/lib/ffmpeg/libswscale/x86/rgb2rgb_template.c
C | 2595 lines | 2386 code | 131 blank | 78 comment | 78 complexity | 69133a8698389e86b7d9121a15e4a22d MD5 | raw file
Possible License(s): GPL-3.0, CC-BY-SA-3.0, BSD-3-Clause, GPL-2.0, LGPL-3.0, 0BSD, LGPL-2.0, AGPL-1.0, LGPL-2.1
Large files files are truncated, but you can click here to view the full file
- /*
- * software RGB to RGB converter
- * pluralize by software PAL8 to RGB converter
- * software YUV to YUV converter
- * software YUV to RGB converter
- * Written by Nick Kurshev.
- * 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.
- *
- * FFmpeg is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * FFmpeg is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with FFmpeg; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
- #include <stddef.h>
- #undef PREFETCH
- #undef MOVNTQ
- #undef EMMS
- #undef SFENCE
- #undef PAVGB
- #if COMPILE_TEMPLATE_AMD3DNOW
- #define PREFETCH "prefetch"
- #define PAVGB "pavgusb"
- #elif COMPILE_TEMPLATE_MMX2
- #define PREFETCH "prefetchnta"
- #define PAVGB "pavgb"
- #else
- #define PREFETCH " # nop"
- #endif
- #if COMPILE_TEMPLATE_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 COMPILE_TEMPLATE_MMX2
- #define MOVNTQ "movntq"
- #define SFENCE "sfence"
- #else
- #define MOVNTQ "movq"
- #define SFENCE " # nop"
- #endif
- #if !COMPILE_TEMPLATE_SSE2
- #if !COMPILE_TEMPLATE_AMD3DNOW
- static inline void RENAME(rgb24tobgr32)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- uint8_t *dest = dst;
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- end = s + src_size;
- __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");
- while (s < end) {
- *dest++ = *s++;
- *dest++ = *s++;
- *dest++ = *s++;
- *dest++ = 255;
- }
- }
- #define STORE_BGR24_MMX \
- "psrlq $8, %%mm2 \n\t" \
- "psrlq $8, %%mm3 \n\t" \
- "psrlq $8, %%mm6 \n\t" \
- "psrlq $8, %%mm7 \n\t" \
- "pand "MANGLE(mask24l)", %%mm0\n\t" \
- "pand "MANGLE(mask24l)", %%mm1\n\t" \
- "pand "MANGLE(mask24l)", %%mm4\n\t" \
- "pand "MANGLE(mask24l)", %%mm5\n\t" \
- "pand "MANGLE(mask24h)", %%mm2\n\t" \
- "pand "MANGLE(mask24h)", %%mm3\n\t" \
- "pand "MANGLE(mask24h)", %%mm6\n\t" \
- "pand "MANGLE(mask24h)", %%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 "MANGLE(mask24hh)", %%mm2\n\t" \
- "pand "MANGLE(mask24hhh)", %%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 "MANGLE(mask24hhhh)", %%mm5\n\t" \
- "por %%mm5, %%mm4 \n\t" \
- \
- MOVNTQ" %%mm0, %0 \n\t" \
- MOVNTQ" %%mm1, 8%0 \n\t" \
- MOVNTQ" %%mm4, 16%0"
- static inline void RENAME(rgb32tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- uint8_t *dest = dst;
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- end = s + src_size;
- __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"
- STORE_BGR24_MMX
- :"=m"(*dest)
- :"m"(*s)
- :"memory");
- dest += 24;
- s += 32;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- while (s < end) {
- *dest++ = *s++;
- *dest++ = *s++;
- *dest++ = *s++;
- s++;
- }
- }
- /*
- 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, 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;
- __asm__ volatile(PREFETCH" %0"::"m"(*s));
- __asm__ volatile("movq %0, %%mm4"::"m"(mask15s));
- mm_end = end - 15;
- while (s<mm_end) {
- __asm__ volatile(
- PREFETCH" 32%1 \n\t"
- "movq %1, %%mm0 \n\t"
- "movq 8%1, %%mm2 \n\t"
- "movq %%mm0, %%mm1 \n\t"
- "movq %%mm2, %%mm3 \n\t"
- "pand %%mm4, %%mm0 \n\t"
- "pand %%mm4, %%mm2 \n\t"
