U = (gint)U1+(gint)U2 - 127; \
V = (gint)V1+(gint)V2 - 127; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
#define BLEND_HARDLIGHT(Y1,U1,V1,Y2,U2,V2,alpha,Y,U,V) \
Y = (gint)Y1+(gint)Y2*2 - 255; \
U = (gint)U1+(gint)U2 - 127; \
V = (gint)V1+(gint)V2 - 127; \
Y = ((Y*alpha)+(Y1*(255-alpha)))>>8; \
U = ((U*alpha)+(U1*(255-alpha)))>>8; \
V = ((V*alpha)+(V1*(255-alpha)))>>8; \
if (Y>255) { \
gint mult = MAX (0, 288-Y); \
U = ((U*mult) + (127*(32-mult)))>>5; \
V = ((V*mult) + (127*(32-mult)))>>5; \
Y = 255; \
} else if (Y<0) { \
gint mult = MIN (32, -Y); \
U = ((U*(32-mult)) + (127*mult))>>5; \
V = ((V*(32-mult)) + (127*mult))>>5; \
Y = 0; \
} \
#define BLEND_MODE BLEND_NORMAL
#if 0
#define BLEND_MODE BLEND_NORMAL
#define BLEND_MODE BLEND_ADD
#define BLEND_MODE BLEND_SUBTRACT
#define BLEND_MODE BLEND_LIGHTEN
#define BLEND_MODE BLEND_DARKEN
#define BLEND_MODE BLEND_MULTIPLY
#define BLEND_MODE BLEND_DIFFERENCE
#define BLEND_MODE BLEND_EXCLUSION
#define BLEND_MODE BLEND_SOFTLIGHT
#define BLEND_MODE BLEND_HARDLIGHT
#endif
/* I420 */
/* Copied from jpegenc */
#define VIDEO_Y_ROWSTRIDE(width) (GST_ROUND_UP_4(width))
#define VIDEO_U_ROWSTRIDE(width) (GST_ROUND_UP_8(width)/2)
#define VIDEO_V_ROWSTRIDE(width) ((GST_ROUND_UP_8(VIDEO_Y_ROWSTRIDE(width)))/2)
#define VIDEO_Y_OFFSET(w,h) (0)
#define VIDEO_U_OFFSET(w,h) (VIDEO_Y_OFFSET(w,h)+(VIDEO_Y_ROWSTRIDE(w)*GST_ROUND_UP_2(h)))
#define VIDEO_V_OFFSET(w,h) (VIDEO_U_OFFSET(w,h)+(VIDEO_U_ROWSTRIDE(w)*GST_ROUND_UP_2(h)/2))
#define VIDEO_SIZE(w,h) (VIDEO_V_OFFSET(w,h)+(VIDEO_V_ROWSTRIDE(w)*GST_ROUND_UP_2(h)/2))
inline static void
gst_i420_do_blend (guint8 * src, guint8 * dest,
gint src_stride, gint dest_stride, gint src_width, gint src_height,
gint dest_width, gdouble src_alpha)
{
int i, j;
gint b_alpha;
/* If it's completely transparent... we just return */
if (G_UNLIKELY (src_alpha == 0.0)) {
GST_INFO ("Fast copy (alpha == 0.0)");
return;
}
/* If it's completely opaque, we do a fast copy */
if (G_UNLIKELY (src_alpha == 1.0)) {
GST_INFO ("Fast copy (alpha == 1.0)");
for (i = 0; i < src_height; i++) {
memcpy (dest, src, src_width);
src += src_stride;
dest += dest_stride;
}
return;
}
b_alpha = CLAMP ((gint) (src_alpha * 255), 0, 255);
for (i = 0; i < src_height; i++) {
for (j = 0; j < src_width; j++) {
*dest = (b_alpha * (*src) + (255 - b_alpha) * (*dest)) >> 8;
dest++;
src++;
}
src += src_stride - src_width;
dest += dest_stride - dest_width;
}
}
/* note that this function does packing conversion and blending at the
* same time */
void
gst_videomixer_blend_i420_i420 (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height)
{
guint8 *b_src;
guint8 *b_dest;
gint b_src_width = src_width;
gint b_src_height = src_height;
gint xoffset = 0;
gint yoffset = 0;
xpos = GST_ROUND_UP_2 (xpos);
ypos = GST_ROUND_UP_2 (ypos);
/* adjust src pointers for negative sizes */
