X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libswscale%2Frgb2rgb_template.c;h=d1a43e01cbfe1c67a4c8a1bf612c84b92c84ef70;hb=791de61bbb0d2bceb1037597b310e2a4a94494fd;hp=e95b6280490c684aef6c664bde98d32d4ed03d66;hpb=f8a138be5257f751ef7d3c6b7ab534c0434e90e7;p=ffmpeg diff --git a/libswscale/rgb2rgb_template.c b/libswscale/rgb2rgb_template.c index e95b6280490..d1a43e01cbf 100644 --- a/libswscale/rgb2rgb_template.c +++ b/libswscale/rgb2rgb_template.c @@ -7,1997 +7,587 @@ * palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at) * lot of big-endian byte order fixes by Alex Beregszaszi * - * This file is part of FFmpeg. + * This file is part of Libav. * - * FFmpeg 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. + * Libav is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. * - * FFmpeg is distributed in the hope that it will be useful, + * Libav is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. * - * You should have received a copy of the GNU General Public License - * along with FFmpeg; if not, write to the Free Software + * You should have received a copy of the GNU Lesser General Public + * License along with Libav; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA - * - * The C code (not assembly, MMX, ...) of this file can be used - * under the LGPL license. */ #include -#undef PREFETCH -#undef MOVNTQ -#undef EMMS -#undef SFENCE -#undef MMREG_SIZE -#undef PREFETCHW -#undef PAVGB - -#if HAVE_SSE2 -#define MMREG_SIZE 16 -#else -#define MMREG_SIZE 8 -#endif - -#if HAVE_AMD3DNOW -#define PREFETCH "prefetch" -#define PREFETCHW "prefetchw" -#define PAVGB "pavgusb" -#elif HAVE_MMX2 -#define PREFETCH "prefetchnta" -#define PREFETCHW "prefetcht0" -#define PAVGB "pavgb" -#else -#define PREFETCH " # nop" -#define PREFETCHW " # nop" -#endif - -#if HAVE_AMD3DNOW -/* On K6 femms is faster than emms. On K7 femms is directly mapped to emms. */ -#define EMMS "femms" -#else -#define EMMS "emms" -#endif - -#if HAVE_MMX2 -#define MOVNTQ "movntq" -#define SFENCE "sfence" +static inline void rgb24tobgr32_c(const uint8_t *src, uint8_t *dst, + int src_size) +{ + uint8_t *dest = dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { +#if HAVE_BIGENDIAN + /* RGB24 (= R, G, B) -> RGB32 (= A, B, G, R) */ + *dest++ = 255; + *dest++ = s[2]; + *dest++ = s[1]; + *dest++ = s[0]; + s += 3; #else -#define MOVNTQ "movq" -#define SFENCE " # nop" + *dest++ = *s++; + *dest++ = *s++; + *dest++ = *s++; + *dest++ = 255; #endif - -static inline void RENAME(rgb24tobgr32)(const uint8_t *src, uint8_t *dst, long src_size) -{ - uint8_t *dest = dst; - const uint8_t *s = src; - const uint8_t *end; - #if HAVE_MMX - const uint8_t *mm_end; - #endif - end = s + src_size; - #if HAVE_MMX - __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory"); - mm_end = end - 23; - __asm__ volatile("movq %0, %%mm7"::"m"(mask32a):"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" - "por %%mm7, %%mm0 \n\t" - "por %%mm7, %%mm1 \n\t" - "por %%mm7, %%mm2 \n\t" - "por %%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++ = 255; - *dest++ = s[2]; - *dest++ = s[1]; - *dest++ = s[0]; - s+=3; - #else - *dest++ = *s++; - *dest++ = *s++; - *dest++ = *s++; - *dest++ = 255; - #endif } } -static inline void RENAME(rgb32tobgr24)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb32tobgr24_c(const uint8_t *src, uint8_t *dst, + int src_size) { - uint8_t *dest = dst; - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - end = s + src_size; -#if 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) */ + uint8_t *dest = dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { +#if HAVE_BIGENDIAN + /* RGB32 (= A, B, G, R) -> RGB24 (= R, G, B) */ s++; - dest[2] = *s++; - dest[1] = *s++; - dest[0] = *s++; - dest += 3; + dest[2] = *s++; + dest[1] = *s++; + dest[0] = *s++; + dest += 3; #else - *dest++ = *s++; - *dest++ = *s++; - *dest++ = *s++; + *dest++ = *s++; + *dest++ = *s++; + *dest++ = *s++; s++; #endif } } /* - original by Strepto/Astral - ported to gcc & bugfixed: A'rpi - MMX2, 3DNOW optimization by Nick Kurshev - 32-bit C version, and and&add trick by Michael Niedermayer -*/ -static inline void RENAME(rgb15to16)(const uint8_t *src, uint8_t *dst, long src_size) + * original by Strepto/Astral + * ported to gcc & bugfixed: A'rpi + * MMX2, 3DNOW optimization by Nick Kurshev + * 32-bit C version, and and&add trick by Michael Niedermayer + */ +static inline void rgb15to16_c(const uint8_t *src, uint8_t *dst, int 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; -#if 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; + register uint8_t *d = dst; + register const uint8_t *s = src; + register const uint8_t *end = s + src_size; + const uint8_t *mm_end = end - 3; + + while (s < mm_end) { + register uint32_t x = *((const uint32_t *)s); + *((uint32_t *)d) = ((x >> 1) & 0x7FE07FE0) | (x & 0x001F001F); + s += 4; + d += 4; } - if (s < end) - { - register uint16_t x= *((const uint16_t*)s); - *((uint16_t *)d) = ((x>>1)&0x7FE0) | (x&0x001F); - s+=2; - d+=2; + if (s < end) { + register uint16_t x = *((const uint16_t *)s); + *((uint16_t *)d) = ((x >> 1) & 0x7FE0) | (x & 0x001F); } } -static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb32to16_c(const uint8_t *src, uint8_t *dst, int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if HAVE_MMX - mm_end = end - 15; -#if 1 //is faster only if multiplies are reasonably fast (FIXME figure out on which CPUs this is faster, on Athlon it is slightly faster) - __asm__ volatile( - "movq %3, %%mm5 \n\t" - "movq %4, %%mm6 \n\t" - "movq %5, %%mm7 \n\t" - "jmp 2f \n\t" - ASMALIGN(4) - "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" - "2: \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 = *(const uint32_t*)s; s += 4; - *d++ = ((rgb&0xFF)>>3) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>8); + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { + register int rgb = *(const 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) +static inline void rgb32tobgr16_c(const uint8_t *src, uint8_t *dst, + int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if 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 = *(const uint32_t*)s; s += 4; - *d++ = ((rgb&0xF8)<<8) + ((rgb&0xFC00)>>5) + ((rgb&0xF80000)>>19); + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { + register int rgb = *(const 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) +static inline void rgb32to15_c(const uint8_t *src, uint8_t *dst, int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if HAVE_MMX - mm_end = end - 15; -#if 1 //is faster only if multiplies are reasonably fast (FIXME figure out on which CPUs this is faster, on Athlon it is slightly faster) - __asm__ volatile( - "movq %3, %%mm5 \n\t" - "movq %4, %%mm6 \n\t" - "movq %5, %%mm7 \n\t" - "jmp 2f \n\t" - ASMALIGN(4) - "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" - "2: \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 = *(const uint32_t*)s; s += 4; - *d++ = ((rgb&0xFF)>>3) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>9); + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { + register int rgb = *(const 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) +static inline void rgb32tobgr15_c(const uint8_t *src, uint8_t *dst, + int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if 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 = *(const uint32_t*)s; s += 4; - *d++ = ((rgb&0xF8)<<7) + ((rgb&0xF800)>>6) + ((rgb&0xF80000)>>19); + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { + register int rgb = *(const uint32_t *)s; + s += 4; + *d++ = ((rgb & 0xF8) << 7) + + ((rgb & 0xF800) >> 6) + + ((rgb & 0xF80000) >> 19); } } -static inline void RENAME(rgb24tobgr16)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb24tobgr16_c(const uint8_t *src, uint8_t *dst, + int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if 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) - { + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { const int b = *s++; const int g = *s++; const int r = *s++; - *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8); + *d++ = (b >> 3) | ((g & 0xFC) << 3) | ((r & 0xF8) << 8); } } -static inline void RENAME(rgb24to16)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb24to16_c(const uint8_t *src, uint8_t *dst, int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if 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) - { + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { const int r = *s++; const int g = *s++; const int b = *s++; - *d++ = (b>>3) | ((g&0xFC)<<3) | ((r&0xF8)<<8); + *d++ = (b >> 3) | ((g & 0xFC) << 3) | ((r & 0xF8) << 8); } } -static inline void RENAME(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb24tobgr15_c(const uint8_t *src, uint8_t *dst, + int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if 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) - { + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { const int b = *s++; const int g = *s++; const int r = *s++; - *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7); + *d++ = (b >> 3) | ((g & 0xF8) << 2) | ((r & 0xF8) << 7); } } -static inline void RENAME(rgb24to15)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb24to15_c(const uint8_t *src, uint8_t *dst, int src_size) { - const uint8_t *s = src; - const uint8_t *end; -#if HAVE_MMX - const uint8_t *mm_end; -#endif - uint16_t *d = (uint16_t *)dst; - end = s + src_size; -#if 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) - { + uint16_t *d = (uint16_t *)dst; + const uint8_t *s = src; + const uint8_t *end = s + src_size; + + while (s < end) { const int r = *s++; const int g = *s++; const int b = *s++; - *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7); + *d++ = (b >> 3) | ((g & 0xF8) << 2) | ((r & 0xF8) << 7); } } /* - I use less accurate approximation here by simply left-shifting the input - value and filling the low order bits with zeroes. This method improves PNG - 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(rgb15tobgr24)(const uint8_t *src, uint8_t *dst, long src_size) + * I use less accurate approximation here by simply left-shifting the input + * value and filling the low order bits with zeroes. This method improves PNG + * 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 rgb15tobgr24_c(const uint8_t *src, uint8_t *dst, + int src_size) { - const uint16_t *end; -#if HAVE_MMX - const uint16_t *mm_end; -#endif - uint8_t *d = dst; - const uint16_t *s = (const uint16_t*)src; - end = s + src_size/2; -#if 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; + uint8_t *d = dst; + const uint16_t *s = (const uint16_t *)src; + const uint16_t *end = s + src_size / 2; + + while (s < end) { + register uint16_t bgr = *s++; + *d++ = (bgr & 0x1F) << 3; + *d++ = (bgr & 0x3E0) >> 2; + *d++ = (bgr & 0x7C00) >> 7; } } -static inline void RENAME(rgb16tobgr24)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb16tobgr24_c(const uint8_t *src, uint8_t *dst, + int src_size) { - const uint16_t *end; -#if 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; -#if 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" - "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" - :"=m"(*d) - :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"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&0x7E0)>>3; - *d++ = (bgr&0xF800)>>8; + uint8_t *d = (uint8_t *)dst; + const uint16_t *s = (const uint16_t *)src; + const uint16_t *end = s + src_size / 2; + + while (s < end) { + register uint16_t bgr = *s++; + *d++ = (bgr & 0x1F) << 3; + *d++ = (bgr & 0x7E0) >> 3; + *d++ = (bgr & 0xF800) >> 8; } } -static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb15to32_c(const uint8_t *src, uint8_t *dst, int src_size) { - const uint16_t *end; -#if HAVE_MMX - const uint16_t *mm_end; -#endif - uint8_t *d = dst; - const uint16_t *s = (const uint16_t *)src; - end = s + src_size/2; -#if HAVE_MMX - __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory"); - __asm__ volatile("pxor %%mm7,%%mm7 \n\t":::"memory"); - mm_end = end - 3; - 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 %%mm7, %%mm0 \n\t" - "punpcklwd %%mm7, %%mm1 \n\t" - "punpcklwd %%mm7, %%mm2 \n\t" - "punpckhwd %%mm7, %%mm3 \n\t" - "punpckhwd %%mm7, %%mm4 \n\t" - "punpckhwd %%mm7, %%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" - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm3, 8%0 \n\t" - :"=m"(*d) - :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r) - :"memory"); - d += 16; - s += 4; - } - __asm__ volatile(SFENCE:::"memory"); - __asm__ volatile(EMMS:::"memory"); -#endif - while (s < end) - { -#if 0 //slightly slower on Athlon - int bgr= *s++; - *((uint32_t*)d)++ = ((bgr&0x1F)<<3) + ((bgr&0x3E0)<<6) + ((bgr&0x7C00)<<9); -#else - register uint16_t bgr; - bgr = *s++; -#ifdef WORDS_BIGENDIAN + uint8_t *d = dst; + const uint16_t *s = (const uint16_t *)src; + const uint16_t *end = s + src_size / 2; + + while (s < end) { + register uint16_t bgr = *s++; +#if HAVE_BIGENDIAN *d++ = 255; - *d++ = (bgr&0x7C00)>>7; - *d++ = (bgr&0x3E0)>>2; - *d++ = (bgr&0x1F)<<3; + *d++ = (bgr & 0x7C00) >> 7; + *d++ = (bgr & 0x3E0) >> 2; + *d++ = (bgr & 0x1F) << 3; #else - *d++ = (bgr&0x1F)<<3; - *d++ = (bgr&0x3E0)>>2; - *d++ = (bgr&0x7C00)>>7; + *d++ = (bgr & 0x1F) << 3; + *d++ = (bgr & 0x3E0) >> 2; + *d++ = (bgr & 0x7C00) >> 7; *d++ = 255; -#endif - #endif } } -static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb16to32_c(const uint8_t *src, uint8_t *dst, int src_size) { - const uint16_t *end; -#if HAVE_MMX - const uint16_t *mm_end; -#endif - uint8_t *d = dst; - const uint16_t *s = (const uint16_t*)src; - end = s + src_size/2; -#if HAVE_MMX - __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory"); - __asm__ volatile("pxor %%mm7,%%mm7 \n\t":::"memory"); - mm_end = end - 3; - 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 %%mm7, %%mm0 \n\t" - "punpcklwd %%mm7, %%mm1 \n\t" - "punpcklwd %%mm7, %%mm2 \n\t" - "punpckhwd %%mm7, %%mm3 \n\t" - "punpckhwd %%mm7, %%mm4 \n\t" - "punpckhwd %%mm7, %%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" - MOVNTQ" %%mm0, %0 \n\t" - MOVNTQ" %%mm3, 8%0 \n\t" - :"=m"(*d) - :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r) - :"memory"); - d += 16; - s += 4; - } - __asm__ volatile(SFENCE:::"memory"); - __asm__ volatile(EMMS:::"memory"); -#endif - while (s < end) - { - register uint16_t bgr; - bgr = *s++; -#ifdef WORDS_BIGENDIAN + uint8_t *d = dst; + const uint16_t *s = (const uint16_t *)src; + const uint16_t *end = s + src_size / 2; + + while (s < end) { + register uint16_t bgr = *s++; +#if HAVE_BIGENDIAN *d++ = 255; - *d++ = (bgr&0xF800)>>8; - *d++ = (bgr&0x7E0)>>3; - *d++ = (bgr&0x1F)<<3; + *d++ = (bgr & 0xF800) >> 8; + *d++ = (bgr & 0x7E0) >> 3; + *d++ = (bgr & 0x1F) << 3; #else - *d++ = (bgr&0x1F)<<3; - *d++ = (bgr&0x7E0)>>3; - *d++ = (bgr&0xF800)>>8; + *d++ = (bgr & 0x1F) << 3; + *d++ = (bgr & 0x7E0) >> 3; + *d++ = (bgr & 0xF800) >> 8; *d++ = 255; #endif } } -static inline void RENAME(rgb32tobgr32)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void shuffle_bytes_2103_c(const uint8_t *src, uint8_t *dst, + int src_size) { - long idx = 15 - src_size; - const uint8_t *s = src-idx; - uint8_t *d = dst-idx; -#if HAVE_MMX - __asm__ volatile( - "test %0, %0 \n\t" - "jns 2f \n\t" - PREFETCH" (%1, %0) \n\t" - "movq %3, %%mm7 \n\t" - "pxor %4, %%mm7 \n\t" - "movq %%mm7, %%mm6 \n\t" - "pxor %5, %%mm7 \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 32(%1, %0) \n\t" - "movq (%1, %0), %%mm0 \n\t" - "movq 8(%1, %0), %%mm1 \n\t" -# if HAVE_MMX2 - "pshufw $177, %%mm0, %%mm3 \n\t" - "pshufw $177, %%mm1, %%mm5 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm6, %%mm3 \n\t" - "pand %%mm7, %%mm1 \n\t" - "pand %%mm6, %%mm5 \n\t" - "por %%mm3, %%mm0 \n\t" - "por %%mm5, %%mm1 \n\t" -# else - "movq %%mm0, %%mm2 \n\t" - "movq %%mm1, %%mm4 \n\t" - "pand %%mm7, %%mm0 \n\t" - "pand %%mm6, %%mm2 \n\t" - "pand %%mm7, %%mm1 \n\t" - "pand %%mm6, %%mm4 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq %%mm4, %%mm5 \n\t" - "pslld $16, %%mm2 \n\t" - "psrld $16, %%mm3 \n\t" - "pslld $16, %%mm4 \n\t" - "psrld $16, %%mm5 \n\t" - "por %%mm2, %%mm0 \n\t" - "por %%mm4, %%mm1 \n\t" - "por %%mm3, %%mm0 \n\t" - "por %%mm5, %%mm1 \n\t" -# endif - MOVNTQ" %%mm0, (%2, %0) \n\t" - MOVNTQ" %%mm1, 8(%2, %0) \n\t" - "add $16, %0 \n\t" - "js 1b \n\t" - SFENCE" \n\t" - EMMS" \n\t" - "2: \n\t" - : "+&r"(idx) - : "r" (s), "r" (d), "m" (mask32b), "m" (mask32r), "m" (mmx_one) - : "memory"); -#endif - for (; idx<15; idx+=4) { - register int v = *(const uint32_t *)&s[idx], g = v & 0xff00ff00; - v &= 0xff00ff; - *(uint32_t *)&d[idx] = (v>>16) + g + (v<<16); + int idx = 15 - src_size; + const uint8_t *s = src - idx; + uint8_t *d = dst - idx; + + for (; idx < 15; idx += 4) { + register int v = *(const uint32_t *)&s[idx], g = v & 0xff00ff00; + v &= 0xff00ff; + *(uint32_t *)&d[idx] = (v >> 16) + g + (v << 16); } } -static inline void RENAME(rgb24tobgr24)(const uint8_t *src, uint8_t *dst, long src_size) +static inline void rgb24tobgr24_c(const uint8_t *src, uint8_t *dst, int src_size) { unsigned i; -#if HAVE_MMX - long mmx_size= 23 - src_size; - __asm__ volatile ( - "test %%"REG_a", %%"REG_a" \n\t" - "jns 2f \n\t" - "movq "MANGLE(mask24r)", %%mm5 \n\t" - "movq "MANGLE(mask24g)", %%mm6 \n\t" - "movq "MANGLE(mask24b)", %%mm7 \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 32(%1, %%"REG_a") \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG - "movq (%1, %%"REG_a"), %%mm1 \n\t" // BGR BGR BG - "movq 2(%1, %%"REG_a"), %%mm2 \n\t" // R BGR BGR B - "psllq $16, %%mm0 \n\t" // 00 BGR BGR - "pand %%mm5, %%mm0 \n\t" - "pand %%mm6, %%mm1 \n\t" - "pand %%mm7, %%mm2 \n\t" - "por %%mm0, %%mm1 \n\t" - "por %%mm2, %%mm1 \n\t" - "movq 6(%1, %%"REG_a"), %%mm0 \n\t" // BGR BGR BG - MOVNTQ" %%mm1, (%2, %%"REG_a") \n\t" // RGB RGB RG - "movq 8(%1, %%"REG_a"), %%mm1 \n\t" // R BGR BGR B - "movq 10(%1, %%"REG_a"), %%mm2 \n\t" // GR BGR BGR - "pand %%mm7, %%mm0 \n\t" - "pand %%mm5, %%mm1 \n\t" - "pand %%mm6, %%mm2 \n\t" - "por %%mm0, %%mm1 \n\t" - "por %%mm2, %%mm1 \n\t" - "movq 14(%1, %%"REG_a"), %%mm0 \n\t" // R BGR BGR B - MOVNTQ" %%mm1, 8(%2, %%"REG_a") \n\t" // B RGB RGB R - "movq 16(%1, %%"REG_a"), %%mm1 \n\t" // GR BGR BGR - "movq 18(%1, %%"REG_a"), %%mm2 \n\t" // BGR BGR BG - "pand %%mm6, %%mm0 \n\t" - "pand %%mm7, %%mm1 \n\t" - "pand %%mm5, %%mm2 \n\t" - "por %%mm0, %%mm1 \n\t" - "por %%mm2, %%mm1 \n\t" - MOVNTQ" %%mm1, 16(%2, %%"REG_a") \n\t" - "add $24, %%"REG_a" \n\t" - " js 1b \n\t" - "2: \n\t" - : "+a" (mmx_size) - : "r" (src-mmx_size), "r"(dst-mmx_size) - ); - - __asm__ volatile(SFENCE:::"memory"); - __asm__ volatile(EMMS:::"memory"); - - if (mmx_size==23) return; //finished, was multiple of 8 - - src+= src_size; - dst+= src_size; - src_size= 23-mmx_size; - src-= src_size; - dst-= src_size; -#endif - for (i=0; i>1; - for (y=0; y> 1; -#if ARCH_ALPHA && HAVE_MVI -#define pl2yuy2(n) \ - y1 = yc[n]; \ - y2 = yc2[n]; \ - u = uc[n]; \ - v = vc[n]; \ - __asm__("unpkbw %1, %0" : "=r"(y1) : "r"(y1)); \ - __asm__("unpkbw %1, %0" : "=r"(y2) : "r"(y2)); \ - __asm__("unpkbl %1, %0" : "=r"(u) : "r"(u)); \ - __asm__("unpkbl %1, %0" : "=r"(v) : "r"(v)); \ - yuv1 = (u << 8) + (v << 24); \ - yuv2 = yuv1 + y2; \ - yuv1 += y1; \ - qdst[n] = yuv1; \ - qdst2[n] = yuv2; - - int i; - uint64_t *qdst = (uint64_t *) dst; - uint64_t *qdst2 = (uint64_t *) (dst + dstStride); - const uint32_t *yc = (uint32_t *) ysrc; - const uint32_t *yc2 = (uint32_t *) (ysrc + lumStride); - const uint16_t *uc = (uint16_t*) usrc, *vc = (uint16_t*) vsrc; - for (i = 0; i < chromWidth; i += 8){ - uint64_t y1, y2, yuv1, yuv2; - uint64_t u, v; - /* Prefetch */ - __asm__("ldq $31,64(%0)" :: "r"(yc)); - __asm__("ldq $31,64(%0)" :: "r"(yc2)); - __asm__("ldq $31,64(%0)" :: "r"(uc)); - __asm__("ldq $31,64(%0)" :: "r"(vc)); - - pl2yuy2(0); - pl2yuy2(1); - pl2yuy2(2); - pl2yuy2(3); - - yc += 4; - yc2 += 4; - uc += 4; - vc += 4; - qdst += 4; - qdst2 += 4; - } - y++; - ysrc += lumStride; - dst += dstStride; - -#elif HAVE_FAST_64BIT - int i; - uint64_t *ldst = (uint64_t *) dst; + for (y = 0; y < height; y++) { +#if HAVE_FAST_64BIT + uint64_t *ldst = (uint64_t *)dst; const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc; - for (i = 0; i < chromWidth; i += 2){ - uint64_t k, l; - k = yc[0] + (uc[0] << 8) + - (yc[1] << 16) + (vc[0] << 24); - l = yc[2] + (uc[1] << 8) + - (yc[3] << 16) + (vc[1] << 24); + for (i = 0; i < chromWidth; i += 2) { + uint64_t k = yc[0] + (uc[0] << 8) + + (yc[1] << 16) + (vc[0] << 24); + uint64_t l = yc[2] + (uc[1] << 8) + + (yc[3] << 16) + (vc[1] << 24); *ldst++ = k + (l << 32); - yc += 4; - uc += 2; - vc += 2; + yc += 4; + uc += 2; + vc += 2; } #else - int i, *idst = (int32_t *) dst; + int *idst = (int32_t *)dst; const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc; - for (i = 0; i < chromWidth; i++){ -#ifdef WORDS_BIGENDIAN - *idst++ = (yc[0] << 24)+ (uc[0] << 16) + - (yc[1] << 8) + (vc[0] << 0); + + for (i = 0; i < chromWidth; i++) { +#if HAVE_BIGENDIAN + *idst++ = (yc[0] << 24) + (uc[0] << 16) + + (yc[1] << 8) + (vc[0] << 0); #else *idst++ = yc[0] + (uc[0] << 8) + - (yc[1] << 16) + (vc[0] << 24); + (yc[1] << 16) + (vc[0] << 24); #endif yc += 2; uc++; vc++; } #endif -#endif - if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) - { + if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) { usrc += chromStride; vsrc += chromStride; } ysrc += lumStride; dst += dstStride; } -#if HAVE_MMX -__asm__( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); -#endif } /** * 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 I will fix it.) */ -static inline void RENAME(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) +static inline void yv12toyuy2_c(const uint8_t *ysrc, const uint8_t *usrc, + const uint8_t *vsrc, uint8_t *dst, + int width, int height, int lumStride, + int chromStride, int dstStride) { //FIXME interpolate chroma - RENAME(yuvPlanartoyuy2)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 2); + yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride, + chromStride, dstStride, 2); } -static inline void RENAME(yuvPlanartouyvy)(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, long vertLumPerChroma) +static inline void yuvPlanartouyvy_c(const uint8_t *ysrc, const uint8_t *usrc, + const uint8_t *vsrc, uint8_t *dst, + int width, int height, + int lumStride, int chromStride, + int dstStride, int vertLumPerChroma) { - long y; - const long chromWidth= width>>1; - for (y=0; yyuy2 + int y, i; + const int chromWidth = width >> 1; + for (y = 0; y < height; y++) { #if HAVE_FAST_64BIT - int i; - uint64_t *ldst = (uint64_t *) dst; + uint64_t *ldst = (uint64_t *)dst; const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc; - for (i = 0; i < chromWidth; i += 2){ - uint64_t k, l; - k = uc[0] + (yc[0] << 8) + - (vc[0] << 16) + (yc[1] << 24); - l = uc[1] + (yc[2] << 8) + - (vc[1] << 16) + (yc[3] << 24); + for (i = 0; i < chromWidth; i += 2) { + uint64_t k = uc[0] + (yc[0] << 8) + + (vc[0] << 16) + (yc[1] << 24); + uint64_t l = uc[1] + (yc[2] << 8) + + (vc[1] << 16) + (yc[3] << 24); *ldst++ = k + (l << 32); - yc += 4; - uc += 2; - vc += 2; + yc += 4; + uc += 2; + vc += 2; } #else - int i, *idst = (int32_t *) dst; + int *idst = (int32_t *)dst; const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc; - for (i = 0; i < chromWidth; i++){ -#ifdef WORDS_BIGENDIAN - *idst++ = (uc[0] << 24)+ (yc[0] << 16) + - (vc[0] << 8) + (yc[1] << 0); + + for (i = 0; i < chromWidth; i++) { +#if HAVE_BIGENDIAN + *idst++ = (uc[0] << 24) + (yc[0] << 16) + + (vc[0] << 8) + (yc[1] << 0); #else *idst++ = uc[0] + (yc[0] << 8) + - (vc[0] << 16) + (yc[1] << 24); + (vc[0] << 16) + (yc[1] << 24); #endif yc += 2; uc++; vc++; } #endif -#endif - if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) - { + if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) { usrc += chromStride; vsrc += chromStride; } ysrc += lumStride; - dst += dstStride; + dst += dstStride; } -#if HAVE_MMX -__asm__( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); -#endif } /** * 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 I will fix it.) */ -static inline void RENAME(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) +static inline void yv12touyvy_c(const uint8_t *ysrc, const uint8_t *usrc, + const uint8_t *vsrc, uint8_t *dst, + int width, int height, int lumStride, + int chromStride, int dstStride) { //FIXME interpolate chroma - RENAME(yuvPlanartouyvy)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 2); + yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride, + chromStride, dstStride, 2); } /** * Width should be a multiple of 16. */ -static inline void RENAME(yuv422ptouyvy)(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) +static inline void yuv422ptouyvy_c(const uint8_t *ysrc, const uint8_t *usrc, + const uint8_t *vsrc, uint8_t *dst, + int width, int height, int lumStride, + int chromStride, int dstStride) { - RENAME(yuvPlanartouyvy)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 1); + yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride, + chromStride, dstStride, 1); } /** * Width should be a multiple of 16. */ -static inline void RENAME(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) +static inline void yuv422ptoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc, + const uint8_t *vsrc, uint8_t *dst, + int width, int height, int lumStride, + int chromStride, int dstStride) { - RENAME(yuvPlanartoyuy2)(ysrc, usrc, vsrc, dst, width, height, lumStride, chromStride, dstStride, 1); + yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride, + chromStride, dstStride, 1); } /** * 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 I will fix it.) */ -static inline void RENAME(yuy2toyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, - long width, long height, - long lumStride, long chromStride, long srcStride) +static inline void yuy2toyv12_c(const uint8_t *src, uint8_t *ydst, + uint8_t *udst, uint8_t *vdst, + int width, int height, int lumStride, + int chromStride, int srcStride) { - long y; - const long chromWidth= width>>1; - for (y=0; y> 1; - __asm__ volatile( - "xor %%"REG_a", %%"REG_a" \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_a", 4) \n\t" - "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) - "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) - "movq 16(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(8) - "movq 24(%0, %%"REG_a", 4), %%mm3 \n\t" // YUYV YUYV(12) - "pand %%mm7, %%mm0 \n\t" // Y0Y0 Y0Y0(0) - "pand %%mm7, %%mm1 \n\t" // Y0Y0 Y0Y0(4) - "pand %%mm7, %%mm2 \n\t" // Y0Y0 Y0Y0(8) - "pand %%mm7, %%mm3 \n\t" // Y0Y0 Y0Y0(12) - "packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0) - "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8) - - MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \n\t" - - "add $8, %%"REG_a" \n\t" - "cmp %4, %%"REG_a" \n\t" - " jb 1b \n\t" - - ::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth) - : "memory", "%"REG_a - ); -#else - long i; - for (i=0; i>2; - dst[2*x+2]= ( src[x] + 3*src[x+1])>>2; + for (x = 0; x < srcWidth - 1; x++) { + dst[2 * x + 1] = (3 * src[x] + src[x + 1]) >> 2; + dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2; } - dst[2*srcWidth-1]= src[srcWidth-1]; - - dst+= dstStride; - - for (y=1; y>2; - dst[dstStride]= ( src[0] + 3*src[srcStride])>>2; - - for (x=mmxSize-1; x>2; - dst[2*x+dstStride+2]= ( src[x+0] + 3*src[x+srcStride+1])>>2; - dst[2*x+dstStride+1]= ( src[x+1] + 3*src[x+srcStride ])>>2; - dst[2*x +2]= (3*src[x+1] + src[x+srcStride ])>>2; + dst[2 * srcWidth - 1] = src[srcWidth - 1]; + + dst += dstStride; + + for (y = 1; y < srcHeight; y++) { + const int mmxSize = 1; + + dst[0] = (src[0] * 3 + src[srcStride]) >> 2; + dst[dstStride] = (src[0] + 3 * src[srcStride]) >> 2; + + for (x = mmxSize - 1; x < srcWidth - 1; x++) { + dst[2 * x + 1] = (src[x + 0] * 3 + src[x + srcStride + 1]) >> 2; + dst[2 * x + dstStride + 2] = (src[x + 0] + 3 * src[x + srcStride + 1]) >> 2; + dst[2 * x + dstStride + 1] = (src[x + 1] + 3 * src[x + srcStride]) >> 2; + dst[2 * x + 2] = (src[x + 1] * 3 + src[x + srcStride]) >> 2; } - dst[srcWidth*2 -1 ]= (3*src[srcWidth-1] + src[srcWidth-1 + srcStride])>>2; - dst[srcWidth*2 -1 + dstStride]= ( src[srcWidth-1] + 3*src[srcWidth-1 + srcStride])>>2; + dst[srcWidth * 2 - 1] = (src[srcWidth - 1] * 3 + src[srcWidth - 1 + srcStride]) >> 2; + dst[srcWidth * 2 - 1 + dstStride] = (src[srcWidth - 1] + 3 * src[srcWidth - 1 + srcStride]) >> 2; - dst+=dstStride*2; - src+=srcStride; + dst += dstStride * 2; + src += srcStride; } // last line -#if 1 - dst[0]= src[0]; + dst[0] = src[0]; - for (x=0; x>2; - dst[2*x+2]= ( src[x] + 3*src[x+1])>>2; - } - dst[2*srcWidth-1]= src[srcWidth-1]; -#else - for (x=0; x> 2; + dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2; } -#endif - -#if HAVE_MMX -__asm__ volatile( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); -#endif + dst[2 * srcWidth - 1] = src[srcWidth - 1]; } /** @@ -2006,124 +596,34 @@ __asm__ volatile( EMMS" \n\t" * Chrominance data is only taken from every second line, others are ignored. * FIXME: Write HQ version. */ -static inline void RENAME(uyvytoyv12)(const uint8_t *src, uint8_t *ydst, uint8_t *udst, uint8_t *vdst, - long width, long height, - long lumStride, long chromStride, long srcStride) +static inline void uyvytoyv12_c(const uint8_t *src, uint8_t *ydst, + uint8_t *udst, uint8_t *vdst, + int width, int height, int lumStride, + int chromStride, int srcStride) { - long y; - const long chromWidth= width>>1; - for (y=0; y> 1; - __asm__ volatile( - "xor %%"REG_a", %%"REG_a" \n\t" - ASMALIGN(4) - "1: \n\t" - PREFETCH" 64(%0, %%"REG_a", 4) \n\t" - "movq (%0, %%"REG_a", 4), %%mm0 \n\t" // YUYV YUYV(0) - "movq 8(%0, %%"REG_a", 4), %%mm1 \n\t" // YUYV YUYV(4) - "movq 16(%0, %%"REG_a", 4), %%mm2 \n\t" // YUYV YUYV(8) - "movq 24(%0, %%"REG_a", 4), %%mm3 \n\t" // YUYV YUYV(12) - "psrlw $8, %%mm0 \n\t" // Y0Y0 Y0Y0(0) - "psrlw $8, %%mm1 \n\t" // Y0Y0 Y0Y0(4) - "psrlw $8, %%mm2 \n\t" // Y0Y0 Y0Y0(8) - "psrlw $8, %%mm3 \n\t" // Y0Y0 Y0Y0(12) - "packuswb %%mm1, %%mm0 \n\t" // YYYY YYYY(0) - "packuswb %%mm3, %%mm2 \n\t" // YYYY YYYY(8) - - MOVNTQ" %%mm0, (%1, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm2, 8(%1, %%"REG_a", 2) \n\t" - - "add $8, %%"REG_a" \n\t" - "cmp %4, %%"REG_a" \n\t" - " jb 1b \n\t" - - ::"r"(src), "r"(ydst), "r"(udst), "r"(vdst), "g" (chromWidth) - : "memory", "%"REG_a - ); -#else - long i; - for (i=0; i>1; -#if HAVE_MMX - for (y=0; y> 1; - udst += chromStride; - vdst += chromStride; - src += srcStride*2; - } + for (y = 0; y < height; y += 2) { + int i; + for (i = 0; i < chromWidth; i++) { + unsigned int b = src[6 * i + 0]; + unsigned int g = src[6 * i + 1]; + unsigned int r = src[6 * i + 2]; - __asm__ volatile( EMMS" \n\t" - SFENCE" \n\t" - :::"memory"); -#else - y=0; -#endif - for (; y>RGB2YUV_SHIFT) + 16; - unsigned int V = ((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128; - unsigned int U = ((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128; + unsigned int Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16; + unsigned int V = ((RV * r + GV * g + BV * b) >> RGB2YUV_SHIFT) + 128; + unsigned int U = ((RU * r + GU * g + BU * b) >> RGB2YUV_SHIFT) + 128; udst[i] = U; vdst[i] = V; - ydst[2*i] = Y; + ydst[2 * i] = Y; - b = src[6*i+3]; - g = src[6*i+4]; - r = src[6*i+5]; + b = src[6 * i + 3]; + g = src[6 * i + 4]; + r = src[6 * i + 5]; - Y = ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16; - ydst[2*i+1] = Y; + Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16; + ydst[2 * i + 1] = Y; } ydst += lumStride; src += srcStride; - for (i=0; i>RGB2YUV_SHIFT) + 16; + unsigned int Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16; - ydst[2*i] = Y; + ydst[2 * i] = Y; - b = src[6*i+3]; - g = src[6*i+4]; - r = src[6*i+5]; + b = src[6 * i + 3]; + g = src[6 * i + 4]; + r = src[6 * i + 5]; - Y = ((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16; - ydst[2*i+1] = Y; + Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16; + ydst[2 * i + 1] = Y; } udst += chromStride; vdst += chromStride; @@ -2429,310 +688,258 @@ static inline void RENAME(rgb24toyv12)(const uint8_t *src, uint8_t *ydst, uint8_ } } -static void RENAME(interleaveBytes)(uint8_t *src1, uint8_t *src2, uint8_t *dest, - long width, long height, long src1Stride, - long src2Stride, long dstStride){ - long h; - - for (h=0; h < height; h++) - { - long w; - -#if HAVE_MMX -#if HAVE_SSE2 - __asm__( - "xor %%"REG_a", %%"REG_a" \n\t" - "1: \n\t" - PREFETCH" 64(%1, %%"REG_a") \n\t" - PREFETCH" 64(%2, %%"REG_a") \n\t" - "movdqa (%1, %%"REG_a"), %%xmm0 \n\t" - "movdqa (%1, %%"REG_a"), %%xmm1 \n\t" - "movdqa (%2, %%"REG_a"), %%xmm2 \n\t" - "punpcklbw %%xmm2, %%xmm0 \n\t" - "punpckhbw %%xmm2, %%xmm1 \n\t" - "movntdq %%xmm0, (%0, %%"REG_a", 2) \n\t" - "movntdq %%xmm1, 16(%0, %%"REG_a", 2) \n\t" - "add $16, %%"REG_a" \n\t" - "cmp %3, %%"REG_a" \n\t" - " jb 1b \n\t" - ::"r"(dest), "r"(src1), "r"(src2), "r" (width-15) - : "memory", "%"REG_a"" - ); -#else - __asm__( - "xor %%"REG_a", %%"REG_a" \n\t" - "1: \n\t" - PREFETCH" 64(%1, %%"REG_a") \n\t" - PREFETCH" 64(%2, %%"REG_a") \n\t" - "movq (%1, %%"REG_a"), %%mm0 \n\t" - "movq 8(%1, %%"REG_a"), %%mm2 \n\t" - "movq %%mm0, %%mm1 \n\t" - "movq %%mm2, %%mm3 \n\t" - "movq (%2, %%"REG_a"), %%mm4 \n\t" - "movq 8(%2, %%"REG_a"), %%mm5 \n\t" - "punpcklbw %%mm4, %%mm0 \n\t" - "punpckhbw %%mm4, %%mm1 \n\t" - "punpcklbw %%mm5, %%mm2 \n\t" - "punpckhbw %%mm5, %%mm3 \n\t" - MOVNTQ" %%mm0, (%0, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm1, 8(%0, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm2, 16(%0, %%"REG_a", 2) \n\t" - MOVNTQ" %%mm3, 24(%0, %%"REG_a", 2) \n\t" - "add $16, %%"REG_a" \n\t" - "cmp %3, %%"REG_a" \n\t" - " jb 1b \n\t" - ::"r"(dest), "r"(src1), "r"(src2), "r" (width-15) - : "memory", "%"REG_a - ); -#endif - for (w= (width&(~15)); w < width; w++) - { - dest[2*w+0] = src1[w]; - dest[2*w+1] = src2[w]; - } -#else - for (w=0; w < width; w++) - { - dest[2*w+0] = src1[w]; - dest[2*w+1] = src2[w]; +static void interleaveBytes_c(const uint8_t *src1, const uint8_t *src2, + uint8_t *dest, int width, int height, + int src1Stride, int src2Stride, int dstStride) +{ + int h; + + for (h = 0; h < height; h++) { + int w; + for (w = 0; w < width; w++) { + dest[2 * w + 0] = src1[w]; + dest[2 * w + 1] = src2[w]; } -#endif dest += dstStride; - src1 += src1Stride; - src2 += src2Stride; + src1 += src1Stride; + src2 += src2Stride; } -#if HAVE_MMX - __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" - ); -#endif } -static inline void RENAME(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) +static inline void vu9_to_vu12_c(const uint8_t *src1, const uint8_t *src2, + uint8_t *dst1, uint8_t *dst2, + int width, int height, + int srcStride1, int srcStride2, + int dstStride1, int dstStride2) { - long y,x,w,h; - w=width/2; h=height/2; -#if HAVE_MMX - __asm__ volatile( - PREFETCH" %0 \n\t" - PREFETCH" %1 \n\t" - ::"m"(*(src1+srcStride1)),"m"(*(src2+srcStride2)):"memory"); -#endif - for (y=0;y>1); - uint8_t* d=dst1+dstStride1*y; - x=0; -#if HAVE_MMX - for (;x> 1); + uint8_t *d = dst1 + dstStride1 * y; + for (x = 0; x < w; x++) + d[2 * x] = d[2 * x + 1] = s1[x]; } -#endif - for (;x> 1); + uint8_t *d = dst2 + dstStride2 * y; + for (x = 0; x < w; x++) + d[2 * x] = d[2 * x + 1] = s2[x]; } - for (y=0;y>1); - uint8_t* d=dst2+dstStride2*y; - x=0; -#if HAVE_MMX - for (;x> 2); + const uint8_t *vp = src3 + srcStride3 * (y >> 2); + uint8_t *d = dst + dstStride * y; + for (x = 0; x < w; x++) { + const int x2 = x << 2; + d[8 * x + 0] = yp[x2]; + d[8 * x + 1] = up[x]; + d[8 * x + 2] = yp[x2 + 1]; + d[8 * x + 3] = vp[x]; + d[8 * x + 4] = yp[x2 + 2]; + d[8 * x + 