From cc30aae17bf521fc5453c7ee173cceefde255faf Mon Sep 17 00:00:00 2001 From: lucabe Date: Fri, 30 Jun 2006 12:00:31 +0000 Subject: Move postproc ---> libswscale git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@18866 b3059339-0415-0410-9bf9-f77b7e298cf2 --- libswscale/Makefile | 52 + libswscale/cs_test.c | 288 ++++ libswscale/rgb2rgb.c | 644 ++++++++ libswscale/rgb2rgb.h | 144 ++ libswscale/rgb2rgb_template.c | 2675 ++++++++++++++++++++++++++++++ libswscale/swscale-example.c | 216 +++ libswscale/swscale.c | 2707 ++++++++++++++++++++++++++++++ libswscale/swscale.h | 129 ++ libswscale/swscale_altivec_template.c | 541 ++++++ libswscale/swscale_internal.h | 167 ++ libswscale/swscale_template.c | 2928 +++++++++++++++++++++++++++++++++ libswscale/yuv2rgb.c | 848 ++++++++++ libswscale/yuv2rgb_altivec.c | 953 +++++++++++ libswscale/yuv2rgb_mlib.c | 88 + libswscale/yuv2rgb_template.c | 540 ++++++ 15 files changed, 12920 insertions(+) create mode 100644 libswscale/Makefile create mode 100644 libswscale/cs_test.c create mode 100644 libswscale/rgb2rgb.c create mode 100644 libswscale/rgb2rgb.h create mode 100644 libswscale/rgb2rgb_template.c create mode 100644 libswscale/swscale-example.c create mode 100644 libswscale/swscale.c create mode 100644 libswscale/swscale.h create mode 100644 libswscale/swscale_altivec_template.c create mode 100644 libswscale/swscale_internal.h create mode 100644 libswscale/swscale_template.c create mode 100644 libswscale/yuv2rgb.c create mode 100644 libswscale/yuv2rgb_altivec.c create mode 100644 libswscale/yuv2rgb_mlib.c create mode 100644 libswscale/yuv2rgb_template.c (limited to 'libswscale') diff --git a/libswscale/Makefile b/libswscale/Makefile new file mode 100644 index 0000000000..475bd31381 --- /dev/null +++ b/libswscale/Makefile @@ -0,0 +1,52 @@ + +include ../config.mak + +SWSLIB = libswscale.a + +SWSSRCS=swscale.c rgb2rgb.c yuv2rgb.c + +SWSOBJS=$(SWSSRCS:.c=.o) + +ifeq ($(TARGET_ALTIVEC),yes) +SWSOBJS += yuv2rgb_altivec.o +endif + +CFLAGS = $(OPTFLAGS) $(MLIB_INC) -I. -I.. $(EXTRA_INC) +# -I/usr/X11R6/include/ + +.SUFFIXES: .c .o + +# .PHONY: all clean + +.c.o: + $(CC) -c $(CFLAGS) -I.. -o $@ $< + +all: $(SWSLIB) + +$(SWSLIB): $(SWSOBJS) + $(AR) r $(SWSLIB) $(SWSOBJS) + $(RANLIB) $(SWSLIB) + +clean: + rm -f *.o *.a *~ *.so cs_test swscale-example + +distclean: clean + rm -f .depend + +dep: depend + +depend: + $(CC) -MM $(CFLAGS) $(SWSSRCS) 1>.depend + +cs_test: cs_test.o $(SWSLIB) + $(CC) cs_test.o $(SWSLIB) ../cpudetect.o -DFOR_MENCODER ../mp_msg.c -o cs_test -W -Wall + +swscale-example: swscale-example.o $(SWSLIB) + $(CC) swscale-example.o $(SWSLIB) ../libmpcodecs/img_format.o -lm -o swscale-example -W -Wall +# +# include dependency files if they exist +# +ifneq ($(wildcard .depend),) +include .depend +endif + diff --git a/libswscale/cs_test.c b/libswscale/cs_test.c new file mode 100644 index 0000000000..b1963f92f6 --- /dev/null +++ b/libswscale/cs_test.c @@ -0,0 +1,288 @@ +/* + Copyright (C) 2002 Michael Niedermayer + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + This program 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 General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA +*/ + +#include +#include +#include + +#include "swscale.h" +#include "rgb2rgb.h" +#include "cpudetect.h" + +#define SIZE 1000 +#define srcByte 0x55 +#define dstByte 0xBB + +#ifdef __APPLE_CC__ +#define memalign(x,y) malloc(y) +#endif + +static int get_sws_cpuflags() +{ + return (gCpuCaps.hasMMX ? SWS_CPU_CAPS_MMX : 0) | + (gCpuCaps.hasMMX2 ? SWS_CPU_CAPS_MMX2 : 0) | + (gCpuCaps.has3DNow ? SWS_CPU_CAPS_3DNOW : 0); +} + +main(int argc, char **argv) +{ + int i, funcNum; + uint8_t *srcBuffer= (uint8_t*)memalign(128, SIZE); + uint8_t *dstBuffer= (uint8_t*)memalign(128, SIZE); + int failedNum=0; + int passedNum=0; + + printf("memory corruption test ...\n"); + + if(argc==2){ + GetCpuCaps(&gCpuCaps); + printf("testing mmx\n"); + } + + sws_rgb2rgb_init(get_sws_cpuflags()); + + for(funcNum=0; funcNum<100; funcNum++){ + int width; + int failed=0; + int srcBpp=0; + int dstBpp=0; + + printf("."); fflush(stdout); + memset(srcBuffer, srcByte, SIZE); + + for(width=32; width<64; width++){ + int dstOffset; + for(dstOffset=128; dstOffset<196; dstOffset++){ + int srcOffset; + memset(dstBuffer, dstByte, SIZE); + + for(srcOffset=128; srcOffset<196; srcOffset++){ + uint8_t *src= srcBuffer+srcOffset; + uint8_t *dst= dstBuffer+dstOffset; + char *name=NULL; + + if(failed) break; //don't fill the screen with shit ... + + switch(funcNum){ + case 0: + srcBpp=2; + dstBpp=2; + name="rgb15to16"; + rgb15to16(src, dst, width*srcBpp); + break; + case 1: + srcBpp=2; + dstBpp=3; + name="rgb15to24"; + rgb15to24(src, dst, width*srcBpp); + break; + case 2: + srcBpp=2; + dstBpp=4; + name="rgb15to32"; + rgb15to32(src, dst, width*srcBpp); + break; + case 3: + srcBpp=2; + dstBpp=3; + name="rgb16to24"; + rgb16to24(src, dst, width*srcBpp); + break; + case 4: + srcBpp=2; + dstBpp=4; + name="rgb16to32"; + rgb16to32(src, dst, width*srcBpp); + break; + case 5: + srcBpp=3; + dstBpp=2; + name="rgb24to15"; + rgb24to15(src, dst, width*srcBpp); + break; + case 6: + srcBpp=3; + dstBpp=2; + name="rgb24to16"; + rgb24to16(src, dst, width*srcBpp); + break; + case 7: + srcBpp=3; + dstBpp=4; + name="rgb24to32"; + rgb24to32(src, dst, width*srcBpp); + break; + case 8: + srcBpp=4; + dstBpp=2; + name="rgb32to15"; + rgb32to15(src, dst, width*srcBpp); + break; + case 9: + srcBpp=4; + dstBpp=2; + name="rgb32to16"; + rgb32to16(src, dst, width*srcBpp); + break; + case 10: + srcBpp=4; + dstBpp=3; + name="rgb32to24"; + rgb32to24(src, dst, width*srcBpp); + break; + case 11: + srcBpp=2; + dstBpp=2; + name="rgb16to15"; + rgb16to15(src, dst, width*srcBpp); + break; + + case 14: + srcBpp=2; + dstBpp=2; + name="rgb15tobgr15"; + rgb15tobgr15(src, dst, width*srcBpp); + break; + case 15: + srcBpp=2; + dstBpp=2; + name="rgb15tobgr16"; + rgb15tobgr16(src, dst, width*srcBpp); + break; + case 16: + srcBpp=2; + dstBpp=3; + name="rgb15tobgr24"; + rgb15tobgr24(src, dst, width*srcBpp); + break; + case 17: + srcBpp=2; + dstBpp=4; + name="rgb15tobgr32"; + rgb15tobgr32(src, dst, width*srcBpp); + break; + case 18: + srcBpp=2; + dstBpp=2; + name="rgb16tobgr15"; + rgb16tobgr15(src, dst, width*srcBpp); + break; + case 19: + srcBpp=2; + dstBpp=2; + name="rgb16tobgr16"; + rgb16tobgr16(src, dst, width*srcBpp); + break; + case 20: + srcBpp=2; + dstBpp=3; + name="rgb16tobgr24"; + rgb16tobgr24(src, dst, width*srcBpp); + break; + case 21: + srcBpp=2; + dstBpp=4; + name="rgb16tobgr32"; + rgb16tobgr32(src, dst, width*srcBpp); + break; + case 22: + srcBpp=3; + dstBpp=2; + name="rgb24tobgr15"; + rgb24tobgr15(src, dst, width*srcBpp); + break; + case 23: + srcBpp=3; + dstBpp=2; + name="rgb24tobgr16"; + rgb24tobgr16(src, dst, width*srcBpp); + break; + case 24: + srcBpp=3; + dstBpp=3; + name="rgb24tobgr24"; + rgb24tobgr24(src, dst, width*srcBpp); + break; + case 25: + srcBpp=3; + dstBpp=4; + name="rgb24tobgr32"; + rgb24tobgr32(src, dst, width*srcBpp); + break; + case 26: + srcBpp=4; + dstBpp=2; + name="rgb32tobgr15"; + rgb32tobgr15(src, dst, width*srcBpp); + break; + case 27: + srcBpp=4; + dstBpp=2; + name="rgb32tobgr16"; + rgb32tobgr16(src, dst, width*srcBpp); + break; + case 28: + srcBpp=4; + dstBpp=3; + name="rgb32tobgr24"; + rgb32tobgr24(src, dst, width*srcBpp); + break; + case 29: + srcBpp=4; + dstBpp=4; + name="rgb32tobgr32"; + rgb32tobgr32(src, dst, width*srcBpp); + break; + + } + if(!srcBpp) break; + + for(i=0; i +#include "config.h" +#include "rgb2rgb.h" +#include "swscale.h" +#include "cpudetect.h" +#include "mangle.h" +#include "bswap.h" +#include "libvo/fastmemcpy.h" + +#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit + +void (*rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb24to16)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb24to15)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb32to16)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb32to15)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb15to24)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb15to32)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb16to24)(const uint8_t *src,uint8_t *dst,long src_size); +void (*rgb16to32)(const uint8_t *src,uint8_t *dst,long src_size); +//void (*rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size); +void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size); +void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size); +void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size); +void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size); +//void (*rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size); +void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size); +void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size); + +void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, + long width, long height, + long lumStride, long chromStride, long dstStride); +void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, + long width, long height, + long lumStride, long chromStride, long dstStride); +void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, + long width, long height, + long lumStride, long chromStride, long dstStride); +void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + long width, long height, + long lumStride, long chromStride, long srcStride); +void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + long width, long height, + long lumStride, long chromStride, long srcStride); +void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height, + long srcStride, long dstStride); +void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst, + long width, long height, long src1Stride, + long src2Stride, long dstStride); +void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2, + uint8_t *dst1, uint8_t *dst2, + long width, long height, + long srcStride1, long srcStride2, + long dstStride1, long dstStride2); +void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, + uint8_t *dst, + long width, long height, + long srcStride1, long srcStride2, + long srcStride3, long dstStride); + +#if defined(ARCH_X86) || defined(ARCH_X86_64) +static const uint64_t mmx_null __attribute__((aligned(8))) = 0x0000000000000000ULL; +static const uint64_t mmx_one __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL; +static const uint64_t mask32b attribute_used __attribute__((aligned(8))) = 0x000000FF000000FFULL; +static const uint64_t mask32g attribute_used __attribute__((aligned(8))) = 0x0000FF000000FF00ULL; +static const uint64_t mask32r attribute_used __attribute__((aligned(8))) = 0x00FF000000FF0000ULL; +static const uint64_t mask32 __attribute__((aligned(8))) = 0x00FFFFFF00FFFFFFULL; +static const uint64_t mask3216br __attribute__((aligned(8)))=0x00F800F800F800F8ULL; +static const uint64_t mask3216g __attribute__((aligned(8)))=0x0000FC000000FC00ULL; +static const uint64_t mask3215g __attribute__((aligned(8)))=0x0000F8000000F800ULL; +static const uint64_t mul3216 __attribute__((aligned(8))) = 0x2000000420000004ULL; +static const uint64_t mul3215 __attribute__((aligned(8))) = 0x2000000820000008ULL; +static const uint64_t mask24b attribute_used __attribute__((aligned(8))) = 0x00FF0000FF0000FFULL; +static const uint64_t mask24g attribute_used __attribute__((aligned(8))) = 0xFF0000FF0000FF00ULL; +static const uint64_t mask24r attribute_used __attribute__((aligned(8))) = 0x0000FF0000FF0000ULL; +static const uint64_t mask24l __attribute__((aligned(8))) = 0x0000000000FFFFFFULL; +static const uint64_t mask24h __attribute__((aligned(8))) = 0x0000FFFFFF000000ULL; +static const uint64_t mask24hh __attribute__((aligned(8))) = 0xffff000000000000ULL; +static const uint64_t mask24hhh __attribute__((aligned(8))) = 