📄 sdl_yuv_mmx.c
字号:
"pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
"punpckhbw %%mm2,%%mm6\n" // 0 R3 0 R2 0 R1 0 R0
"punpckhbw %%mm1,%%mm5\n" // G3 B3 G2 B2 G1 B1 G0 B0
"movq %%mm5,%%mm1\n" // G3 B3 G2 B2 G1 B1 G0 B0
"punpcklwd %%mm6,%%mm1\n" // 0 R1 G1 B1 0 R0 G0 B0
"movq %%mm1,(%5)\n" // wrote out ! row2
"punpckhwd %%mm6,%%mm5\n" // 0 R3 G3 B3 0 R2 G2 B2
"movq %%mm5,8(%5)\n" // wrote out ! row2
"addl $4,%2\n" // lum+4
"leal 16(%3),%3\n" // row1+16
"leal 16(%5),%5\n" // row2+16
"addl $2, %%ebx\n" // cr+2
"addl $2, %1\n" // cb+2
"addl $4,%6\n" // x+4
"cmpl %4,%6\n"
"jl 1b\n"
"addl %4, %2\n" // lum += cols
"addl %8, %3\n" // row1+= mod
"addl %8, %5\n" // row2+= mod
"movl $0, %6\n" // x=0
"cmpl %7, %2\n"
"jl 1b\n"
"emms\n"
"popl %%ebx\n"
:
: "m" (cr), "r"(cb),"r"(lum),
"r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)
: "%ebx"
);
}
void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
unsigned char *lum, unsigned char *cr,
unsigned char *cb, unsigned char *out,
int rows, int cols, int mod )
{
Uint16 *row1;
Uint16 *row2;
unsigned char* y = lum +cols*rows; /* Pointer to the end */
int x=0;
row1 = (Uint16 *)out; /* 16 bit target */
row2 = (Uint16 *)out+cols+mod; /* start of second row */
mod = (mod+cols+mod)*2; /* increment for row1 in byte */
__asm__ __volatile__(
"pushl %%ebx\n"
"movl %0, %%ebx\n"
".align 8\n"
"1:\n"
"movd (%1), %%mm0\n" // 4 Cb 0 0 0 0 u3 u2 u1 u0
"pxor %%mm7, %%mm7\n"
"movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0
"punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0
"punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0
"psubw _MMX_0080w, %%mm0\n"
"psubw _MMX_0080w, %%mm1\n"
"movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0
"movq %%mm1, %%mm3\n" // Cr
"pmullw _MMX_Ugrn565, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0
"movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0
"pmullw _MMX_Ublu5x5, %%mm0\n" // Cb2blue
"pand _MMX_00FFw, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0
"pmullw _MMX_Vgrn565, %%mm3\n" // Cr2green
"movq (%2), %%mm7\n" // L2
"pmullw _MMX_Vred5x5, %%mm1\n" // Cr2red
"psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1
"pmullw _MMX_Ycoeff, %%mm6\n" // lum1
"paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green
"pmullw _MMX_Ycoeff, %%mm7\n" // lum2
"movq %%mm6, %%mm4\n" // lum1
"paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0
"movq %%mm4, %%mm5\n" // lum1
"paddw %%mm1, %%mm4\n" // lum1 +red 00 R6 00 R4 00 R2 00 R0
"paddw %%mm2, %%mm5\n" // lum1 +green 00 G6 00 G4 00 G2 00 G0
"psraw $6, %%mm4\n" // R1 0 .. 64
"movq %%mm7, %%mm3\n" // lum2 00 L7 00 L5 00 L3 00 L1
"psraw $6, %%mm5\n" // G1 - .. +
"paddw %%mm0, %%mm7\n" // Lum2 +blue 00 B7 00 B5 00 B3 00 B1
"psraw $6, %%mm6\n" // B1 0 .. 64
"packuswb %%mm4, %%mm4\n" // R1 R1
"packuswb %%mm5, %%mm5\n" // G1 G1
"packuswb %%mm6, %%mm6\n" // B1 B1
"punpcklbw %%mm4, %%mm4\n"
"punpcklbw %%mm5, %%mm5\n"
"pand _MMX_red565, %%mm4\n"