5] = up[x]; + d[8 * x + 6] = yp[x2 + 3]; + d[8 * x + 7] = vp[x]; + } } -#endif - for (;x>2); - const uint8_t* vp=src3+srcStride3*(y>>2); - uint8_t* d=dst+dstStride*y; - x=0; -#if HAVE_MMX - for (;x> 1; + dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1; + count++; } +} + +static void extract_odd2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1, + int count) +{ + dst0 += count; + dst1 += count; + src += count * 4; + count = -count; + src++; + while (count < 0) { + dst0[count] = src[4 * count + 0]; + dst1[count] = src[4 * count + 2]; + count++; + } +} + +static void extract_odd2avg_c(const uint8_t *src0, const uint8_t *src1, + uint8_t *dst0, uint8_t *dst1, int count) +{ + dst0 += count; + dst1 += count; + src0 += count * 4; + src1 += count * 4; + count = -count; + src0++; + src1++; + while (count < 0) { + dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1; + dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1; + count++; + } +} + +static void yuyvtoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + const uint8_t *src, int width, int height, + int lumStride, int chromStride, int srcStride) +{ + int y; + const int chromWidth = -((-width) >> 1); + + for (y = 0; y < height; y++) { + extract_even_c(src, ydst, width); + if (y & 1) { + extract_odd2avg_c(src - srcStride, src, udst, vdst, chromWidth); + udst += chromStride; + vdst += chromStride; + } + + src += srcStride; + ydst += lumStride; } -#if HAVE_MMX - __asm__( - EMMS" \n\t" - SFENCE" \n\t" - ::: "memory" - ); -#endif } -static inline void RENAME(rgb2rgb_init)(void){ - rgb15to16 = RENAME(rgb15to16); - rgb15tobgr24 = RENAME(rgb15tobgr24); - rgb15to32 = RENAME(rgb15to32); - rgb16tobgr24 = RENAME(rgb16tobgr24); - rgb16to32 = RENAME(rgb16to32); - rgb16to15 = RENAME(rgb16to15); - rgb24tobgr16 = RENAME(rgb24tobgr16); - rgb24tobgr15 = RENAME(rgb24tobgr15); - rgb24tobgr32 = RENAME(rgb24tobgr32); - rgb32to16 = RENAME(rgb32to16); - rgb32to15 = RENAME(rgb32to15); - rgb32tobgr24 = RENAME(rgb32tobgr24); - rgb24to15 = RENAME(rgb24to15); - rgb24to16 = RENAME(rgb24to16); - rgb24tobgr24 = RENAME(rgb24tobgr24); - rgb32tobgr32 = RENAME(rgb32tobgr32); - rgb32tobgr16 = RENAME(rgb32tobgr16); - rgb32tobgr15 = RENAME(rgb32tobgr15); - yv12toyuy2 = RENAME(yv12toyuy2); - yv12touyvy = RENAME(yv12touyvy); - yuv422ptoyuy2 = RENAME(yuv422ptoyuy2); - yuv422ptouyvy = RENAME(yuv422ptouyvy); - yuy2toyv12 = RENAME(yuy2toyv12); -// uyvytoyv12 = RENAME(uyvytoyv12); -// yvu9toyv12 = RENAME(yvu9toyv12); - planar2x = RENAME(planar2x); - rgb24toyv12 = RENAME(rgb24toyv12); - interleaveBytes = RENAME(interleaveBytes); - vu9_to_vu12 = RENAME(vu9_to_vu12); - yvu9_to_yuy2 = RENAME(yvu9_to_yuy2); +static void yuyvtoyuv422_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + const uint8_t *src, int width, int height, + int lumStride, int chromStride, int srcStride) +{ + int y; + const int chromWidth = -((-width) >> 1); + + for (y = 0; y < height; y++) { + extract_even_c(src, ydst, width); + extract_odd2_c(src, udst, vdst, chromWidth); + + src += srcStride; + ydst += lumStride; + udst += chromStride; + vdst += chromStride; + } +} + +static void uyvytoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + const uint8_t *src, int width, int height, + int lumStride, int chromStride, int srcStride) +{ + int y; + const int chromWidth = -((-width) >> 1); + + for (y = 0; y < height; y++) { + extract_even_c(src + 1, ydst, width); + if (y & 1) { + extract_even2avg_c(src - srcStride, src, udst, vdst, chromWidth); + udst += chromStride; + vdst += chromStride; + } + + src += srcStride; + ydst += lumStride; + } +} + +static void uyvytoyuv422_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, + const uint8_t *src, int width, int height, + int lumStride, int chromStride, int srcStride) +{ + int y; + const int chromWidth = -((-width) >> 1); + + for (y = 0; y < height; y++) { + extract_even_c(src + 1, ydst, width); + extract_even2_c(src, udst, vdst, chromWidth); + + src += srcStride; + ydst += lumStride; + udst += chromStride; + vdst += chromStride; + } +} + +static inline void rgb2rgb_init_c(void) +{ + rgb15to16 = rgb15to16_c; + rgb15tobgr24 = rgb15tobgr24_c; + rgb15to32 = rgb15to32_c; + rgb16tobgr24 = rgb16tobgr24_c; + rgb16to32 = rgb16to32_c; + rgb16to15 = rgb16to15_c; + rgb24tobgr16 = rgb24tobgr16_c; + rgb24tobgr15 = rgb24tobgr15_c; + rgb24tobgr32 = rgb24tobgr32_c; + rgb32to16 = rgb32to16_c; + rgb32to15 = rgb32to15_c; + rgb32tobgr24 = rgb32tobgr24_c; + rgb24to15 = rgb24to15_c; + rgb24to16 = rgb24to16_c; + rgb24tobgr24 = rgb24tobgr24_c; + shuffle_bytes_2103 = shuffle_bytes_2103_c; + rgb32tobgr16 = rgb32tobgr16_c; + rgb32tobgr15 = rgb32tobgr15_c; + yv12toyuy2 = yv12toyuy2_c; + yv12touyvy = yv12touyvy_c; + yuv422ptoyuy2 = yuv422ptoyuy2_c; + yuv422ptouyvy = yuv422ptouyvy_c; + yuy2toyv12 = yuy2toyv12_c; + planar2x = planar2x_c; + rgb24toyv12 = rgb24toyv12_c; + interleaveBytes = interleaveBytes_c; + vu9_to_vu12 = vu9_to_vu12_c; + yvu9_to_yuy2 = yvu9_to_yuy2_c; + + uyvytoyuv420 = uyvytoyuv420_c; + uyvytoyuv422 = uyvytoyuv422_c; + yuyvtoyuv420 = yuyvtoyuv420_c; + yuyvtoyuv422 = yuyvtoyuv422_c; }