if (xpos < 0) {
xoffset = -xpos;
b_src_width -= -xpos;
xpos = 0;
}
if (ypos < 0) {
yoffset += -ypos;
b_src_height -= -ypos;
ypos = 0;
}
/* If x or y offset are larger then the source it's outside of the picture */
if (xoffset > src_width || yoffset > src_width) {
return;
}
/* adjust width/height if the src is bigger than dest */
if (xpos + src_width > dest_width) {
b_src_width = dest_width - xpos;
}
if (ypos + src_height > dest_height) {
b_src_height = dest_height - ypos;
}
if (b_src_width < 0 || b_src_height < 0) {
return;
}
/* First mix Y, then U, then V */
b_src = src + VIDEO_Y_OFFSET (src_width, src_height);
b_dest = dest + VIDEO_Y_OFFSET (dest_width, dest_height);
gst_i420_do_blend (b_src + xoffset + yoffset * VIDEO_Y_ROWSTRIDE (src_width),
b_dest + xpos + ypos * VIDEO_Y_ROWSTRIDE (dest_width),
VIDEO_Y_ROWSTRIDE (src_width),
VIDEO_Y_ROWSTRIDE (dest_width), b_src_width, b_src_height,
dest_width, src_alpha);
b_src = src + VIDEO_U_OFFSET (src_width, src_height);
b_dest = dest + VIDEO_U_OFFSET (dest_width, dest_height);
gst_i420_do_blend (b_src + xoffset / 2 +
yoffset / 2 * VIDEO_U_ROWSTRIDE (src_width),
b_dest + xpos / 2 + ypos / 2 * VIDEO_U_ROWSTRIDE (dest_width),
VIDEO_U_ROWSTRIDE (src_width), VIDEO_U_ROWSTRIDE (dest_width),
b_src_width / 2, GST_ROUND_UP_2 (b_src_height) / 2, dest_width / 2,
src_alpha);
b_src = src + VIDEO_V_OFFSET (src_width, src_height);
b_dest = dest + VIDEO_V_OFFSET (dest_width, dest_height);
gst_i420_do_blend (b_src + xoffset / 2 +
yoffset / 2 * VIDEO_V_ROWSTRIDE (src_width),
b_dest + xpos / 2 + ypos / 2 * VIDEO_V_ROWSTRIDE (dest_width),
VIDEO_V_ROWSTRIDE (src_width), VIDEO_V_ROWSTRIDE (dest_width),
b_src_width / 2, GST_ROUND_UP_2 (b_src_height) / 2, dest_width / 2,
src_alpha);
}
#undef BLEND_MODE
/* fill a buffer with a checkerboard pattern */
void
gst_videomixer_fill_i420_checker (guint8 * dest, gint width, gint height)
{
int size;
gint i, j;
static const int tab[] = { 80, 160, 80, 160 };
guint8 *p = dest;
for (i = 0; i < height; i++) {
for (j = 0; j < width; j++) {
*p++ = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)];
}
p += VIDEO_Y_ROWSTRIDE (width) - width;
}
size = (VIDEO_U_ROWSTRIDE (width) * height) / 2;
memset (dest + VIDEO_U_OFFSET (width, height), 0x80, size);
size = (VIDEO_V_ROWSTRIDE (width) * height) / 2;
memset (dest + VIDEO_V_OFFSET (width, height), 0x80, size);
}
void
gst_videomixer_fill_i420_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV)
{
int size;
size = VIDEO_Y_ROWSTRIDE (width) * height;
memset (dest, colY, size);
size = (VIDEO_U_ROWSTRIDE (width) * height) / 2;
memset (dest + VIDEO_U_OFFSET (width, height), colU, size);
size = (VIDEO_V_ROWSTRIDE (width) * height) / 2;
memset (dest + VIDEO_V_OFFSET (width, height), colV, size);
}

168
gst/videomixer/blend_mmx.h Normal file
View file

@ -0,0 +1,168 @@
#ifdef A32
static inline void