"psllw $3, %%mm5\n" // GREEN 1
"punpcklbw %%mm6, %%mm6\n"
"pand _MMX_grn565, %%mm5\n"
"pand _MMX_red565, %%mm6\n"
"por %%mm5, %%mm4\n" //
"psrlw $11, %%mm6\n" // BLUE 1
"movq %%mm3, %%mm5\n" // lum2
"paddw %%mm1, %%mm3\n" // lum2 +red 00 R7 00 R5 00 R3 00 R1
"paddw %%mm2, %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1
"psraw $6, %%mm3\n" // R2
"por %%mm6, %%mm4\n" // MM4
"psraw $6, %%mm5\n" // G2
"movq (%2, %4), %%mm6\n" // L3 load lum2
"psraw $6, %%mm7\n"
"packuswb %%mm3, %%mm3\n"
"packuswb %%mm5, %%mm5\n"
"packuswb %%mm7, %%mm7\n"
"pand _MMX_00FFw, %%mm6\n" // L3
"punpcklbw %%mm3, %%mm3\n"
"punpcklbw %%mm5, %%mm5\n"
"pmullw _MMX_Ycoeff, %%mm6\n" // lum3
"punpcklbw %%mm7, %%mm7\n"
"psllw $3, %%mm5\n" // GREEN 2
"pand _MMX_red565, %%mm7\n"
"pand _MMX_red565, %%mm3\n"
"psrlw $11, %%mm7\n" // BLUE 2
"pand _MMX_grn565, %%mm5\n"
"por %%mm7, %%mm3\n"
"movq (%2,%4), %%mm7\n" // L4 load lum2
"por %%mm5, %%mm3\n" //
"psrlw $8, %%mm7\n" // L4
"movq %%mm4, %%mm5\n"
"punpcklwd %%mm3, %%mm4\n"
"pmullw _MMX_Ycoeff, %%mm7\n" // lum4
"punpckhwd %%mm3, %%mm5\n"
"movq %%mm4, (%3)\n" // write row1
"movq %%mm5, 8(%3)\n" // write row1
"movq %%mm6, %%mm4\n" // Lum3
"paddw %%mm0, %%mm6\n" // Lum3 +blue
"movq %%mm4, %%mm5\n" // Lum3
"paddw %%mm1, %%mm4\n" // Lum3 +red
"paddw %%mm2, %%mm5\n" // Lum3 +green
"psraw $6, %%mm4\n"
"movq %%mm7, %%mm3\n" // Lum4
"psraw $6, %%mm5\n"
"paddw %%mm0, %%mm7\n" // Lum4 +blue
"psraw $6, %%mm6\n" // Lum3 +blue
"movq %%mm3, %%mm0\n" // Lum4
"packuswb %%mm4, %%mm4\n"
"paddw %%mm1, %%mm3\n" // Lum4 +red
"packuswb %%mm5, %%mm5\n"
"paddw %%mm2, %%mm0\n" // Lum4 +green
"packuswb %%mm6, %%mm6\n"
"punpcklbw %%mm4, %%mm4\n"
"punpcklbw %%mm5, %%mm5\n"
"punpcklbw %%mm6, %%mm6\n"
"psllw $3, %%mm5\n" // GREEN 3
"pand _MMX_red565, %%mm4\n"
"psraw $6, %%mm3\n" // psr 6
"psraw $6, %%mm0\n"
"pand _MMX_red565, %%mm6\n" // BLUE
"pand _MMX_grn565, %%mm5\n"
"psrlw $11, %%mm6\n" // BLUE 3
"por %%mm5, %%mm4\n"
"psraw $6, %%mm7\n"
"por %%mm6, %%mm4\n"
"packuswb %%mm3, %%mm3\n"
"packuswb %%mm0, %%mm0\n"
"packuswb %%mm7, %%mm7\n"
"punpcklbw %%mm3, %%mm3\n"
"punpcklbw %%mm0, %%mm0\n"
"punpcklbw %%mm7, %%mm7\n"
"pand _MMX_red565, %%mm3\n"
"pand _MMX_red565, %%mm7\n" // BLUE
"psllw $3, %%mm0\n" // GREEN 4
"psrlw $11, %%mm7\n"
"pand _MMX_grn565, %%mm0\n"
"por %%mm7, %%mm3\n"
"por %%mm0, %%mm3\n"
"movq %%mm4, %%mm5\n"
"punpcklwd %%mm3, %%mm4\n"
"punpckhwd %%mm3, %%mm5\n"
"movq %%mm4, (%5)\n"
"movq %%mm5, 8(%5)\n"
"addl $8, %6\n"
"addl $8, %2\n"
"addl $4, %%ebx\n"
"addl $4, %1\n"
"cmpl %4, %6\n"
"leal 16(%3), %3\n"
"leal 16(%5),%5\n" // row2+16
"jl 1b\n"
"addl %4, %2\n" // lum += cols
"addl %8, %3\n" // row1+= mod
"addl %8, %5\n" // row2+= mod
"movl $0, %6\n" // x=0
"cmpl %7, %2\n"
"jl 1b\n"
"emms\n"
"popl %%ebx\n"
:
:"m" (cr), "r"(cb),"r"(lum),
"r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod)
: "%ebx"
);
}
#endif /* GCC i386 inline assembly */
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -