dsputil_mmx_avg.h

来自「这是著名的TCPMP播放器在WINDWOWS,和WINCE下编译通过的源程序.笔」· C头文件 代码 · 共 821 行 · 第 1/2 页

H
821 行
字号
12
	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
	:"S"((long)src1Stride), "D"((long)dstStride)
	:"memory"); 
//the following should be used, though better not with gcc ...
/*	:"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
	:"r"(src1Stride), "r"(dstStride)
	:"memory");*/
}

static void DEF(avg_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
{
    __asm __volatile(
	"testl $1, %0			\n\t"
	    " jz 1f				\n\t"
	"movq	(%1), %%mm0		\n\t"
	"movq	8(%1), %%mm1		\n\t"
	PAVGB" (%2), %%mm0		\n\t"
	PAVGB" 8(%2), %%mm1		\n\t"
	"add	%4, %1			\n\t"
	"add	$16, %2			\n\t"
	PAVGB" (%3), %%mm0		\n\t"
	PAVGB" 8(%3), %%mm1		\n\t"
	"movq	%%mm0, (%3)		\n\t"
	"movq	%%mm1, 8(%3)		\n\t"
	"add	%5, %3			\n\t"
	"decl	%0			\n\t"
	"1:				\n\t"
	"movq	(%1), %%mm0		\n\t"
	"movq	8(%1), %%mm1		\n\t"
	"add	%4, %1			\n\t"
	PAVGB" (%2), %%mm0		\n\t"
	PAVGB" 8(%2), %%mm1		\n\t"
	PAVGB" (%3), %%mm0		\n\t"
	PAVGB" 8(%3), %%mm1		\n\t"
	"movq	%%mm0, (%3)		\n\t"
	"movq	%%mm1, 8(%3)		\n\t"
	"add	%5, %3			\n\t"
	"movq	(%1), %%mm0		\n\t"
	"movq	8(%1), %%mm1		\n\t"
	"add	%4, %1			\n\t"
	PAVGB" 16(%2), %%mm0		\n\t"
	PAVGB" 24(%2), %%mm1		\n\t"
	PAVGB" (%3), %%mm0		\n\t"
	PAVGB" 8(%3), %%mm1		\n\t"
	"movq	%%mm0, (%3)		\n\t"
	"movq	%%mm1, 8(%3)		\n\t"
	"add	%5, %3			\n\t"
	"add	$32, %2			\n\t"
	"subl	$2, %0			\n\t"
	"jnz	1b			\n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
	:"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#else
	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
	:"S"((long)src1Stride), "D"((long)dstStride)
	:"memory"); 
//the following should be used, though better not with gcc ...
/*	:"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
	:"r"(src1Stride), "r"(dstStride)
	:"memory");*/
}