NAME_BLEND (guint8 * dest, const guint8 * src, gint src_height, gint src_width,
gint src_stride, gint dest_stride, guint s_alpha)
{
gint i;
gint src_add = src_stride - (4 * src_width);
gint dest_add = dest_stride - (4 * src_width);
for (i = 0; i < src_height; i++) {
gulong old_ebx;
/* (P1 * (256 - A) + (P2 * A)) / 256
* => (P1 * 256 - P1 * A + P2 * A) / 256
* => (P1 * 256 + A * (P2 - P1) / 256
* => P1 + (A * (P2 - P1)) / 256
*/
/* *INDENT-OFF* */
__asm__ __volatile__ (
" movl %%ebx , %6 \n\t"
" pcmpeqd %%mm5 , %%mm5 \n\t" /* mm5 = 0xffff... */
#if A_OFF == 0
" psrld $24 , %%mm5 \n\t" /* mm5 = 00 00 00 ff 00 00 00 ff, selector for alpha */
#else
" pslld $24 , %%mm5 \n\t" /* mm5 = ff 00 00 00 ff 00 00 00, selector for alpha */
#endif
" mov %4 , %%eax \n\t" /* eax = s_alpha */
" movd %%eax , %%mm6 \n\t" /* mm6 = s_alpha */
" punpckldq %%mm6 , %%mm6 \n\t" /* mm6 = 00 00 00 aa 00 00 00 aa, alpha scale factor */
" movl %5 , %%ebx \n\t" /* ebx = src_width */
" test $1 , %%ebx \n\t" /* check odd pixel */
" je 1f \n\t"
/* do odd pixel */
" movd (%2) , %%mm2 \n\t" /* mm2 = src, 00 00 00 00 sv su sy sa */
" movd (%3) , %%mm1 \n\t" /* mm1 = dest, 00 00 00 00 dv du dy da */
" movq %%mm2 , %%mm0 \n\t"
" punpcklbw %%mm7 , %%mm2 \n\t" /* mm2 = 00 sv 00 su 00 sy 00 sa */
" pand %%mm5 , %%mm0 \n\t" /* mm0 = 00 00 00 00 00 00 00 sa, get alpha component */
#if A_OFF != 0
" psrld $24 , %%mm0 \n\t"
#endif
" punpcklbw %%mm7 , %%mm1 \n\t" /* mm1 = 00 dv 00 du 00 dy 00 da */
" pmullw %%mm6 , %%mm0 \n\t" /* mult with scale */
" psubw %%mm1 , %%mm2 \n\t" /* mm2 = mm2 - mm1 */
" punpcklwd %%mm0 , %%mm0 \n\t"
" punpckldq %%mm0 , %%mm0 \n\t" /* mm0 == 00 aa 00 aa 00 aa 00 aa */
" psrlw $8 , %%mm0 \n\t"
" pmullw %%mm0 , %%mm2 \n\t" /* mm2 == a * mm2 */
" psllw $8 , %%mm1 \n\t" /* scale up */
" paddw %%mm1 , %%mm2 \n\t" /* mm2 == mm2 + mm1 */
" psrlw $8 , %%mm2 \n\t" /* scale down */
" por %%mm5 , %%mm2 \n\t" /* set alpha to ff */
" packuswb %%mm2 , %%mm2 \n\t"
" movd %%mm2 , (%3) \n\t" /* dest = mm1 */
" add $4 , %1 \n\t"
" add $4 , %0 \n\t"
"1: \n\t"
" sar $1 , %%ebx \n\t" /* prepare for 2 pixel per loop */
" cmp $0 , %%ebx \n\t"
" je 3f \n\t"
"2: \n\t"
/* do even pixels */
" movq (%2) , %%mm2 \n\t" /* mm2 = src, sv1 su1 sy1 sa1 sv0 su0 sy0 sa0 */
" movq (%3) , %%mm1 \n\t" /* mm1 = dest, dv1 du1 dy1 da1 dv0 du0 dy0 da0 */
" movq %%mm2 , %%mm4 \n\t"
" movq %%mm1 , %%mm3 \n\t"
" movq %%mm2 , %%mm0 \n\t" /* copy for doing the alpha */
" pxor %%mm7 , %%mm7 \n\t"
" punpcklbw %%mm7 , %%mm2 \n\t" /* mm2 = 00 sv0 00 su0 00 sy0 00 sa0 */
" punpckhbw %%mm7 , %%mm4 \n\t" /* mm4 = 00 sv1 00 su1 00 sy1 00 sa1 */
" punpcklbw %%mm7 , %%mm1 \n\t" /* mm1 = 00 dv0 00 du0 00 dy0 00 da0 */
" punpckhbw %%mm7 , %%mm3 \n\t" /* mm2 = 00 dv1 00 du1 00 dy1 00 da1 */
" pand %%mm5 , %%mm0 \n\t" /* mm0 = 00 00 00 sa1 00 00 00 sa0 */
#if A_OFF != 0