- "paddw %%mm1, %%mm0 \n\t"
- "paddw %%mm3, %%mm2 \n\t"
- MOVNTQ" %%mm0, %0 \n\t"
- MOVNTQ" %%mm2, 8%0"
- :"=m"(*d)
- :"m"(*s)
- );
- d+=16;
- s+=16;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- mm_end = end - 3;
- while (s < mm_end) {
- register unsigned x= *((const uint32_t *)s);
- *((uint32_t *)d) = (x&0x7FFF7FFF) + (x&0x7FE07FE0);
- d+=4;
- s+=4;
- }
- if (s < end) {
- register unsigned short x= *((const uint16_t *)s);
- *((uint16_t *)d) = (x&0x7FFF) + (x&0x7FE0);
- }
- }
- static inline void RENAME(rgb16to15)(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;
- __asm__ volatile(PREFETCH" %0"::"m"(*s));
- __asm__ volatile("movq %0, %%mm7"::"m"(mask15rg));
- __asm__ volatile("movq %0, %%mm6"::"m"(mask15b));
- mm_end = end - 15;
- while (s<mm_end) {
- __asm__ volatile(
- PREFETCH" 32%1 \n\t"
- "movq %1, %%mm0 \n\t"
- "movq 8%1, %%mm2 \n\t"
- "movq %%mm0, %%mm1 \n\t"
- "movq %%mm2, %%mm3 \n\t"
- "psrlq $1, %%mm0 \n\t"
- "psrlq $1, %%mm2 \n\t"
- "pand %%mm7, %%mm0 \n\t"
- "pand %%mm7, %%mm2 \n\t"
- "pand %%mm6, %%mm1 \n\t"
- "pand %%mm6, %%mm3 \n\t"
- "por %%mm1, %%mm0 \n\t"
- "por %%mm3, %%mm2 \n\t"
- MOVNTQ" %%mm0, %0 \n\t"
- MOVNTQ" %%mm2, 8%0"
- :"=m"(*d)
- :"m"(*s)
- );
- d+=16;
- s+=16;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- 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);
- }
- }
- static inline void RENAME(rgb32to16)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- 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"
- ".p2align 4 \n\t"
- "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");
- 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, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- __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");
- 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, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- 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"
- ".p2align 4 \n\t"
- "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");
- 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, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- __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");
- 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, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- __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");
- 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(rgb24to16)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- __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");
- 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(rgb24tobgr15)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- __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");
- 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(rgb24to15)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint8_t *s = src;
- const uint8_t *end;
- const uint8_t *mm_end;
- uint16_t *d = (uint16_t *)dst;
- end = s + src_size;
- __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");
- while (s < end) {
- const int r = *s++;
- const int g = *s++;
- const int b = *s++;
- *d++ = (b>>3) | ((g&0xF8)<<2) | ((r&0xF8)<<7);
- }
- }
- static inline void RENAME(rgb15tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint16_t *end;
- const uint16_t *mm_end;
- uint8_t *d = dst;
- const uint16_t *s = (const uint16_t*)src;
- end = s + src_size/2;
- __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 $5, %%mm0 \n\t"
- "pmulhw %6, %%mm0 \n\t"
- "pmulhw %6, %%mm1 \n\t"
- "pmulhw %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 $5, %%mm0 \n\t"
- "pmulhw %6, %%mm0 \n\t"
- "pmulhw %6, %%mm1 \n\t"
- "pmulhw %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),"m"(mul15_mid),"m"(mul15_hi)
- :"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"
- STORE_BGR24_MMX
- :"=m"(*d)
- :"m"(*s)
- :"memory");
- d += 24;
- s += 8;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- while (s < end) {
- register uint16_t bgr;
- bgr = *s++;
- *d++ = ((bgr&0x1F)<<3) | ((bgr&0x1F)>>2);
- *d++ = ((bgr&0x3E0)>>2) | ((bgr&0x3E0)>>7);
- *d++ = ((bgr&0x7C00)>>7) | ((bgr&0x7C00)>>12);
- }
- }
- static inline void RENAME(rgb16tobgr24)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint16_t *end;
- const uint16_t *mm_end;
- uint8_t *d = (uint8_t *)dst;
- const uint16_t *s = (const uint16_t *)src;
- end = s + src_size/2;
- __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 $5, %%mm0 \n\t"
- "psrlq $1, %%mm2 \n\t"
- "pmulhw %6, %%mm0 \n\t"
- "pmulhw %8, %%mm1 \n\t"
- "pmulhw %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 $5, %%mm0 \n\t"
- "psrlq $1, %%mm2 \n\t"
- "pmulhw %6, %%mm0 \n\t"
- "pmulhw %8, %%mm1 \n\t"
- "pmulhw %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"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mmx_null),"m"(mul15_mid),"m"(mul15_hi),"m"(mul16_mid)
- :"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"
- STORE_BGR24_MMX
- :"=m"(*d)
- :"m"(*s)
- :"memory");
- d += 24;
- s += 8;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- while (s < end) {
- register uint16_t bgr;
- bgr = *s++;
- *d++ = ((bgr&0x1F)<<3) | ((bgr&0x1F)>>2);
- *d++ = ((bgr&0x7E0)>>3) | ((bgr&0x7E0)>>9);
- *d++ = ((bgr&0xF800)>>8) | ((bgr&0xF800)>>13);
- }
- }
- /*
- * mm0 = 00 B3 00 B2 00 B1 00 B0
- * mm1 = 00 G3 00 G2 00 G1 00 G0
- * mm2 = 00 R3 00 R2 00 R1 00 R0
- * mm6 = FF FF FF FF FF FF FF FF
- * mm7 = 00 00 00 00 00 00 00 00
- */
- #define PACK_RGB32 \
- "packuswb %%mm7, %%mm0 \n\t" /* 00 00 00 00 B3 B2 B1 B0 */ \
- "packuswb %%mm7, %%mm1 \n\t" /* 00 00 00 00 G3 G2 G1 G0 */ \
- "packuswb %%mm7, %%mm2 \n\t" /* 00 00 00 00 R3 R2 R1 R0 */ \
- "punpcklbw %%mm1, %%mm0 \n\t" /* G3 B3 G2 B2 G1 B1 G0 B0 */ \
- "punpcklbw %%mm6, %%mm2 \n\t" /* FF R3 FF R2 FF R1 FF R0 */ \
- "movq %%mm0, %%mm3 \n\t" \
- "punpcklwd %%mm2, %%mm0 \n\t" /* FF R1 G1 B1 FF R0 G0 B0 */ \
- "punpckhwd %%mm2, %%mm3 \n\t" /* FF R3 G3 B3 FF R2 G2 B2 */ \
- MOVNTQ" %%mm0, %0 \n\t" \
- MOVNTQ" %%mm3, 8%0 \n\t" \
- static inline void RENAME(rgb15to32)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint16_t *end;
- const uint16_t *mm_end;
- uint8_t *d = dst;
- const uint16_t *s = (const uint16_t *)src;
- end = s + src_size/2;
- __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
- __asm__ volatile("pxor %%mm7,%%mm7 \n\t":::"memory");
- __asm__ volatile("pcmpeqd %%mm6,%%mm6 \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 $5, %%mm0 \n\t"
- "pmulhw %5, %%mm0 \n\t"
- "pmulhw %5, %%mm1 \n\t"
- "pmulhw %6, %%mm2 \n\t"
- PACK_RGB32
- :"=m"(*d)
- :"m"(*s),"m"(mask15b),"m"(mask15g),"m"(mask15r),"m"(mul15_mid),"m"(mul15_hi)
- :"memory");
- d += 16;
- s += 4;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- while (s < end) {
- register uint16_t bgr;
- bgr = *s++;
- *d++ = ((bgr&0x1F)<<3) | ((bgr&0x1F)>>2);
- *d++ = ((bgr&0x3E0)>>2) | ((bgr&0x3E0)>>7);
- *d++ = ((bgr&0x7C00)>>7) | ((bgr&0x7C00)>>12);
- *d++ = 255;
- }
- }
- static inline void RENAME(rgb16to32)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- const uint16_t *end;
- const uint16_t *mm_end;
- uint8_t *d = dst;
- const uint16_t *s = (const uint16_t*)src;
- end = s + src_size/2;
- __asm__ volatile(PREFETCH" %0"::"m"(*s):"memory");
- __asm__ volatile("pxor %%mm7,%%mm7 \n\t":::"memory");
- __asm__ volatile("pcmpeqd %%mm6,%%mm6 \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 $5, %%mm0 \n\t"
- "psrlq $1, %%mm2 \n\t"
- "pmulhw %5, %%mm0 \n\t"
- "pmulhw %7, %%mm1 \n\t"
- "pmulhw %6, %%mm2 \n\t"
- PACK_RGB32
- :"=m"(*d)
- :"m"(*s),"m"(mask16b),"m"(mask16g),"m"(mask16r),"m"(mul15_mid),"m"(mul15_hi),"m"(mul16_mid)
- :"memory");
- d += 16;
- s += 4;
- }
- __asm__ volatile(SFENCE:::"memory");
- __asm__ volatile(EMMS:::"memory");
- while (s < end) {
- register uint16_t bgr;
- bgr = *s++;
- *d++ = ((bgr&0x1F)<<3) | ((bgr&0x1F)>>2);
- *d++ = ((bgr&0x7E0)>>3) | ((bgr&0x7E0)>>9);
- *d++ = ((bgr&0xF800)>>8) | ((bgr&0xF800)>>13);
- *d++ = 255;
- }
- }
- static inline void RENAME(shuffle_bytes_2103)(const uint8_t *src, uint8_t *dst, int src_size)
- {
- x86_reg idx = 15 - src_size;
- const uint8_t *s = src-idx;
- uint8_t *d = dst-idx;
- __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"
- ".p2align 4 \n\t"
- "1: \n\t"
- PREFETCH" 32(%1, %0) \n\t"
- "movq (%1, %0), %%mm0 \n\t"
- "movq 8(%1, %0), %%mm1 \n\t"
- # if COMPILE_TEMPLATE_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");
- 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, int src_size)
- {
- unsigned i;
- x86_reg 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"
- ".p2align 4 \n\t"
- "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;
- for (i=0; i<src_size; i+=3) {
- register uint8_t x;
- x = src[i + 2];
- dst[i + 1] = src[i + 1];
- dst[i + 2] = src[i + 0];
- dst[i + 0] = x;
- }
- }
- static inline void RENAME(yuvPlanartoyuy2)(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)
- {
- int y;
- const x86_reg chromWidth= width>>1;
- for (y=0; y<height; y++) {
- //FIXME handle 2 lines at once (fewer prefetches, reuse some chroma, but very likely memory-limited anyway)
- __asm__ volatile(
- "xor %%"REG_a", %%"REG_a" \n\t"
- ".p2align 4 \n\t"
- "1: \n\t"
- PREFETCH" 32(%1, %%"REG_a", 2) \n\t"
- PREFETCH" 32(%2, %%"REG_a") \n\t"
- PREFETCH" 32(%3, %%"REG_a") \n\t"
- "movq (%2, %%"REG_a"), %%mm0 \n\t" // U(0)
- "movq %%mm0, %%mm2 \n\t" // U(0)
- "movq (%3, %%"REG_a"), %%mm1 \n\t" // V(0)
- "punpcklbw %%mm1, %%mm0 \n\t" // UVUV UVUV(0)
- "punpckhbw %%mm1, %%mm2 \n\t" // UVUV UVUV(8)
- "movq (%1, %%"REG_a",2), %%mm3 \n\t" // Y(0)
- "movq 8(%1, %%"REG_a",2), %%mm5 \n\t" // Y(8)
- "movq %%mm3, %%mm4 \n\t" // Y(0)
- "movq %%mm5, %%mm6 \n\t" // Y(8)
- "punpcklbw %%mm0, %%mm3 \n\t" // YUYV YUYV(0)
- "punpckhbw %%mm0, %%mm4 \n\t" // YUYV YUYV(4)
- "punpcklbw %%mm2, %%mm5 \n\t" // YUYV YUYV(8)
- "punpckhbw %%mm2, %%mm6 \n\t" // YUYV YUYV(12)
- MOVNTQ" %%mm3, (%0, %%"REG_a", 4) \n\t"
- MOVNTQ" %%mm4, 8(%0, %%"REG_a", 4) \n\t"
- MOVNTQ" %%mm5, 16(%0, %%"REG_a", 4) \n\t"
- MOVNTQ" %%mm6, 24(%0, %%"REG_a", 4) \n\t"
- "add $8, %%"REG_a" \n\t"
- "cmp %4, %%"REG_a" \n\t"
- " jb 1b \n\t"
- ::"r"(dst), "r"(ysrc), "r"(usrc), "r"(vsrc), "g" (chromWidth)
- : "%"REG_a
- );
- if ((y&(vertLumPerChroma-1)) == vertLumPerChroma-1) {
- usrc += chromStride;
- vsrc += chromStride;
- }
- ysrc += lumStride;
- dst += dstStride;
- }
- __asm__(EMMS" …
Large files files are truncated, but you can click here to view the full file