static void DEF(put_no_rnd_pixels16_l2)(uint8_t *dst, uint8_t *src1, uint8_t *src2, int dstStride, int src1Stride, int h)
{
    __asm __volatile(
	"pcmpeqb %%mm6, %%mm6\n\t"
	"testl $1, %0			\n\t"
	    " jz 1f				\n\t"
	"movq	(%1), %%mm0		\n\t"
	"movq	8(%1), %%mm1		\n\t"
	"movq	(%2), %%mm2		\n\t"
	"movq	8(%2), %%mm3		\n\t"
	"pxor %%mm6, %%mm0		\n\t"
	"pxor %%mm6, %%mm1		\n\t"
	"pxor %%mm6, %%mm2		\n\t"
	"pxor %%mm6, %%mm3		\n\t"
	PAVGB" %%mm2, %%mm0		\n\t"
	PAVGB" %%mm3, %%mm1		\n\t"
	"pxor %%mm6, %%mm0		\n\t"
	"pxor %%mm6, %%mm1		\n\t"
	"add	%4, %1			\n\t"
	"add	$16, %2			\n\t"
	"movq	%%mm0, (%3)		\n\t"
	"movq	%%mm1, 8(%3)		\n\t"
	"add	%5, %3			\n\t"
	"decl	%0			\n\t"
	"1:				\n\t"
	"movq	(%1), %%mm0		\n\t"
	"movq	8(%1), %%mm1		\n\t"
	"add	%4, %1			\n\t"
	"movq	(%2), %%mm2		\n\t"
	"movq	8(%2), %%mm3		\n\t"
	"pxor %%mm6, %%mm0		\n\t"
	"pxor %%mm6, %%mm1		\n\t"
	"pxor %%mm6, %%mm2		\n\t"
	"pxor %%mm6, %%mm3		\n\t"
	PAVGB" %%mm2, %%mm0		\n\t"
	PAVGB" %%mm3, %%mm1		\n\t"
	"pxor %%mm6, %%mm0		\n\t"
	"pxor %%mm6, %%mm1		\n\t"
	"movq	%%mm0, (%3)		\n\t"
	"movq	%%mm1, 8(%3)		\n\t"
	"add	%5, %3			\n\t"
	"movq	(%1), %%mm0		\n\t"
	"movq	8(%1), %%mm1		\n\t"
	"add	%4, %1			\n\t"
	"movq	16(%2), %%mm2		\n\t"
	"movq	24(%2), %%mm3		\n\t"
	"pxor %%mm6, %%mm0		\n\t"
	"pxor %%mm6, %%mm1		\n\t"
	"pxor %%mm6, %%mm2		\n\t"
	"pxor %%mm6, %%mm3		\n\t"
	PAVGB" %%mm2, %%mm0		\n\t"
	PAVGB" %%mm3, %%mm1		\n\t"
	"pxor %%mm6, %%mm0		\n\t"
	"pxor %%mm6, %%mm1		\n\t"
	"movq	%%mm0, (%3)		\n\t"
	"movq	%%mm1, 8(%3)		\n\t"
	"add	%5, %3			\n\t"
	"add	$32, %2			\n\t"
	"subl	$2, %0			\n\t"
	"jnz	1b			\n\t"
#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
	:"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#else
	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
#endif
	:"S"((long)src1Stride), "D"((long)dstStride)
	:"memory"); 
//the following should be used, though better not with gcc ...
/*	:"+g"(h), "+r"(src1), "+r"(src2), "+r"(dst)
	:"r"(src1Stride), "r"(dstStride)
	:"memory");*/
}
 
/* GL: this function does incorrect rounding if overflow */
static void DEF(put_no_rnd_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    MOVQ_BONE(mm6);
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"1:				\n\t"
	"movq (%1), %%mm0		\n\t"
	"movq (%1, %3), %%mm2		\n\t"
	"movq 1(%1), %%mm1		\n\t"
	"movq 1(%1, %3), %%mm3		\n\t"
	"add %%"REG_a", %1		\n\t"
	"psubusb %%mm6, %%mm0		\n\t"
	"psubusb %%mm6, %%mm2		\n\t"
	PAVGB" %%mm1, %%mm0		\n\t"
	PAVGB" %%mm3, %%mm2		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"movq (%1), %%mm0		\n\t"
	"movq 1(%1), %%mm1		\n\t"
	"movq (%1, %3), %%mm2		\n\t"
	"movq 1(%1, %3), %%mm3		\n\t"
	"add %%"REG_a", %2		\n\t"
	"add %%"REG_a", %1		\n\t"
	"psubusb %%mm6, %%mm0		\n\t"
	"psubusb %%mm6, %%mm2		\n\t"
	PAVGB" %%mm1, %%mm0		\n\t"
	PAVGB" %%mm3, %%mm2		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D"(block)
	:"r" ((long)line_size)
	:"%"REG_a, "memory");
}

static void DEF(put_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"movq (%1), %%mm0		\n\t"
	"sub %3, %2			\n\t"
	"1:				\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm2	\n\t"
	"add %%"REG_a", %1		\n\t"
	PAVGB" %%mm1, %%mm0		\n\t"
	PAVGB" %%mm2, %%mm1		\n\t"
	"movq %%mm0, (%2, %3)		\n\t"
	"movq %%mm1, (%2, %%"REG_a")	\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm0	\n\t"
	"add %%"REG_a", %2		\n\t"
	"add %%"REG_a", %1		\n\t"
	PAVGB" %%mm1, %%mm2		\n\t"
	PAVGB" %%mm0, %%mm1		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"movq %%mm1, (%2, %%"REG_a")	\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D" (block)
	:"r" ((long)line_size)
	:"%"REG_a, "memory");
}