" psrld $24 , %%mm0 \n\t"
#endif
" psubw %%mm1 , %%mm2 \n\t" /* mm2 = mm2 - mm1 */
" pmullw %%mm6 , %%mm0 \n\t" /* mult with scale */
" psubw %%mm3 , %%mm4 \n\t" /* mm4 = mm4 - mm3 */
" psrlw $8 , %%mm0 \n\t" /* scale back */
" movq %%mm0 , %%mm7 \n\t" /* save copy */
" punpcklwd %%mm0 , %%mm0 \n\t" /* mm0 = 00 00 00 00 00 aa0 00 aa0 */
" punpckhwd %%mm7 , %%mm7 \n\t" /* mm7 = 00 00 00 00 00 aa1 00 aa1 */
" punpckldq %%mm0 , %%mm0 \n\t" /* mm0 = 00 aa0 00 aa0 00 aa0 00 aa0 */
" punpckldq %%mm7 , %%mm7 \n\t" /* mm7 = 00 aa1 00 aa1 00 aa1 00 aa1 */
" pmullw %%mm0 , %%mm2 \n\t" /* mm2 == aa * mm2 */
" pmullw %%mm7 , %%mm4 \n\t" /* mm2 == aa * mm2 */
" psllw $8 , %%mm1 \n\t"
" psllw $8 , %%mm3 \n\t"
" paddw %%mm1 , %%mm2 \n\t" /* mm2 == mm2 + mm1 */
" paddw %%mm3 , %%mm4 \n\t" /* mm2 == mm2 + mm1 */
" psrlw $8 , %%mm2 \n\t"
" psrlw $8 , %%mm4 \n\t"
" packuswb %%mm4 , %%mm2 \n\t"
" por %%mm5 , %%mm2 \n\t" /* set alpha to ff */
" movq %%mm2 , (%3) \n\t"
" add $8 , %1 \n\t"
" add $8 , %0 \n\t"
" dec %%ebx \n\t"
" jne 2b \n\t"
"3: \n\t"
" movl %6 , %%ebx \n\t"
:"=r" (src), "=r" (dest)
:"0" (src), "1" (dest), "m" (s_alpha), "m" (src_width), "m" (old_ebx)
:"%eax", "memory"
#ifdef __MMX__
, "mm0", "mm1", "mm2", "mm5", "mm6", "mm7"
#endif
);
/* *INDENT-ON* */
src += src_add;
dest += dest_add;
}
__asm__ __volatile__ ("emms");
}
static inline void
NAME_FILL_COLOR (guint8 * dest, gint height, gint width, gint c1, gint c2,
gint c3)
{
guint64 val;
guint nvals = width * height;
val = (((guint64) 0xff << A_OFF)) | (((guint64) c1) << C1_OFF) |
(((guint64) c2) << C2_OFF) | (((guint64) c3) << C3_OFF);
val = (val << 32) | val;
/* *INDENT-OFF* */
__asm__ __volatile__ (
"movq %4 , %%mm0 \n\t"
"test $1 , %0 \n\t"
"je 1f \n\t"
"movd %%mm0 , (%1) \n\t"
"add $4 , %1 \n\t"
"dec %0 \n\t"
"1: \n\t"
"sar $1 , %0 \n\t"
"cmp $0 , %0 \n\t"
"je 3f \n\t"
"2: \n\t"
"movq %%mm0 , (%1) \n\t"
"add $8 , %1 \n\t"
"dec %0 \n\t"
"jne 2b \n\t"
"3: \n\t"
"emms \n\t"
: "=r" (nvals), "=r" (dest)
: "0" (nvals), "1" (dest), "m" (val)
: "memory"
#ifdef __MMX__
, "mm0"
#endif
);
/* *INDENT-ON* */
}
#endif

155
gst/videomixer/blend_rgb.c
View file

@ -1,155 +0,0 @@
/*
* Copyright (C) 2009 Alex Ugarte <augarte@vicomtech.org>
* Copyright (C) 2009 Sebastian Dröge <sebastian.droege@collabora.co.uk>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#include <gst/gst.h>
#include <string.h>
#include "videomixer.h"
#define BLEND_NORMAL(R1,G1,B1,R2,G2,B2,alpha,R,G,B) \
R = ((R1*(255-alpha))+(R2*alpha))>>8; \
G = ((G1*(255-alpha))+(G2*alpha))>>8; \
B = ((B1*(255-alpha))+(B2*alpha))>>8;
#define BLEND_MODE BLEND_NORMAL
#define CREATE_FUNCTIONS(name, bpp, r, g, b) \
void \
gst_videomixer_blend_##name##_##name (guint8 * src, gint xpos, gint ypos, \
gint src_width, gint src_height, gdouble src_alpha, \
guint8 * dest, gint dest_width, gint dest_height) \
{ \
gint b_alpha; \
gint i, j; \
gint src_stride, dest_stride; \
gint src_add, dest_add; \
gint R, G, B; \
\
src_stride = GST_ROUND_UP_4 (src_width * bpp); \
dest_stride = GST_ROUND_UP_4 (dest_width * bpp); \
\
b_alpha = CLAMP ((gint) (src_alpha * 255), 0, 255); \
\
/* adjust src pointers for negative sizes */ \
if (xpos < 0) { \
src += -xpos * bpp; \
src_width -= -xpos; \
xpos = 0; \
} \
if (ypos < 0) { \
src += -ypos * src_stride; \
src_height -= -ypos; \
ypos = 0; \
} \
/* adjust width/height if the src is bigger than dest */ \
if (xpos + src_width > dest_width) { \
src_width = dest_width - xpos; \
} \
if (ypos + src_height > dest_height) { \
src_height = dest_height - ypos; \
} \
\
src_add = src_stride - (bpp * src_width); \
dest_add = dest_stride - (bpp * src_width); \
\
dest = dest + bpp * xpos + (ypos * dest_stride); \
/* If it's completely transparent... we just return */ \
if (G_UNLIKELY (src_alpha == 0.0)) { \
GST_INFO ("Fast copy (alpha == 0.0)"); \
return; \
} \
\
/* If it's completely opaque, we do a fast copy */ \
if (G_UNLIKELY (src_alpha == 1.0)) { \
GST_INFO ("Fast copy (alpha == 1.0)"); \
for (i = 0; i < src_height; i++) { \
memcpy (dest, src, bpp * src_width); \
src += src_stride; \
dest += dest_stride; \
} \
return; \
} \
\
for (i = 0; i < src_height; i++) { \
for (j = 0; j < src_width; j++) { \
BLEND_MODE (dest[r], dest[g], dest[b], src[r], src[g], src[b], \
b_alpha, R, G, B); \
dest[r] = R; \
dest[g] = G; \
dest[b] = B; \
\
src += bpp; \
dest += bpp; \
} \
src += src_add; \
dest += dest_add; \
} \
} \
\
/* fill a buffer with a checkerboard pattern */ \
void \
gst_videomixer_fill_##name##_checker (guint8 * dest, gint width, gint height) \
{ \
gint i, j; \
static const int tab[] = { 80, 160, 80, 160 }; \
gint dest_add = GST_ROUND_UP_4 (width * bpp) - width * bpp; \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[r] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* red */ \
dest[g] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* green */ \
dest[b] = tab[((i & 0x8) >> 3) + ((j & 0x8) >> 3)]; /* blue */ \
dest += bpp; \
} \
dest += dest_add; \
} \
} \
\
void \
gst_videomixer_fill_##name##_color (guint8 * dest, gint width, gint height, \
gint colY, gint colU, gint colV) \
{ \
gint red, green, blue; \
gint i, j; \
gint dest_add = GST_ROUND_UP_4 (width * bpp) - width * bpp; \
\
red = CLAMP (1.164 * (colY - 16) + 1.596 * (colV - 128), 0, 255); \
green = \
CLAMP (1.164 * (colY - 16) - 0.813 * (colV - 128) - 0.391 * (colU - 128), \
0, 255); \
blue = CLAMP (1.164 * (colY - 16) + 2.018 * (colU - 128), 0, 255); \
\
for (i = 0; i < height; i++) { \
for (j = 0; j < width; j++) { \
dest[r] = red; \
dest[g] = green; \
dest[b] = blue; \
dest += bpp; \
} \
dest += dest_add; \
} \
}
CREATE_FUNCTIONS (rgb, 3, 0, 1, 2);
CREATE_FUNCTIONS (bgr, 3, 2, 1, 0);
CREATE_FUNCTIONS (xrgb, 4, 1, 2, 3);
CREATE_FUNCTIONS (xbgr, 4, 3, 2, 1);
CREATE_FUNCTIONS (rgbx, 4, 0, 1, 2);
CREATE_FUNCTIONS (bgrx, 4, 2, 1, 0);

166
gst/videomixer/videomixer.c
View file

@ -68,10 +68,6 @@
#include <gst/controller/gstcontroller.h>
#include <gst/video/video.h>
#include <liboil/liboil.h>
#include <liboil/liboilcpu.h>
#include <liboil/liboilfunction.h>
#ifdef HAVE_STDLIB_H
#include <stdlib.h>
#endif
@ -105,79 +101,6 @@ static gboolean gst_videomixer_sink_event (GstPad * pad, GstEvent * event);
static void gst_videomixer_sort_pads (GstVideoMixer * mix);
/*AYUV function definitions see file: blend_ayuv*/
void gst_videomixer_blend_ayuv_ayuv (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_ayuv_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_ayuv_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
/*BGRA/ARGB function definitions see file: blend_bgra*/
void gst_videomixer_blend_bgra_bgra (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_bgra_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_bgra_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
void gst_videomixer_blend_argb_argb (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_argb_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_argb_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
/* RGB function definitions see file: blend_rgb.c */
void gst_videomixer_blend_rgb_rgb (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_rgb_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_rgb_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
void gst_videomixer_blend_bgr_bgr (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_bgr_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_bgr_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
void gst_videomixer_blend_xrgb_xrgb (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_xrgb_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_xrgb_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
void gst_videomixer_blend_xbgr_xbgr (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_xbgr_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_xbgr_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
void gst_videomixer_blend_rgbx_rgbx (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_rgbx_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_rgbx_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
void gst_videomixer_blend_bgrx_bgrx (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_bgrx_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_bgrx_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
/*I420 function definitions see file: blend_i420.c*/
void gst_videomixer_blend_i420_i420 (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_heighty);
void gst_videomixer_fill_i420_checker (guint8 * dest, gint width, gint height);
void gst_videomixer_fill_i420_color (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
#ifdef BUILD_X86_ASM
void gst_videomixer_blend_ayuv_ayuv_mmx (guint8 * src, gint xpos, gint ypos,
gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void gst_videomixer_fill_ayuv_color_mmx (guint8 * dest, gint width, gint height,
gint colY, gint colU, gint colV);
#endif
#define DEFAULT_PAD_ZORDER 0
#define DEFAULT_PAD_XPOS 0
#define DEFAULT_PAD_YPOS 0
@ -938,77 +861,64 @@ gst_videomixer_setcaps (GstPad * pad, GstCaps * caps)
goto done;
switch (mixer->fmt) {
case GST_VIDEO_FORMAT_AYUV:{
#ifdef BUILD_X86_ASM
guint cpu_flags = oil_cpu_get_flags ();
mixer->blend =
(cpu_flags & OIL_IMPL_FLAG_MMX) ? gst_videomixer_blend_ayuv_ayuv_mmx :
gst_videomixer_blend_ayuv_ayuv;
mixer->fill_checker = gst_videomixer_fill_ayuv_checker;
mixer->fill_color =
(cpu_flags & OIL_IMPL_FLAG_MMX) ? gst_videomixer_fill_ayuv_color_mmx :
gst_videomixer_fill_ayuv_color;
#else
mixer->blend = gst_videomixer_blend_ayuv_ayuv;
mixer->fill_checker = gst_videomixer_fill_ayuv_checker;
mixer->fill_color = gst_videomixer_fill_ayuv_color;
#endif
ret = TRUE;
break;
}
case GST_VIDEO_FORMAT_I420:
mixer->blend = gst_videomixer_blend_i420_i420;
mixer->fill_checker = gst_videomixer_fill_i420_checker;
mixer->fill_color = gst_videomixer_fill_i420_color;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_BGRA:
mixer->blend = gst_videomixer_blend_bgra_bgra;
mixer->fill_checker = gst_videomixer_fill_bgra_checker;
mixer->fill_color = gst_videomixer_fill_bgra_color;
case GST_VIDEO_FORMAT_AYUV:
mixer->blend = gst_video_mixer_blend_ayuv;
mixer->fill_checker = gst_video_mixer_fill_checker_ayuv;
mixer->fill_color = gst_video_mixer_fill_color_ayuv;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_ARGB:
mixer->blend = gst_videomixer_blend_argb_argb;
mixer->fill_checker = gst_videomixer_fill_argb_checker;
mixer->fill_color = gst_videomixer_fill_argb_color;
mixer->blend = gst_video_mixer_blend_argb;
mixer->fill_checker = gst_video_mixer_fill_checker_argb;
mixer->fill_color = gst_video_mixer_fill_color_argb;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_BGRA:
mixer->blend = gst_video_mixer_blend_bgra;
mixer->fill_checker = gst_video_mixer_fill_checker_bgra;
mixer->fill_color = gst_video_mixer_fill_color_bgra;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_I420:
mixer->blend = gst_video_mixer_blend_i420;
mixer->fill_checker = gst_video_mixer_fill_checker_i420;
mixer->fill_color = gst_video_mixer_fill_color_i420;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_RGB:
mixer->blend = gst_videomixer_blend_rgb_rgb;
mixer->fill_checker = gst_videomixer_fill_rgb_checker;
mixer->fill_color = gst_videomixer_fill_rgb_color;
mixer->blend = gst_video_mixer_blend_rgb;
mixer->fill_checker = gst_video_mixer_fill_checker_rgb;
mixer->fill_color = gst_video_mixer_fill_color_rgb;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_BGR:
mixer->blend = gst_videomixer_blend_bgr_bgr;
mixer->fill_checker = gst_videomixer_fill_bgr_checker;
mixer->fill_color = gst_videomixer_fill_bgr_color;
mixer->blend = gst_video_mixer_blend_bgr;
mixer->fill_checker = gst_video_mixer_fill_checker_bgr;
mixer->fill_color = gst_video_mixer_fill_color_bgr;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_xRGB:
mixer->blend = gst_videomixer_blend_xrgb_xrgb;
mixer->fill_checker = gst_videomixer_fill_xrgb_checker;
mixer->fill_color = gst_videomixer_fill_xrgb_color;
mixer->blend = gst_video_mixer_blend_xrgb;
mixer->fill_checker = gst_video_mixer_fill_checker_xrgb;
mixer->fill_color = gst_video_mixer_fill_color_xrgb;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_xBGR:
mixer->blend = gst_videomixer_blend_xbgr_xbgr;
mixer->fill_checker = gst_videomixer_fill_xbgr_checker;
mixer->fill_color = gst_videomixer_fill_xbgr_color;
mixer->blend = gst_video_mixer_blend_xbgr;
mixer->fill_checker = gst_video_mixer_fill_checker_xbgr;
mixer->fill_color = gst_video_mixer_fill_color_xbgr;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_RGBx:
mixer->blend = gst_videomixer_blend_rgbx_rgbx;
mixer->fill_checker = gst_videomixer_fill_rgbx_checker;
mixer->fill_color = gst_videomixer_fill_rgbx_color;
mixer->blend = gst_video_mixer_blend_rgbx;
mixer->fill_checker = gst_video_mixer_fill_checker_rgbx;
mixer->fill_color = gst_video_mixer_fill_color_rgbx;
ret = TRUE;
break;
case GST_VIDEO_FORMAT_BGRx:
mixer->blend = gst_videomixer_blend_bgrx_bgrx;
mixer->fill_checker = gst_videomixer_fill_bgrx_checker;
mixer->fill_color = gst_videomixer_fill_bgrx_color;
mixer->blend = gst_video_mixer_blend_bgrx;
mixer->fill_checker = gst_video_mixer_fill_checker_bgrx;
mixer->fill_color = gst_video_mixer_fill_color_bgrx;
ret = TRUE;
break;
default:
@ -1649,7 +1559,7 @@ plugin_init (GstPlugin * plugin)
GST_DEBUG_CATEGORY_INIT (gst_videomixer_debug, "videomixer", 0,
"video mixer");
oil_init ();
gst_video_mixer_init_blend ();
return gst_element_register (plugin, "videomixer", GST_RANK_PRIMARY,
GST_TYPE_VIDEO_MIXER);

9
gst/videomixer/videomixer.h
View file

@ -23,6 +23,7 @@
#include <gst/gst.h>
#include <gst/video/video.h>
#include "videomixerpad.h"
#include "blend.h"
GST_DEBUG_CATEGORY_EXTERN (gst_videomixer_debug);
#define GST_CAT_DEFAULT gst_videomixer_debug
@ -105,11 +106,9 @@ struct _GstVideoMixer
guint64 segment_position;
gdouble segment_rate;
void (*blend) (guint8 * src, gint xpos, gint ypos, gint src_width, gint src_height, gdouble src_alpha,
guint8 * dest, gint dest_width, gint dest_height);
void (*fill_checker) (guint8 * dest, gint width, gint height);
void (*fill_color) (guint8 * dest, gint width, gint height, gint colY, gint colU, gint colV);
BlendFunction blend;
FillCheckerFunction fill_checker;
FillColorFunction fill_color;
};
struct _GstVideoMixerClass