0xffffffff00000000ULL; +static const uint64_t mask24hhhh __attribute__((aligned(8))) = 0xffffffffffff0000ULL; +static const uint64_t mask15b __attribute__((aligned(8))) = 0x001F001F001F001FULL; /* 00000000 00011111 xxB */ +static const uint64_t mask15rg __attribute__((aligned(8))) = 0x7FE07FE07FE07FE0ULL; /* 01111111 11100000 RGx */ +static const uint64_t mask15s __attribute__((aligned(8))) = 0xFFE0FFE0FFE0FFE0ULL; +static const uint64_t mask15g __attribute__((aligned(8))) = 0x03E003E003E003E0ULL; +static const uint64_t mask15r __attribute__((aligned(8))) = 0x7C007C007C007C00ULL; +#define mask16b mask15b +static const uint64_t mask16g __attribute__((aligned(8))) = 0x07E007E007E007E0ULL; +static const uint64_t mask16r __attribute__((aligned(8))) = 0xF800F800F800F800ULL; +static const uint64_t red_16mask __attribute__((aligned(8))) = 0x0000f8000000f800ULL; +static const uint64_t green_16mask __attribute__((aligned(8)))= 0x000007e0000007e0ULL; +static const uint64_t blue_16mask __attribute__((aligned(8))) = 0x0000001f0000001fULL; +static const uint64_t red_15mask __attribute__((aligned(8))) = 0x00007c000000f800ULL; +static const uint64_t green_15mask __attribute__((aligned(8)))= 0x000003e0000007e0ULL; +static const uint64_t blue_15mask __attribute__((aligned(8))) = 0x0000001f0000001fULL; + +#ifdef FAST_BGR2YV12 +static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL; +static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL; +static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL; +#else +static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL; +static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL; +static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL; +#endif +static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL; +static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL; +static const uint64_t w1111 attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL; + +#if 0 +static volatile uint64_t __attribute__((aligned(8))) b5Dither; +static volatile uint64_t __attribute__((aligned(8))) g5Dither; +static volatile uint64_t __attribute__((aligned(8))) g6Dither; +static volatile uint64_t __attribute__((aligned(8))) r5Dither; + +static uint64_t __attribute__((aligned(8))) dither4[2]={ + 0x0103010301030103LL, + 0x0200020002000200LL,}; + +static uint64_t __attribute__((aligned(8))) dither8[2]={ + 0x0602060206020602LL, + 0x0004000400040004LL,}; +#endif +#endif + +#define RGB2YUV_SHIFT 8 +#define BY ((int)( 0.098*(1<rgb16 Original by Strepto/Astral + ported to gcc & bugfixed : A'rpi + MMX2, 3DNOW optimization by Nick Kurshev + 32bit c version, and and&add trick by Michael Niedermayer +*/ + +void sws_rgb2rgb_init(int flags){ +#if defined(ARCH_X86) || defined(ARCH_X86_64) + if(flags & SWS_CPU_CAPS_MMX2){ + rgb15to16= rgb15to16_MMX2; + rgb15to24= rgb15to24_MMX2; + rgb15to32= rgb15to32_MMX2; + rgb16to24= rgb16to24_MMX2; + rgb16to32= rgb16to32_MMX2; + rgb16to15= rgb16to15_MMX2; + rgb24to16= rgb24to16_MMX2; + rgb24to15= rgb24to15_MMX2; + rgb24to32= rgb24to32_MMX2; + rgb32to16= rgb32to16_MMX2; + rgb32to15= rgb32to15_MMX2; + rgb32to24= rgb32to24_MMX2; + rgb24tobgr15= rgb24tobgr15_MMX2; + rgb24tobgr16= rgb24tobgr16_MMX2; + rgb24tobgr24= rgb24tobgr24_MMX2; + rgb32tobgr32= rgb32tobgr32_MMX2; + rgb32tobgr16= rgb32tobgr16_MMX2; + rgb32tobgr15= rgb32tobgr15_MMX2; + yv12toyuy2= yv12toyuy2_MMX2; + yv12touyvy= yv12touyvy_MMX2; + yuv422ptoyuy2= yuv422ptoyuy2_MMX2; + yuy2toyv12= yuy2toyv12_MMX2; +// uyvytoyv12= uyvytoyv12_MMX2; +// yvu9toyv12= yvu9toyv12_MMX2; + planar2x= planar2x_MMX2; + rgb24toyv12= rgb24toyv12_MMX2; + interleaveBytes= interleaveBytes_MMX2; + vu9_to_vu12= vu9_to_vu12_MMX2; + yvu9_to_yuy2= yvu9_to_yuy2_MMX2; + }else if(flags & SWS_CPU_CAPS_3DNOW){ + rgb15to16= rgb15to16_3DNOW; + rgb15to24= rgb15to24_3DNOW; + rgb15to32= rgb15to32_3DNOW; + rgb16to24= rgb16to24_3DNOW; + rgb16to32= rgb16to32_3DNOW; + rgb16to15= rgb16to15_3DNOW; + rgb24to16= rgb24to16_3DNOW; + rgb24to15= rgb24to15_3DNOW; + rgb24to32= rgb24to32_3DNOW; + rgb32to16= rgb32to16_3DNOW; + rgb32to15= rgb32to15_3DNOW; + rgb32to24= rgb32to24_3DNOW; + rgb24tobgr15= rgb24tobgr15_3DNOW; + rgb24tobgr16= rgb24tobgr16_3DNOW; + rgb24tobgr24= rgb24tobgr24_3DNOW; + rgb32tobgr32= rgb32tobgr32_3DNOW; + rgb32tobgr16= rgb32tobgr16_3DNOW; + rgb32tobgr15= rgb32tobgr15_3DNOW; + yv12toyuy2= yv12toyuy2_3DNOW; + yv12touyvy= yv12touyvy_3DNOW; + yuv422ptoyuy2= yuv422ptoyuy2_3DNOW; + yuy2toyv12= yuy2toyv12_3DNOW; +// uyvytoyv12= uyvytoyv12_3DNOW; +// yvu9toyv12= yvu9toyv12_3DNOW; + planar2x= planar2x_3DNOW; + rgb24toyv12= rgb24toyv12_3DNOW; + interleaveBytes= interleaveBytes_3DNOW; + vu9_to_vu12= vu9_to_vu12_3DNOW; + yvu9_to_yuy2= yvu9_to_yuy2_3DNOW; + }else if(flags & SWS_CPU_CAPS_MMX){ + rgb15to16= rgb15to16_MMX; + rgb15to24= rgb15to24_MMX; + rgb15to32= rgb15to32_MMX; + rgb16to24= rgb16to24_MMX; + rgb16to32= rgb16to32_MMX; + rgb16to15= rgb16to15_MMX; + rgb24to16= rgb24to16_MMX; + rgb24to15= rgb24to15_MMX; + rgb24to32= rgb24to32_MMX; + rgb32to16= rgb32to16_MMX; + rgb32to15= rgb32to15_MMX; + rgb32to24= rgb32to24_MMX; + rgb24tobgr15= rgb24tobgr15_MMX; + rgb24tobgr16= rgb24tobgr16_MMX; + rgb24tobgr24= rgb24tobgr24_MMX; + rgb32tobgr32= rgb32tobgr32_MMX; + rgb32tobgr16= rgb32tobgr16_MMX; + rgb32tobgr15= rgb32tobgr15_MMX; + yv12toyuy2= yv12toyuy2_MMX; + yv12touyvy= yv12touyvy_MMX; + yuv422ptoyuy2= yuv422ptoyuy2_MMX; + yuy2toyv12= yuy2toyv12_MMX; +// uyvytoyv12= uyvytoyv12_MMX; +// yvu9toyv12= yvu9toyv12_MMX; + planar2x= planar2x_MMX; + rgb24toyv12= rgb24toyv12_MMX; + interleaveBytes= interleaveBytes_MMX; + vu9_to_vu12= vu9_to_vu12_MMX; + yvu9_to_yuy2= yvu9_to_yuy2_MMX; + }else +#endif + { + rgb15to16= rgb15to16_C; + rgb15to24= rgb15to24_C; + rgb15to32= rgb15to32_C; + rgb16to24= rgb16to24_C; + rgb16to32= rgb16to32_C; + rgb16to15= rgb16to15_C; + rgb24to16= rgb24to16_C; + rgb24to15= rgb24to15_C; + rgb24to32= rgb24to32_C; + rgb32to16= rgb32to16_C; + rgb32to15= rgb32to15_C; + rgb32to24= rgb32to24_C; + rgb24tobgr15= rgb24tobgr15_C; + rgb24tobgr16= rgb24tobgr16_C; + rgb24tobgr24= rgb24tobgr24_C; + rgb32tobgr32= rgb32tobgr32_C; + rgb32tobgr16= rgb32tobgr16_C; + rgb32tobgr15= rgb32tobgr15_C; + yv12toyuy2= yv12toyuy2_C; + yv12touyvy= yv12touyvy_C; + yuv422ptoyuy2= yuv422ptoyuy2_C; + yuy2toyv12= yuy2toyv12_C; +// uyvytoyv12= uyvytoyv12_C; +// yvu9toyv12= yvu9toyv12_C; + planar2x= planar2x_C; + rgb24toyv12= rgb24toyv12_C; + interleaveBytes= interleaveBytes_C; + vu9_to_vu12= vu9_to_vu12_C; + yvu9_to_yuy2= yvu9_to_yuy2_C; + } +} + +/** + * Pallete is assumed to contain bgr32 + */ +void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette) +{ + long i; + +/* + for(i=0; i> 2; + for(i=0; i BGR24 (= B,G,R) */ + dst[3*i + 0] = src[4*i + 1]; + dst[3*i + 1] = src[4*i + 2]; + dst[3*i + 2] = src[4*i + 3]; + #else + dst[3*i + 0] = src[4*i + 2]; + dst[3*i + 1] = src[4*i + 1]; + dst[3*i + 2] = src[4*i + 0]; + #endif + } +} + +void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size) +{ + long i; + for(i=0; 3*i BGR32 (= A,R,G,B) */ + dst[4*i + 0] = 0; + dst[4*i + 1] = src[3*i + 0]; + dst[4*i + 2] = src[3*i + 1]; + dst[4*i + 3] = src[3*i + 2]; + #else + dst[4*i + 0] = src[3*i + 2]; + dst[4*i + 1] = src[3*i + 1]; + dst[4*i + 2] = src[3*i + 0]; + dst[4*i + 3] = 0; + #endif + } +} + +void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint16_t *end; + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (uint16_t *)src; + end = s + src_size/2; + while(s < end) + { + register uint16_t bgr; + bgr = *s++; + #ifdef WORDS_BIGENDIAN + *d++ = 0; + *d++ = (bgr&0x1F)<<3; + *d++ = (bgr&0x7E0)>>3; + *d++ = (bgr&0xF800)>>8; + #else + *d++ = (bgr&0xF800)>>8; + *d++ = (bgr&0x7E0)>>3; + *d++ = (bgr&0x1F)<<3; + *d++ = 0; + #endif + } +} + +void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint16_t *end; + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (const uint16_t *)src; + end = s + src_size/2; + while(s < end) + { + register uint16_t bgr; + bgr = *s++; + *d++ = (bgr&0xF800)>>8; + *d++ = (bgr&0x7E0)>>3; + *d++ = (bgr&0x1F)<<3; + } +} + +void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size) +{ + long i; + long num_pixels = src_size >> 1; + + for(i=0; i>5; + b = (rgb&0xF800)>>11; + dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11); + } +} + +void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size) +{ + long i; + long num_pixels = src_size >> 1; + + for(i=0; i>5; + b = (rgb&0xF800)>>11; + dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10); + } +} + +void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint16_t *end; + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (const uint16_t *)src; + end = s + src_size/2; + while(s < end) + { + register uint16_t bgr; + bgr = *s++; + #ifdef WORDS_BIGENDIAN + *d++ = 0; + *d++ = (bgr&0x1F)<<3; + *d++ = (bgr&0x3E0)>>2; + *d++ = (bgr&0x7C00)>>7; + #else + *d++ = (bgr&0x7C00)>>7; + *d++ = (bgr&0x3E0)>>2; + *d++ = (bgr&0x1F)<<3; + *d++ = 0; + #endif + } +} + +void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint16_t *end; + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (uint16_t *)src; + end = s + src_size/2; + while(s < end) + { + register uint16_t bgr; + bgr = *s++; + *d++ = (bgr&0x7C00)>>7; + *d++ = (bgr&0x3E0)>>2; + *d++ = (bgr&0x1F)<<3; + } +} + +void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size) +{ + long i; + long num_pixels = src_size >> 1; + + for(i=0; i>5; + b = (rgb&0x7C00)>>10; + dst[2*i] = (b&0x1F) | ((g&0x3F)<<5) | ((r&0x1F)<<11); + } +} + +void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size) +{ + long i; + long num_pixels = src_size >> 1; + + for(i=0; i>5; + b = (rgb&0x7C00)>>10; + dst[2*i] = (b&0x1F) | ((g&0x1F)<<5) | ((r&0x1F)<<10); + } +} + +void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size) +{ + long i; + long num_pixels = src_size; + for(i=0; i>3; + b = (rgb&0xC0)>>6; + dst[i] = ((b<<1)&0x07) | ((g&0x07)<<3) | ((r&0x03)<<6); + } +} diff --git a/libswscale/rgb2rgb.h b/libswscale/rgb2rgb.h new file mode 100644 index 0000000000..e38893deaf --- /dev/null +++ b/libswscale/rgb2rgb.h @@ -0,0 +1,144 @@ +/* + * + * rgb2rgb.h, Software RGB to RGB convertor + * pluralize by Software PAL8 to RGB convertor + * Software YUV to YUV convertor + * Software YUV to RGB convertor + */ + +#ifndef RGB2RGB_INCLUDED +#define RGB2RGB_INCLUDED + +// Note: do not fix the dependence on stdio.h + +/* A full collection of rgb to rgb(bgr) convertors */ +extern void (*rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb24to16)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb24to15)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb32to16)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb32to15)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb15to24)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb15to32)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb16to15)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb16to24)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb16to32)(const uint8_t *src,uint8_t *dst,long src_size); +extern void (*rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size); +extern void (*rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size); +extern void (*rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size); +extern void (*rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size); +extern void (*rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size); +extern void (*rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size); + +extern void rgb24tobgr32(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb32tobgr24(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb16tobgr32(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb16tobgr24(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb16tobgr16(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb16tobgr15(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb15tobgr32(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb15tobgr24(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb15tobgr16(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb15tobgr15(const uint8_t *src, uint8_t *dst, long src_size); +extern void rgb8tobgr8(const uint8_t *src, uint8_t *dst, long src_size); + + +extern void palette8torgb32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8tobgr32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8torgb24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8tobgr24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8torgb16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8tobgr16(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8torgb15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); +extern void palette8tobgr15(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette); + +/** + * + * height should be a multiple of 2 and width should be a multiple of 16 (if this is a + * problem for anyone then tell me, and ill fix it) + * chrominance data is only taken from every secound line others are ignored FIXME write HQ version + */ +//void uyvytoyv12(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + +/** + * + * height should be a multiple of 2 and width should be a multiple of 16 (if this is a + * problem for anyone then tell me, and ill fix it) + */ +extern void (*yv12toyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, + long width, long height, + long lumStride, long chromStride, long dstStride); + +/** + * + * width should be a multiple of 16 + */ +extern void (*yuv422ptoyuy2)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, + long width, long height, + long lumStride, long chromStride, long dstStride); + +/** + * + * height should be a multiple of 2 and width should be a multiple of 16 (if this is a + * problem for anyone then tell me, and ill fix it) + */ +extern void (*yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + long width, long height, + long lumStride, long chromStride, long srcStride); + +/** + * + * height should be a multiple of 2 and width should be a multiple of 16 (if this is a + * problem for anyone then tell me, and ill fix it) + */ +extern void (*yv12touyvy)(const uint8_t *ysrc, const uint8_t *usrc, const uint8_t *vsrc, uint8_t *dst, + long width, long height, + long lumStride, long chromStride, long dstStride); + +/** + * + * height should be a multiple of 2 and width should be a multiple of 2 (if this is a + * problem for anyone then tell me, and ill fix it) + * chrominance data is only taken from every secound line others are ignored FIXME write HQ version + */ +extern void (*rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + long width, long height, + long lumStride, long chromStride, long srcStride); +extern void (*planar2x)(const uint8_t *src, uint8_t *dst, long width, long height, + long srcStride, long dstStride); + +extern void (*interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dst, + long width, long height, long src1Stride, + long src2Stride, long dstStride); + +extern void (*vu9_to_vu12)(const uint8_t *src1, const uint8_t *src2, + uint8_t *dst1, uint8_t *dst2, + long width, long height, + long srcStride1, long srcStride2, + long dstStride1, long dstStride2); + +extern void (*yvu9_to_yuy2)(const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, + uint8_t *dst, + long width, long height, + long srcStride1, long srcStride2, + long srcStride3, long dstStride); + + +#define MODE_RGB 0x1 +#define MODE_BGR 0x2 + +static void yuv2rgb(uint8_t * image, uint8_t * py, + uint8_t * pu, uint8_t * pv, + unsigned h_size, unsigned v_size, + int rgb_stride, int y_stride, int uv_stride){ +printf("broken, this should use the swscaler\n"); +} + +static void yuv2rgb_init (unsigned bpp, int mode){ +printf("broken, this should use the swscaler\n"); +} + +void sws_rgb2rgb_init(int flags); + +#endif diff --git a/libswscale/rgb2rgb_template.c b/libswscale/rgb2rgb_template.c new file mode 100644 index 0000000000..807da6166f --- /dev/null +++ b/libswscale/rgb2rgb_template.c @@ -0,0 +1,2675 @@ +/* + * + * rgb2rgb.c, Software RGB to RGB convertor + * pluralize by Software PAL8 to RGB convertor + * Software YUV to YUV convertor + * Software YUV to RGB convertor + * Written by Nick Kurshev. + * palette & yuv & runtime cpu stuff by Michael (michaelni@gmx.at) (under GPL) + * lot of big-endian byteorder fixes by Alex Beregszaszi + */ + +#include +#include /* for __WORDSIZE */ + +#include "asmalign.h" + +#ifndef __WORDSIZE +// #warning You have misconfigured system and probably will lose performance! +#define __WORDSIZE MP_WORDSIZE +#endif + +#undef PREFETCH +#undef MOVNTQ +#undef EMMS +#undef SFENCE +#undef MMREG_SIZE +#undef PREFETCHW +#undef PAVGB + +#ifdef HAVE_SSE2 +#define MMREG_SIZE 16 +#else +#define MMREG_SIZE 8 +#endif + +#ifdef HAVE_3DNOW +#define PREFETCH "prefetch" +#define PREFETCHW "prefetchw" +#define PAVGB "pavgusb" +#elif defined ( HAVE_MMX2 ) +#define PREFETCH "prefetchnta" +#define PREFETCHW "prefetcht0" +#define PAVGB "pavgb" +#else +#ifdef __APPLE__ +#define PREFETCH "#" +#define PREFETCHW "#" +#else +#define PREFETCH "/nop" +#define PREFETCHW "/nop" +#endif +#endif + +#ifdef HAVE_3DNOW +/* On K6 femms is faster of emms. On K7 femms is directly mapped on emms. */ +#define EMMS "femms" +#else +#define EMMS "emms" +#endif + +#ifdef HAVE_MMX2 +#define MOVNTQ "movntq" +#define SFENCE "sfence" +#else +#define MOVNTQ "movq" +#ifdef __APPLE__ +#define SFENCE "#" +#else +#define SFENCE "/nop" +#endif +#endif + +static inline void RENAME(rgb24to32)(const uint8_t *src,uint8_t *dst,long src_size) +{ + uint8_t *dest = dst; + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*s):"memory"); + mm_end = end - 23; + __asm __volatile("movq %0, %%mm7"::"m"(mask32):"memory"); + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "punpckldq 3%1, %%mm0\n\t" + "movd 6%1, %%mm1\n\t" + "punpckldq 9%1, %%mm1\n\t" + "movd 12%1, %%mm2\n\t" + "punpckldq 15%1, %%mm2\n\t" + "movd 18%1, %%mm3\n\t" + "punpckldq 21%1, %%mm3\n\t" + "pand %%mm7, %%mm0\n\t" + "pand %%mm7, %%mm1\n\t" + "pand %%mm7, %%mm2\n\t" + "pand %%mm7, %%mm3\n\t" + MOVNTQ" %%mm0, %0\n\t" + MOVNTQ" %%mm1, 8%0\n\t" + MOVNTQ" %%mm2, 16%0\n\t" + MOVNTQ" %%mm3, 24%0" + :"=m"(*dest) + :"m"(*s) + :"memory"); + dest += 32; + s += 24; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { +#ifdef WORDS_BIGENDIAN + /* RGB24 (= R,G,B) -> RGB32 (= A,B,G,R) */ + *dest++ = 0; + *dest++ = s[2]; + *dest++ = s[1]; + *dest++ = s[0]; + s+=3; +#else + *dest++ = *s++; + *dest++ = *s++; + *dest++ = *s++; + *dest++ = 0; +#endif + } +} + +static inline void RENAME(rgb32to24)(const uint8_t *src,uint8_t *dst,long src_size) +{ + uint8_t *dest = dst; + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*s):"memory"); + mm_end = end - 31; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movq %1, %%mm0\n\t" + "movq 8%1, %%mm1\n\t" + "movq 16%1, %%mm4\n\t" + "movq 24%1, %%mm5\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm1, %%mm3\n\t" + "movq %%mm4, %%mm6\n\t" + "movq %%mm5, %%mm7\n\t" + "psrlq $8, %%mm2\n\t" + "psrlq $8, %%mm3\n\t" + "psrlq $8, %%mm6\n\t" + "psrlq $8, %%mm7\n\t" + "pand %2, %%mm0\n\t" + "pand %2, %%mm1\n\t" + "pand %2, %%mm4\n\t" + "pand %2, %%mm5\n\t" + "pand %3, %%mm2\n\t" + "pand %3, %%mm3\n\t" + "pand %3, %%mm6\n\t" + "pand %3, %%mm7\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm3, %%mm1\n\t" + "por %%mm6, %%mm4\n\t" + "por %%mm7, %%mm5\n\t" + + "movq %%mm1, %%mm2\n\t" + "movq %%mm4, %%mm3\n\t" + "psllq $48, %%mm2\n\t" + "psllq $32, %%mm3\n\t" + "pand %4, %%mm2\n\t" + "pand %5, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "psrlq $16, %%mm1\n\t" + "psrlq $32, %%mm4\n\t" + "psllq $16, %%mm5\n\t" + "por %%mm3, %%mm1\n\t" + "pand %6, %%mm5\n\t" + "por %%mm5, %%mm4\n\t" + + MOVNTQ" %%mm0, %0\n\t" + MOVNTQ" %%mm1, 8%0\n\t" + MOVNTQ" %%mm4, 16%0" + :"=m"(*dest) + :"m"(*s),"m"(mask24l), + "m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) + :"memory"); + dest += 24; + s += 32; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { +#ifdef WORDS_BIGENDIAN + /* RGB32 (= A,B,G,R) -> RGB24 (= R,G,B) */ + s++; + dest[2] = *s++; + dest[1] = *s++; + dest[0] = *s++; + dest += 3; +#else + *dest++ = *s++; + *dest++ = *s++; + *dest++ = *s++; + s++; +#endif + } +} + +/* + Original by Strepto/Astral + ported to gcc & bugfixed : A'rpi + MMX2, 3DNOW optimization by Nick Kurshev + 32bit c version, and and&add trick by Michael Niedermayer +*/ +static inline void RENAME(rgb15to16)(const uint8_t *src,uint8_t *dst,long src_size) +{ + register const uint8_t* s=src; + register uint8_t* d=dst; + register const uint8_t *end; + const uint8_t *mm_end; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*s)); + __asm __volatile("movq %0, %%mm4"::"m"(mask15s)); + mm_end = end - 15; + while(s>1)&0x7FE07FE0) | (x&0x001F001F); + s+=4; + d+=4; + } + if(s < end) + { + register uint16_t x= *((uint16_t *)s); + *((uint16_t *)d) = ((x>>1)&0x7FE0) | (x&0x001F); + s+=2; + d+=2; + } +} + +static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + mm_end = end - 15; +#if 1 //is faster only if multiplies are reasonable fast (FIXME figure out on which cpus this is faster, on Athlon its slightly faster) + asm volatile( + "movq %3, %%mm5 \n\t" + "movq %4, %%mm6 \n\t" + "movq %5, %%mm7 \n\t" + ASMALIGN16 + "1: \n\t" + PREFETCH" 32(%1) \n\t" + "movd (%1), %%mm0 \n\t" + "movd 4(%1), %%mm3 \n\t" + "punpckldq 8(%1), %%mm0 \n\t" + "punpckldq 12(%1), %%mm3 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm3, %%mm4 \n\t" + "pand %%mm6, %%mm0 \n\t" + "pand %%mm6, %%mm3 \n\t" + "pmaddwd %%mm7, %%mm0 \n\t" + "pmaddwd %%mm7, %%mm3 \n\t" + "pand %%mm5, %%mm1 \n\t" + "pand %%mm5, %%mm4 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm4, %%mm3 \n\t" + "psrld $5, %%mm0 \n\t" + "pslld $11, %%mm3 \n\t" + "por %%mm3, %%mm0 \n\t" + MOVNTQ" %%mm0, (%0) \n\t" + "add $16, %1 \n\t" + "add $8, %0 \n\t" + "cmp %2, %1 \n\t" + " jb 1b \n\t" + : "+r" (d), "+r"(s) + : "r" (mm_end), "m" (mask3216g), "m" (mask3216br), "m" (mul3216) + ); +#else + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_16mask),"m"(green_16mask)); + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 4%1, %%mm3\n\t" + "punpckldq 8%1, %%mm0\n\t" + "punpckldq 12%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psrlq $3, %%mm0\n\t" + "psrlq $3, %%mm3\n\t" + "pand %2, %%mm0\n\t" + "pand %2, %%mm3\n\t" + "psrlq $5, %%mm1\n\t" + "psrlq $5, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $8, %%mm2\n\t" + "psrlq $8, %%mm5\n\t" + "pand %%mm7, %%mm2\n\t" + "pand %%mm7, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory"); + d += 4; + s += 16; + } +#endif + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + register int rgb = *(uint32_t*)s; s += 4; + *d++ = ((rgb&0xFF)>>3) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>8); + } +} + +static inline void RENAME(rgb32tobgr16)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_16mask),"m"(green_16mask)); + mm_end = end - 15; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 4%1, %%mm3\n\t" + "punpckldq 8%1, %%mm0\n\t" + "punpckldq 12%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psllq $8, %%mm0\n\t" + "psllq $8, %%mm3\n\t" + "pand %%mm7, %%mm0\n\t" + "pand %%mm7, %%mm3\n\t" + "psrlq $5, %%mm1\n\t" + "psrlq $5, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $19, %%mm2\n\t" + "psrlq $19, %%mm5\n\t" + "pand %2, %%mm2\n\t" + "pand %2, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory"); + d += 4; + s += 16; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + register int rgb = *(uint32_t*)s; s += 4; + *d++ = ((rgb&0xF8)<<8) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>19); + } +} + +static inline void RENAME(rgb32to15)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + mm_end = end - 15; +#if 1 //is faster only if multiplies are reasonable fast (FIXME figure out on which cpus this is faster, on Athlon its slightly faster) + asm volatile( + "movq %3, %%mm5 \n\t" + "movq %4, %%mm6 \n\t" + "movq %5, %%mm7 \n\t" + ASMALIGN16 + "1: \n\t" + PREFETCH" 32(%1) \n\t" + "movd (%1), %%mm0 \n\t" + "movd 4(%1), %%mm3 \n\t" + "punpckldq 8(%1), %%mm0 \n\t" + "punpckldq 12(%1), %%mm3 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm3, %%mm4 \n\t" + "pand %%mm6, %%mm0 \n\t" + "pand %%mm6, %%mm3 \n\t" + "pmaddwd %%mm7, %%mm0 \n\t" + "pmaddwd %%mm7, %%mm3 \n\t" + "pand %%mm5, %%mm1 \n\t" + "pand %%mm5, %%mm4 \n\t" + "por %%mm1, %%mm0 \n\t" + "por %%mm4, %%mm3 \n\t" + "psrld $6, %%mm0 \n\t" + "pslld $10, %%mm3 \n\t" + "por %%mm3, %%mm0 \n\t" + MOVNTQ" %%mm0, (%0) \n\t" + "add $16, %1 \n\t" + "add $8, %0 \n\t" + "cmp %2, %1 \n\t" + " jb 1b \n\t" + : "+r" (d), "+r"(s) + : "r" (mm_end), "m" (mask3215g), "m" (mask3216br), "m" (mul3215) + ); +#else + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_15mask),"m"(green_15mask)); + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 4%1, %%mm3\n\t" + "punpckldq 8%1, %%mm0\n\t" + "punpckldq 12%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psrlq $3, %%mm0\n\t" + "psrlq $3, %%mm3\n\t" + "pand %2, %%mm0\n\t" + "pand %2, %%mm3\n\t" + "psrlq $6, %%mm1\n\t" + "psrlq $6, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $9, %%mm2\n\t" + "psrlq $9, %%mm5\n\t" + "pand %%mm7, %%mm2\n\t" + "pand %%mm7, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory"); + d += 4; + s += 16; + } +#endif + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + register int rgb = *(uint32_t*)s; s += 4; + *d++ = ((rgb&0xFF)>>3) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>9); + } +} + +static inline void RENAME(rgb32tobgr15)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_15mask),"m"(green_15mask)); + mm_end = end - 15; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 4%1, %%mm3\n\t" + "punpckldq 8%1, %%mm0\n\t" + "punpckldq 12%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psllq $7, %%mm0\n\t" + "psllq $7, %%mm3\n\t" + "pand %%mm7, %%mm0\n\t" + "pand %%mm7, %%mm3\n\t" + "psrlq $6, %%mm1\n\t" + "psrlq $6, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $19, %%mm2\n\t" + "psrlq $19, %%mm5\n\t" + "pand %2, %%mm2\n\t" + "pand %2, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory"); + d += 4; + s += 16; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + register int rgb = *(uint32_t*)s; s += 4; + *d++ = ((rgb&0xF8)<<7) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>19); + } +} + +static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_16mask),"m"(green_16mask)); + mm_end = end - 11; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 3%1, %%mm3\n\t" + "punpckldq 6%1, %%mm0\n\t" + "punpckldq 9%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psrlq $3, %%mm0\n\t" + "psrlq $3, %%mm3\n\t" + "pand %2, %%mm0\n\t" + "pand %2, %%mm3\n\t" + "psrlq $5, %%mm1\n\t" + "psrlq $5, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $8, %%mm2\n\t" + "psrlq $8, %%mm5\n\t" + "pand %%mm7, %%mm2\n\t" + "pand %%mm7, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory"); + d += 4; + s += 12; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + const int b= *s++; + const int g= *s++; + const int r= *s++; + *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8); + } +} + +static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_16mask),"m"(green_16mask)); + mm_end = end - 15; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 3%1, %%mm3\n\t" + "punpckldq 6%1, %%mm0\n\t" + "punpckldq 9%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psllq $8, %%mm0\n\t" + "psllq $8, %%mm3\n\t" + "pand %%mm7, %%mm0\n\t" + "pand %%mm7, %%mm3\n\t" + "psrlq $5, %%mm1\n\t" + "psrlq $5, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $19, %%mm2\n\t" + "psrlq $19, %%mm5\n\t" + "pand %2, %%mm2\n\t" + "pand %2, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_16mask):"memory"); + d += 4; + s += 12; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + const int r= *s++; + const int g= *s++; + const int b= *s++; + *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8); + } +} + +static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_15mask),"m"(green_15mask)); + mm_end = end - 11; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 3%1, %%mm3\n\t" + "punpckldq 6%1, %%mm0\n\t" + "punpckldq 9%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psrlq $3, %%mm0\n\t" + "psrlq $3, %%mm3\n\t" + "pand %2, %%mm0\n\t" + "pand %2, %%mm3\n\t" + "psrlq $6, %%mm1\n\t" + "psrlq $6, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $9, %%mm2\n\t" + "psrlq $9, %%mm5\n\t" + "pand %%mm7, %%mm2\n\t" + "pand %%mm7, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory"); + d += 4; + s += 12; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + const int b= *s++; + const int g= *s++; + const int r= *s++; + *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7); + } +} + +static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint8_t *s = src; + const uint8_t *end; +#ifdef HAVE_MMX + const uint8_t *mm_end; +#endif + uint16_t *d = (uint16_t *)dst; + end = s + src_size; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*src):"memory"); + __asm __volatile( + "movq %0, %%mm7\n\t" + "movq %1, %%mm6\n\t" + ::"m"(red_15mask),"m"(green_15mask)); + mm_end = end - 15; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movd %1, %%mm0\n\t" + "movd 3%1, %%mm3\n\t" + "punpckldq 6%1, %%mm0\n\t" + "punpckldq 9%1, %%mm3\n\t" + "movq %%mm0, %%mm1\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm3, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "psllq $7, %%mm0\n\t" + "psllq $7, %%mm3\n\t" + "pand %%mm7, %%mm0\n\t" + "pand %%mm7, %%mm3\n\t" + "psrlq $6, %%mm1\n\t" + "psrlq $6, %%mm4\n\t" + "pand %%mm6, %%mm1\n\t" + "pand %%mm6, %%mm4\n\t" + "psrlq $19, %%mm2\n\t" + "psrlq $19, %%mm5\n\t" + "pand %2, %%mm2\n\t" + "pand %2, %%mm5\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm5, %%mm3\n\t" + "psllq $16, %%mm3\n\t" + "por %%mm3, %%mm0\n\t" + MOVNTQ" %%mm0, %0\n\t" + :"=m"(*d):"m"(*s),"m"(blue_15mask):"memory"); + d += 4; + s += 12; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + const int r= *s++; + const int g= *s++; + const int b= *s++; + *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7); + } +} + +/* + I use here less accurate approximation by simply + left-shifting the input + value and filling the low order bits with + zeroes. This method improves png's + compression but this scheme cannot reproduce white exactly, since it does not + generate an all-ones maximum value; the net effect is to darken the + image slightly. + + The better method should be "left bit replication": + + 4 3 2 1 0 + --------- + 1 1 0 1 1 + + 7 6 5 4 3 2 1 0 + ---------------- + 1 1 0 1 1 1 1 0 + |=======| |===| + | Leftmost Bits Repeated to Fill Open Bits + | + Original Bits +*/ +static inline void RENAME(rgb15to24)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint16_t *end; +#ifdef HAVE_MMX + const uint16_t *mm_end; +#endif + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (uint16_t *)src; + end = s + src_size/2; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*s):"memory"); + mm_end = end - 7; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movq %1, %%mm0\n\t" + "movq %1, %%mm1\n\t" + "movq %1, %%mm2\n\t" + "pand %2, %%mm0\n\t" + "pand %3, %%mm1\n\t" + "pand %4, %%mm2\n\t" + "psllq $3, %%mm0\n\t" + "psrlq $2, %%mm1\n\t" + "psrlq $7, %%mm2\n\t" + "movq %%mm0, %%mm3\n\t" + "movq %%mm1, %%mm4\n\t" + "movq %%mm2, %%mm5\n\t" + "punpcklwd %5, %%mm0\n\t" + "punpcklwd %5, %%mm1\n\t" + "punpcklwd %5, %%mm2\n\t" + "punpckhwd %5, %%mm3\n\t" + "punpckhwd %5, %%mm4\n\t" + "punpckhwd %5, %%mm5\n\t" + "psllq $8, %%mm1\n\t" + "psllq $16, %%mm2\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm2, %%mm0\n\t" + "psllq $8, %%mm4\n\t" + "psllq $16, %%mm5\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm5, %%mm3\n\t" + + "movq %%mm0, %%mm6\n\t" + "movq %%mm3, %%mm7\n\t" + + "movq 8%1, %%mm0\n\t" + "movq 8%1, %%mm1\n\t" + "movq 8%1, %%mm2\n\t" + "pand %2, %%mm0\n\t" + "pand %3, %%mm1\n\t" + "pand %4, %%mm2\n\t" + "psllq $3, %%mm0\n\t" + "psrlq $2, %%mm1\n\t" + "psrlq $7, %%mm2\n\t" + "movq %%mm0, %%mm3\n\t" + "movq %%mm1, %%mm4\n\t" + "movq %%mm2, %%mm5\n\t" + "punpcklwd %5, %%mm0\n\t" + "punpcklwd %5, %%mm1\n\t" + "punpcklwd %5, %%mm2\n\t" + "punpckhwd %5, %%mm3\n\t" + "punpckhwd %5, %%mm4\n\t" + "punpckhwd %5, %%mm5\n\t" + "psllq $8, %%mm1\n\t" + "psllq $16, %%mm2\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm2, %%mm0\n\t" + "psllq $8, %%mm4\n\t" + "psllq $16, %%mm5\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm5, %%mm3\n\t" + + :"=m"(*d) + :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r), "m"(mmx_null) + :"memory"); + /* Borrowed 32 to 24 */ + __asm __volatile( + "movq %%mm0, %%mm4\n\t" + "movq %%mm3, %%mm5\n\t" + "movq %%mm6, %%mm0\n\t" + "movq %%mm7, %%mm1\n\t" + + "movq %%mm4, %%mm6\n\t" + "movq %%mm5, %%mm7\n\t" + "movq %%mm0, %%mm2\n\t" + "movq %%mm1, %%mm3\n\t" + + "psrlq $8, %%mm2\n\t" + "psrlq $8, %%mm3\n\t" + "psrlq $8, %%mm6\n\t" + "psrlq $8, %%mm7\n\t" + "pand %2, %%mm0\n\t" + "pand %2, %%mm1\n\t" + "pand %2, %%mm4\n\t" + "pand %2, %%mm5\n\t" + "pand %3, %%mm2\n\t" + "pand %3, %%mm3\n\t" + "pand %3, %%mm6\n\t" + "pand %3, %%mm7\n\t" + "por %%mm2, %%mm0\n\t" + "por %%mm3, %%mm1\n\t" + "por %%mm6, %%mm4\n\t" + "por %%mm7, %%mm5\n\t" + + "movq %%mm1, %%mm2\n\t" + "movq %%mm4, %%mm3\n\t" + "psllq $48, %%mm2\n\t" + "psllq $32, %%mm3\n\t" + "pand %4, %%mm2\n\t" + "pand %5, %%mm3\n\t" + "por %%mm2, %%mm0\n\t" + "psrlq $16, %%mm1\n\t" + "psrlq $32, %%mm4\n\t" + "psllq $16, %%mm5\n\t" + "por %%mm3, %%mm1\n\t" + "pand %6, %%mm5\n\t" + "por %%mm5, %%mm4\n\t" + + MOVNTQ" %%mm0, %0\n\t" + MOVNTQ" %%mm1, 8%0\n\t" + MOVNTQ" %%mm4, 16%0" + + :"=m"(*d) + :"m"(*s),"m"(mask24l),"m"(mask24h),"m"(mask24hh),"m"(mask24hhh),"m"(mask24hhhh) + :"memory"); + d += 24; + s += 8; + } + __asm __volatile(SFENCE:::"memory"); + __asm __volatile(EMMS:::"memory"); +#endif + while(s < end) + { + register uint16_t bgr; + bgr = *s++; + *d++ = (bgr&0x1F)<<3; + *d++ = (bgr&0x3E0)>>2; + *d++ = (bgr&0x7C00)>>7; + } +} + +static inline void RENAME(rgb16to24)(const uint8_t *src, uint8_t *dst, long src_size) +{ + const uint16_t *end; +#ifdef HAVE_MMX + const uint16_t *mm_end; +#endif + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (const uint16_t *)src; + end = s + src_size/2; +#ifdef HAVE_MMX + __asm __volatile(PREFETCH" %0"::"m"(*s):"memory"); + mm_end = end - 7; + while(s < mm_end) + { + __asm __volatile( + PREFETCH" 32%1\n\t" + "movq %1, %%mm0\n\t" + "movq %1, %%mm1\n\t" + "movq %1, %%mm2\n\t" + "pand %2, %%mm0\n\t" + "pand %3, %%mm1\n\t" + "pand %4, %%mm2\n\t" + "psllq $3, %%mm0\n\t" + "psrlq $3, %%mm1\n\t" + "psrlq $8, %%mm2\n\t" + "movq %%mm0, %%mm3\n\t" + "movq %%mm1, %%mm4\n\t" + "movq %%mm2, %%mm5\n\t" + "punpcklwd %5, %%mm0\n\t" + "punpcklwd %5, %%mm1\n\t" + "punpcklwd %5, %%mm2\n\t" + "punpckhwd %5, %%mm3\n\t" + "punpckhwd %5, %%mm4\n\t" + "punpckhwd %5, %%mm5\n\t" + "psllq $8, %%mm1\n\t" + "psllq $16, %%mm2\n\t" + "por %%mm1, %%mm0\n\t" + "por %%mm2, %%mm0\n\t" + "psllq $8, %%mm4\n\t" + "psllq $16, %%mm5\n\t" + "por %%mm4, %%mm3\n\t" + "por %%mm5, %%mm3\n\t" + + "movq %%mm0, %%mm6\n\t" + "movq %%mm3, %%mm7\n\t" + + "movq 8%1, %%mm0\n\t" + "movq 8%1, %%mm1\n\t" + "movq 8%1, %%mm2\n\t" + "pand %2, %%mm0\n\t" + "pand %3, %%mm1\n\t" + "pand %4, %%mm2\n\t" + "psl