/* GL: this function does incorrect rounding if overflow */
static void DEF(put_no_rnd_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    MOVQ_BONE(mm6);
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"movq (%1), %%mm0		\n\t"
	"sub %3, %2			\n\t"
	"1:				\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm2	\n\t"
	"add %%"REG_a", %1		\n\t"
	"psubusb %%mm6, %%mm1		\n\t"
	PAVGB" %%mm1, %%mm0		\n\t"
	PAVGB" %%mm2, %%mm1		\n\t"
	"movq %%mm0, (%2, %3)		\n\t"
	"movq %%mm1, (%2, %%"REG_a")	\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm0	\n\t"
	"add %%"REG_a", %2		\n\t"
	"add %%"REG_a", %1		\n\t"
	"psubusb %%mm6, %%mm1		\n\t"
	PAVGB" %%mm1, %%mm2		\n\t"
	PAVGB" %%mm0, %%mm1		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"movq %%mm1, (%2, %%"REG_a")	\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D" (block)
	:"r" ((long)line_size)
	:"%"REG_a, "memory");
}

static void DEF(avg_pixels8)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"1:				\n\t"
	"movq (%2), %%mm0		\n\t"
	"movq (%2, %3), %%mm1		\n\t"
	PAVGB" (%1), %%mm0		\n\t"
	PAVGB" (%1, %3), %%mm1		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm1, (%2, %3)		\n\t"
	"add %%"REG_a", %1		\n\t"
	"add %%"REG_a", %2		\n\t"
	"movq (%2), %%mm0		\n\t"
	"movq (%2, %3), %%mm1		\n\t"
	PAVGB" (%1), %%mm0		\n\t"
	PAVGB" (%1, %3), %%mm1		\n\t"
	"add %%"REG_a", %1		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm1, (%2, %3)		\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D"(block)
	:"r" ((long)line_size)
	:"%"REG_a, "memory");
}

static void DEF(avg_pixels8_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"1:				\n\t"
	"movq (%1), %%mm0		\n\t"
	"movq (%1, %3), %%mm2		\n\t"
	PAVGB" 1(%1), %%mm0		\n\t"
	PAVGB" 1(%1, %3), %%mm2		\n\t"
	PAVGB" (%2), %%mm0		\n\t"
	PAVGB" (%2, %3), %%mm2		\n\t"
	"add %%"REG_a", %1		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"movq (%1), %%mm0		\n\t"
	"movq (%1, %3), %%mm2		\n\t"
	PAVGB" 1(%1), %%mm0		\n\t"
	PAVGB" 1(%1, %3), %%mm2		\n\t"
	"add %%"REG_a", %2		\n\t"
	"add %%"REG_a", %1		\n\t"
	PAVGB" (%2), %%mm0		\n\t"
	PAVGB" (%2, %3), %%mm2		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D"(block)
	:"r" ((long)line_size)
	:"%"REG_a, "memory");
}

static void DEF(avg_pixels8_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"movq (%1), %%mm0		\n\t"
	"sub %3, %2			\n\t"
	"1:				\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm2	\n\t"
	"add %%"REG_a", %1		\n\t"
	PAVGB" %%mm1, %%mm0		\n\t"
	PAVGB" %%mm2, %%mm1		\n\t"
	"movq (%2, %3), %%mm3		\n\t"
	"movq (%2, %%"REG_a"), %%mm4	\n\t"
	PAVGB" %%mm3, %%mm0		\n\t"
	PAVGB" %%mm4, %%mm1		\n\t"
	"movq %%mm0, (%2, %3)		\n\t"
	"movq %%mm1, (%2, %%"REG_a")	\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm0	\n\t"
	PAVGB" %%mm1, %%mm2		\n\t"
	PAVGB" %%mm0, %%mm1		\n\t"
	"add %%"REG_a", %2		\n\t"
	"add %%"REG_a", %1		\n\t"
	"movq (%2, %3), %%mm3		\n\t"
	"movq (%2, %%"REG_a"), %%mm4	\n\t"
	PAVGB" %%mm3, %%mm2		\n\t"
	PAVGB" %%mm4, %%mm1		\n\t"
	"movq %%mm2, (%2, %3)		\n\t"
	"movq %%mm1, (%2, %%"REG_a")	\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D"(block)
	:"r" ((long)line_size)
	:"%"REG_a, "memory");
}

// Note this is not correctly rounded, but this function is only used for b frames so it doesnt matter 
static void DEF(avg_pixels8_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h)
{
    MOVQ_BONE(mm6);
    __asm __volatile(
	"lea (%3, %3), %%"REG_a"	\n\t"
	"movq (%1), %%mm0		\n\t"
	PAVGB" 1(%1), %%mm0		\n\t"
	".balign 8			\n\t"
	"1:				\n\t"
	"movq (%1, %%"REG_a"), %%mm2	\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"psubusb %%mm6, %%mm2		\n\t"
	PAVGB" 1(%1, %3), %%mm1		\n\t"
	PAVGB" 1(%1, %%"REG_a"), %%mm2	\n\t"
	"add %%"REG_a", %1		\n\t"
	PAVGB" %%mm1, %%mm0		\n\t"
	PAVGB" %%mm2, %%mm1		\n\t"
	PAVGB" (%2), %%mm0		\n\t"
	PAVGB" (%2, %3), %%mm1		\n\t"
	"movq %%mm0, (%2)		\n\t"
	"movq %%mm1, (%2, %3)		\n\t"
	"movq (%1, %3), %%mm1		\n\t"
	"movq (%1, %%"REG_a"), %%mm0	\n\t"
	PAVGB" 1(%1, %3), %%mm1		\n\t"
	PAVGB" 1(%1, %%"REG_a"), %%mm0	\n\t"
	"add %%"REG_a", %2		\n\t"
	"add %%"REG_a", %1		\n\t"
	PAVGB" %%mm1, %%mm2		\n\t"
	PAVGB" %%mm0, %%mm1		\n\t"
	PAVGB" (%2), %%mm2		\n\t"
	PAVGB" (%2, %3), %%mm1		\n\t"
	"movq %%mm2, (%2)		\n\t"
	"movq %%mm1, (%2, %3)		\n\t"
	"add %%"REG_a", %2		\n\t"
	"subl $4, %0			\n\t"
	"jnz 1b				\n\t"
	:"+g"(h), "+S"(pixels), "+D"(block)
	:"r" ((long)line_size)
	:"%"REG_a,  "memory");
}

//FIXME the following could be optimized too ...
static void DEF(put_no_rnd_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(put_no_rnd_pixels8_x2)(block  , pixels  , line_size, h);
    DEF(put_no_rnd_pixels8_x2)(block+8, pixels+8, line_size, h);
}
static void DEF(put_pixels16_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(put_pixels8_y2)(block  , pixels  , line_size, h);
    DEF(put_pixels8_y2)(block+8, pixels+8, line_size, h);
}
static void DEF(put_no_rnd_pixels16_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(put_no_rnd_pixels8_y2)(block  , pixels  , line_size, h);
    DEF(put_no_rnd_pixels8_y2)(block+8, pixels+8, line_size, h);
}
static void DEF(avg_pixels16)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(avg_pixels8)(block  , pixels  , line_size, h);
    DEF(avg_pixels8)(block+8, pixels+8, line_size, h);
}
static void DEF(avg_pixels16_x2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(avg_pixels8_x2)(block  , pixels  , line_size, h);
    DEF(avg_pixels8_x2)(block+8, pixels+8, line_size, h);
}
static void DEF(avg_pixels16_y2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(avg_pixels8_y2)(block  , pixels  , line_size, h);
    DEF(avg_pixels8_y2)(block+8, pixels+8, line_size, h);
}
static void DEF(avg_pixels16_xy2)(uint8_t *block, const uint8_t *pixels, int line_size, int h){
    DEF(avg_pixels8_xy2)(block  , pixels  , line_size, h);
    DEF(avg_pixels8_xy2)(block+8, pixels+8, line_size, h);
}
12

dsputil_mmx_avg.h - 源码说明

本页面展示了「这是著名的TCPMP播放器在WINDWOWS,和WINCE下编译通过的源程序.笔者对其中的LIBMAD库做了针对ARM MPU的优化. 并增加了词幕功能.」中的 dsputil_mmx_avg.h 源码文件，采用 C头文件 编程语言编写，共 821 行代码。您可以在线阅读完整代码内容，也可以返回资源详情页下载完整源码包进行本地学习和开发。

虫虫下载站收录了大量与WINDWOWS相关的技术资源，包括源代码、技术文档、电路图等，是电子工程师和嵌入式开发者的专业学习平台。

⌨️ 快捷键说明

复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?