rlm@1: #include <cstdlib>
rlm@1: #include <cstring>
rlm@1: #include "../Port.h"
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: extern "C" bool cpu_mmx;
rlm@1: #endif
rlm@1: 
rlm@1: /*
rlm@1:  * Thanks to Kawaks' Mr. K for the code
rlm@1: 
rlm@1:    Incorporated into vba by Anthony Di Franco
rlm@1:  */
rlm@1: 
rlm@1: static u8 *frm1 = NULL;
rlm@1: static u8 *frm2 = NULL;
rlm@1: static u8 *frm3 = NULL;
rlm@1: 
rlm@1: extern u32 RGB_LOW_BITS_MASK;
rlm@1: extern u32 qRGB_COLOR_MASK[2];
rlm@1: 
rlm@1: static void Init()
rlm@1: {
rlm@1: 	frm1 = (u8 *)calloc(322 * 242, 4);
rlm@1: 	// 1 frame ago
rlm@1: 	frm2 = (u8 *)calloc(322 * 242, 4);
rlm@1: 	// 2 frames ago
rlm@1: 	frm3 = (u8 *)calloc(322 * 242, 4);
rlm@1: 	// 3 frames ago
rlm@1: }
rlm@1: 
rlm@1: void InterframeCleanup()
rlm@1: {
rlm@1: 	if (frm1)
rlm@1: 		free(frm1);
rlm@1: 	if (frm2)
rlm@1: 		free(frm2);
rlm@1: 	if (frm3)
rlm@1: 		free(frm3);
rlm@1: 	frm1 = frm2 = frm3 = NULL;
rlm@1: }
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: static void SmartIB_MMX(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	u16 *src0 = (u16 *)srcPtr;
rlm@1: 	u16 *src1 = (u16 *)frm1;
rlm@1: 	u16 *src2 = (u16 *)frm2;
rlm@1: 	u16 *src3 = (u16 *)frm3;
rlm@1: 
rlm@1: 	int count = width >> 2;
rlm@1: 
rlm@1: 	for (int i = 0; i < height; i++)
rlm@1: 	{
rlm@1: #ifdef __GNUC__
rlm@1: 		asm volatile (
rlm@1: 		    "push %4\n"
rlm@1: 		    "movq 0(%5), %%mm7\n"       // colorMask
rlm@1: 		    "0:\n"
rlm@1: 		    "movq 0(%0), %%mm0\n"       // src0
rlm@1: 		    "movq 0(%1), %%mm1\n"       // src1
rlm@1: 		    "movq 0(%2), %%mm2\n"       // src2
rlm@1: 		    "movq 0(%3), %%mm3\n"       // src3
rlm@1: 		    "movq %%mm0, 0(%3)\n"       // src3 = src0
rlm@1: 		    "movq %%mm0, %%mm4\n"
rlm@1: 		    "movq %%mm1, %%mm5\n"
rlm@1: 		    "pcmpeqw %%mm2, %%mm5\n"       // src1 == src2 (A)
rlm@1: 		    "pcmpeqw %%mm3, %%mm4\n"       // src3 == src0 (B)
rlm@1: 		    "por %%mm5, %%mm4\n"       // A | B
rlm@1: 		    "movq %%mm2, %%mm5\n"
rlm@1: 		    "pcmpeqw %%mm0, %%mm5\n"       // src0 == src2 (C)
rlm@1: 		    "pcmpeqw %%mm1, %%mm3\n"       // src1 == src3 (D)
rlm@1: 		    "por %%mm3, %%mm5\n"       // C|D
rlm@1: 		    "pandn %%mm5, %%mm4\n"       // (!(A|B))&(C|D)
rlm@1: 		    "movq %%mm0, %%mm2\n"
rlm@1: 		    "pand %%mm7, %%mm2\n"       // color & colorMask
rlm@1: 		    "pand %%mm7, %%mm1\n"       // src1 & colorMask
rlm@1: 		    "psrlw $1, %%mm2\n"       // (color & colorMask) >> 1 (E)
rlm@1: 		    "psrlw $1, %%mm1\n"       // (src & colorMask) >> 1 (F)
rlm@1: 		    "paddw %%mm2, %%mm1\n"       // E+F
rlm@1: 		    "pand %%mm4, %%mm1\n"       // (E+F) & res
rlm@1: 		    "pandn %%mm0, %%mm4\n"       // color& !res
rlm@1: 
rlm@1: 		    "por %%mm1, %%mm4\n"
rlm@1: 		    "movq %%mm4, 0(%0)\n"       // src0 = res
rlm@1: 
rlm@1: 		    "addl $8, %0\n"
rlm@1: 		    "addl $8, %1\n"
rlm@1: 		    "addl $8, %2\n"
rlm@1: 		    "addl $8, %3\n"
rlm@1: 
rlm@1: 		    "decl %4\n"
rlm@1: 		    "jnz 0b\n"
rlm@1: 		    "pop %4\n"
rlm@1: 		    "emms\n"
rlm@1: 			: "+r" (src0), "+r" (src1), "+r" (src2), "+r" (src3)
rlm@1: 			: "r" (count), "r" (qRGB_COLOR_MASK)
rlm@1: 		    );
rlm@1: #else
rlm@1: 		__asm {
rlm@1: 			movq mm7, qword ptr [qRGB_COLOR_MASK];
rlm@1: 			mov	 eax, src0;
rlm@1: 			mov	 ebx, src1;
rlm@1: 			mov	 ecx, src2;
rlm@1: 			mov	 edx, src3;
rlm@1: 			mov	 edi, count;
rlm@1: label0:
rlm@1: 			movq mm0, qword ptr [eax]; // src0
rlm@1: 			movq	   mm1, qword ptr [ebx]; // src1
rlm@1: 			movq	   mm2, qword ptr [ecx]; // src2
rlm@1: 			movq	   mm3, qword ptr [edx]; // src3
rlm@1: 			movq qword ptr [edx], mm0; // src3 = src0
rlm@1: 			movq	   mm4, mm0;
rlm@1: 			movq	   mm5, mm1;
rlm@1: 			pcmpeqw	   mm5, mm2; // src1 == src2 (A)
rlm@1: 			pcmpeqw	   mm4, mm3; // src3 == src0 (B)
rlm@1: 			por		   mm4, mm5; // A | B
rlm@1: 			movq	   mm5, mm2;
rlm@1: 			pcmpeqw	   mm5, mm0; // src0 == src2 (C)
rlm@1: 			pcmpeqw	   mm3, mm1; // src1 == src3 (D)
rlm@1: 			por		   mm5, mm3; // C|D
rlm@1: 			pandn	   mm4, mm5; // (!(A|B))&(C|D)
rlm@1: 			movq	   mm2, mm0;
rlm@1: 			pand	   mm2, mm7; // color & colorMask
rlm@1: 			pand	   mm1, mm7; // src1 & colorMask
rlm@1: 			psrlw	   mm2, 1; // (color & colorMask) >> 1 (E)
rlm@1: 			psrlw	   mm1, 1; // (src & colorMask) >> 1 (F)
rlm@1: 			paddw	   mm1, mm2; // E+F
rlm@1: 			pand	   mm1, mm4; // (E+F) & res
rlm@1: 			pandn	   mm4, mm0; // color & !res
rlm@1: 
rlm@1: 			por		   mm4, mm1;
rlm@1: 			movq qword ptr [eax], mm4; // src0 = res
rlm@1: 
rlm@1: 			add eax, 8;
rlm@1: 			add ebx, 8;
rlm@1: 			add ecx, 8;
rlm@1: 			add edx, 8;
rlm@1: 
rlm@1: 			dec edi;
rlm@1: 			jnz label0;
rlm@1: 			mov src0, eax;
rlm@1: 			mov src1, ebx;
rlm@1: 			mov src2, ecx;
rlm@1: 			mov src3, edx;
rlm@1: 			emms;
rlm@1: 		}
rlm@1: #endif
rlm@1: 		src0 += 2;
rlm@1: 		src1 += 2;
rlm@1: 		src2 += 2;
rlm@1: 		src3 += 2;
rlm@1: 	}
rlm@1: 
rlm@1: 	/* Swap buffers around */
rlm@1: 	u8 *temp = frm1;
rlm@1: 	frm1 = frm3;
rlm@1: 	frm3 = frm2;
rlm@1: 	frm2 = temp;
rlm@1: }
rlm@1: 
rlm@1: #endif
rlm@1: 
rlm@1: void SmartIB(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	if (frm1 == NULL)
rlm@1: 	{
rlm@1: 		Init();
rlm@1: 	}
rlm@1: #ifdef MMX
rlm@1: 	if (cpu_mmx)
rlm@1: 	{
rlm@1: 		SmartIB_MMX(srcPtr, srcPitch, width, height);
rlm@1: 		return;
rlm@1: 	}
rlm@1: #endif
rlm@1: 
rlm@1: 	u16 colorMask = ~RGB_LOW_BITS_MASK;
rlm@1: 
rlm@1: 	u16 *src0 = (u16 *)srcPtr;
rlm@1: 	u16 *src1 = (u16 *)frm1;
rlm@1: 	u16 *src2 = (u16 *)frm2;
rlm@1: 	u16 *src3 = (u16 *)frm3;
rlm@1: 
rlm@1: 	int sPitch = srcPitch >> 1;
rlm@1: 
rlm@1: 	int pos = 0;
rlm@1: 	for (int j = 0; j < height; j++)
rlm@1: 		for (int i = 0; i < sPitch; i++)
rlm@1: 		{
rlm@1: 			u16 color = src0[pos];
rlm@1: 			src0[pos] =
rlm@1: 			    (src1[pos] != src2[pos]) &&
rlm@1: 			    (src3[pos] != color) &&
rlm@1: 			    ((color == src2[pos]) || (src1[pos] == src3[pos]))
rlm@1: 			    ? (((color & colorMask) >> 1) + ((src1[pos] & colorMask) >> 1)) :
rlm@1: 			    color;
rlm@1: 			src3[pos] = color; /* oldest buffer now holds newest frame */
rlm@1: 			pos++;
rlm@1: 		}
rlm@1: 
rlm@1: 	/* Swap buffers around */
rlm@1: 	u8 *temp = frm1;
rlm@1: 	frm1 = frm3;
rlm@1: 	frm3 = frm2;
rlm@1: 	frm2 = temp;
rlm@1: }
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: static void SmartIB32_MMX(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	u32 *src0 = (u32 *)srcPtr;
rlm@1: 	u32 *src1 = (u32 *)frm1;
rlm@1: 	u32 *src2 = (u32 *)frm2;
rlm@1: 	u32 *src3 = (u32 *)frm3;
rlm@1: 
rlm@1: 	int count = width >> 1;
rlm@1: 
rlm@1: 	for (int i = 0; i < height; i++)
rlm@1: 	{
rlm@1: #ifdef __GNUC__
rlm@1: 		asm volatile (
rlm@1: 		    "push %4\n"
rlm@1: 		    "movq 0(%5), %%mm7\n"       // colorMask
rlm@1: 		    "0:\n"
rlm@1: 		    "movq 0(%0), %%mm0\n"       // src0
rlm@1: 		    "movq 0(%1), %%mm1\n"       // src1
rlm@1: 		    "movq 0(%2), %%mm2\n"       // src2
rlm@1: 		    "movq 0(%3), %%mm3\n"       // src3
rlm@1: 		    "movq %%mm0, 0(%3)\n"       // src3 = src0
rlm@1: 		    "movq %%mm0, %%mm4\n"
rlm@1: 		    "movq %%mm1, %%mm5\n"
rlm@1: 		    "pcmpeqd %%mm2, %%mm5\n"       // src1 == src2 (A)
rlm@1: 		    "pcmpeqd %%mm3, %%mm4\n"       // src3 == src0 (B)
rlm@1: 		    "por %%mm5, %%mm4\n"       // A | B
rlm@1: 		    "movq %%mm2, %%mm5\n"
rlm@1: 		    "pcmpeqd %%mm0, %%mm5\n"       // src0 == src2 (C)
rlm@1: 		    "pcmpeqd %%mm1, %%mm3\n"       // src1 == src3 (D)
rlm@1: 		    "por %%mm3, %%mm5\n"       // C|D
rlm@1: 		    "pandn %%mm5, %%mm4\n"       // (!(A|B))&(C|D)
rlm@1: 		    "movq %%mm0, %%mm2\n"
rlm@1: 		    "pand %%mm7, %%mm2\n"       // color & colorMask
rlm@1: 		    "pand %%mm7, %%mm1\n"       // src1 & colorMask
rlm@1: 		    "psrld $1, %%mm2\n"       // (color & colorMask) >> 1 (E)
rlm@1: 		    "psrld $1, %%mm1\n"       // (src & colorMask) >> 1 (F)
rlm@1: 		    "paddd %%mm2, %%mm1\n"       // E+F
rlm@1: 		    "pand %%mm4, %%mm1\n"       // (E+F) & res
rlm@1: 		    "pandn %%mm0, %%mm4\n"       // color& !res
rlm@1: 
rlm@1: 		    "por %%mm1, %%mm4\n"
rlm@1: 		    "movq %%mm4, 0(%0)\n"       // src0 = res
rlm@1: 
rlm@1: 		    "addl $8, %0\n"
rlm@1: 		    "addl $8, %1\n"
rlm@1: 		    "addl $8, %2\n"
rlm@1: 		    "addl $8, %3\n"
rlm@1: 
rlm@1: 		    "decl %4\n"
rlm@1: 		    "jnz 0b\n"
rlm@1: 		    "pop %4\n"
rlm@1: 		    "emms\n"
rlm@1: 			: "+r" (src0), "+r" (src1), "+r" (src2), "+r" (src3)
rlm@1: 			: "r" (count), "r" (qRGB_COLOR_MASK)
rlm@1: 		    );
rlm@1: #else
rlm@1: 		__asm {
rlm@1: 			movq mm7, qword ptr [qRGB_COLOR_MASK];
rlm@1: 			mov	 eax, src0;
rlm@1: 			mov	 ebx, src1;
rlm@1: 			mov	 ecx, src2;
rlm@1: 			mov	 edx, src3;
rlm@1: 			mov	 edi, count;
rlm@1: label0:
rlm@1: 			movq mm0, qword ptr [eax]; // src0
rlm@1: 			movq	   mm1, qword ptr [ebx]; // src1
rlm@1: 			movq	   mm2, qword ptr [ecx]; // src2
rlm@1: 			movq	   mm3, qword ptr [edx]; // src3
rlm@1: 			movq qword ptr [edx], mm0; // src3 = src0
rlm@1: 			movq	   mm4, mm0;
rlm@1: 			movq	   mm5, mm1;
rlm@1: 			pcmpeqd	   mm5, mm2; // src1 == src2 (A)
rlm@1: 			pcmpeqd	   mm4, mm3; // src3 == src0 (B)
rlm@1: 			por		   mm4, mm5; // A | B
rlm@1: 			movq	   mm5, mm2;
rlm@1: 			pcmpeqd	   mm5, mm0; // src0 == src2 (C)
rlm@1: 			pcmpeqd	   mm3, mm1; // src1 == src3 (D)
rlm@1: 			por		   mm5, mm3; // C|D
rlm@1: 			pandn	   mm4, mm5; // (!(A|B))&(C|D)
rlm@1: 			movq	   mm2, mm0;
rlm@1: 			pand	   mm2, mm7; // color & colorMask
rlm@1: 			pand	   mm1, mm7; // src1 & colorMask
rlm@1: 			psrld	   mm2, 1; // (color & colorMask) >> 1 (E)
rlm@1: 			psrld	   mm1, 1; // (src & colorMask) >> 1 (F)
rlm@1: 			paddd	   mm1, mm2; // E+F
rlm@1: 			pand	   mm1, mm4; // (E+F) & res
rlm@1: 			pandn	   mm4, mm0; // color & !res
rlm@1: 
rlm@1: 			por		   mm4, mm1;
rlm@1: 			movq qword ptr [eax], mm4; // src0 = res
rlm@1: 
rlm@1: 			add eax, 8;
rlm@1: 			add ebx, 8;
rlm@1: 			add ecx, 8;
rlm@1: 			add edx, 8;
rlm@1: 
rlm@1: 			dec edi;
rlm@1: 			jnz label0;
rlm@1: 			mov src0, eax;
rlm@1: 			mov src1, ebx;
rlm@1: 			mov src2, ecx;
rlm@1: 			mov src3, edx;
rlm@1: 			emms;
rlm@1: 		}
rlm@1: #endif
rlm@1: 
rlm@1: 		src0++;
rlm@1: 		src1++;
rlm@1: 		src2++;
rlm@1: 		src3++;
rlm@1: 	}
rlm@1: 	/* Swap buffers around */
rlm@1: 	u8 *temp = frm1;
rlm@1: 	frm1 = frm3;
rlm@1: 	frm3 = frm2;
rlm@1: 	frm2 = temp;
rlm@1: }
rlm@1: 
rlm@1: #endif
rlm@1: 
rlm@1: void SmartIB32(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	if (frm1 == NULL)
rlm@1: 	{
rlm@1: 		Init();
rlm@1: 	}
rlm@1: #ifdef MMX
rlm@1: 	if (cpu_mmx)
rlm@1: 	{
rlm@1: 		SmartIB32_MMX(srcPtr, srcPitch, width, height);
rlm@1: 		return;
rlm@1: 	}
rlm@1: #endif
rlm@1: 
rlm@1: 	u32 *src0 = (u32 *)srcPtr;
rlm@1: 	u32 *src1 = (u32 *)frm1;
rlm@1: 	u32 *src2 = (u32 *)frm2;
rlm@1: 	u32 *src3 = (u32 *)frm3;
rlm@1: 
rlm@1: 	u32 colorMask = 0xfefefe;
rlm@1: 
rlm@1: 	int sPitch = srcPitch >> 2;
rlm@1: 	int pos	   = 0;
rlm@1: 
rlm@1: 	for (int j = 0; j < height; j++)
rlm@1: 		for (int i = 0; i < sPitch; i++)
rlm@1: 		{
rlm@1: 			u32 color = src0[pos];
rlm@1: 			src0[pos] =
rlm@1: 			    (src1[pos] != src2[pos]) &&
rlm@1: 			    (src3[pos] != color) &&
rlm@1: 			    ((color == src2[pos]) || (src1[pos] == src3[pos]))
rlm@1: 			    ? (((color & colorMask) >> 1) + ((src1[pos] & colorMask) >> 1)) :
rlm@1: 			    color;
rlm@1: 			src3[pos] = color; /* oldest buffer now holds newest frame */
rlm@1: 			pos++;
rlm@1: 		}
rlm@1: 
rlm@1: 	/* Swap buffers around */
rlm@1: 	u8 *temp = frm1;
rlm@1: 	frm1 = frm3;
rlm@1: 	frm3 = frm2;
rlm@1: 	frm2 = temp;
rlm@1: }
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: static void MotionBlurIB_MMX(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	u16 *src0 = (u16 *)srcPtr;
rlm@1: 	u16 *src1 = (u16 *)frm1;
rlm@1: 
rlm@1: 	int count = width >> 2;
rlm@1: 
rlm@1: 	for (int i = 0; i < height; i++)
rlm@1: 	{
rlm@1: #ifdef __GNUC__
rlm@1: 		asm volatile (
rlm@1: 		    "push %2\n"
rlm@1: 		    "movq 0(%3), %%mm7\n"       // colorMask
rlm@1: 		    "0:\n"
rlm@1: 		    "movq 0(%0), %%mm0\n"       // src0
rlm@1: 		    "movq 0(%1), %%mm1\n"       // src1
rlm@1: 		    "movq %%mm0, 0(%1)\n"       // src1 = src0
rlm@1: 		    "pand %%mm7, %%mm0\n"       // color & colorMask
rlm@1: 		    "pand %%mm7, %%mm1\n"       // src1 & colorMask
rlm@1: 		    "psrlw $1, %%mm0\n"       // (color & colorMask) >> 1 (E)
rlm@1: 		    "psrlw $1, %%mm1\n"       // (src & colorMask) >> 1 (F)
rlm@1: 		    "paddw %%mm1, %%mm0\n"       // E+F
rlm@1: 
rlm@1: 		    "movq %%mm0, 0(%0)\n"       // src0 = res
rlm@1: 
rlm@1: 		    "addl $8, %0\n"
rlm@1: 		    "addl $8, %1\n"
rlm@1: 
rlm@1: 		    "decl %2\n"
rlm@1: 		    "jnz 0b\n"
rlm@1: 		    "pop %2\n"
rlm@1: 		    "emms\n"
rlm@1: 			: "+r" (src0), "+r" (src1)
rlm@1: 			: "r" (count), "r" (qRGB_COLOR_MASK)
rlm@1: 		    );
rlm@1: #else
rlm@1: 		__asm {
rlm@1: 			movq mm7, qword ptr [qRGB_COLOR_MASK];
rlm@1: 			mov	 eax, src0;
rlm@1: 			mov	 ebx, src1;
rlm@1: 			mov	 edi, count;
rlm@1: label0:
rlm@1: 			movq mm0, qword ptr [eax]; // src0
rlm@1: 			movq	   mm1, qword ptr [ebx]; // src1
rlm@1: 			movq qword ptr [ebx], mm0; // src1 = src0
rlm@1: 			pand	   mm0, mm7; // color & colorMask
rlm@1: 			pand	   mm1, mm7; // src1 & colorMask
rlm@1: 			psrlw	   mm0, 1; // (color & colorMask) >> 1 (E)
rlm@1: 			psrlw	   mm1, 1; // (src & colorMask) >> 1 (F)
rlm@1: 			paddw	   mm0, mm1; // E+F
rlm@1: 
rlm@1: 			movq qword ptr [eax], mm0; // src0 = res
rlm@1: 
rlm@1: 			add eax, 8;
rlm@1: 			add ebx, 8;
rlm@1: 
rlm@1: 			dec edi;
rlm@1: 			jnz label0;
rlm@1: 			mov src0, eax;
rlm@1: 			mov src1, ebx;
rlm@1: 			emms;
rlm@1: 		}
rlm@1: #endif
rlm@1: 		src0 += 2;
rlm@1: 		src1 += 2;
rlm@1: 	}
rlm@1: }
rlm@1: 
rlm@1: #endif
rlm@1: 
rlm@1: void MotionBlurIB(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	if (frm1 == NULL)
rlm@1: 	{
rlm@1: 		Init();
rlm@1: 	}
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: 	if (cpu_mmx)
rlm@1: 	{
rlm@1: 		MotionBlurIB_MMX(srcPtr, srcPitch, width, height);
rlm@1: 		return;
rlm@1: 	}
rlm@1: #endif
rlm@1: 
rlm@1: 	u16 colorMask = ~RGB_LOW_BITS_MASK;
rlm@1: 
rlm@1: 	u16 *src0 = (u16 *)srcPtr;
rlm@1: 	u16 *src1 = (u16 *)frm1;
rlm@1: 
rlm@1: 	int sPitch = srcPitch >> 1;
rlm@1: 
rlm@1: 	int pos = 0;
rlm@1: 	for (int j = 0; j < height; j++)
rlm@1: 		for (int i = 0; i < sPitch; i++)
rlm@1: 		{
rlm@1: 			u16 color = src0[pos];
rlm@1: 			src0[pos] =
rlm@1: 			    (((color & colorMask) >> 1) + ((src1[pos] & colorMask) >> 1));
rlm@1: 			src1[pos] = color;
rlm@1: 			pos++;
rlm@1: 		}
rlm@1: }
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: static void MotionBlurIB32_MMX(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	u32 *src0 = (u32 *)srcPtr;
rlm@1: 	u32 *src1 = (u32 *)frm1;
rlm@1: 
rlm@1: 	int count = width >> 1;
rlm@1: 
rlm@1: 	for (int i = 0; i < height; i++)
rlm@1: 	{
rlm@1: #ifdef __GNUC__
rlm@1: 		asm volatile (
rlm@1: 		    "push %2\n"
rlm@1: 		    "movq 0(%3), %%mm7\n"       // colorMask
rlm@1: 		    "0:\n"
rlm@1: 		    "movq 0(%0), %%mm0\n"       // src0
rlm@1: 		    "movq 0(%1), %%mm1\n"       // src1
rlm@1: 		    "movq %%mm0, 0(%1)\n"       // src1 = src0
rlm@1: 		    "pand %%mm7, %%mm0\n"       // color & colorMask
rlm@1: 		    "pand %%mm7, %%mm1\n"       // src1 & colorMask
rlm@1: 		    "psrld $1, %%mm0\n"       // (color & colorMask) >> 1 (E)
rlm@1: 		    "psrld $1, %%mm1\n"       // (src & colorMask) >> 1 (F)
rlm@1: 		    "paddd %%mm1, %%mm0\n"       // E+F
rlm@1: 
rlm@1: 		    "movq %%mm0, 0(%0)\n"       // src0 = res
rlm@1: 
rlm@1: 		    "addl $8, %0\n"
rlm@1: 		    "addl $8, %1\n"
rlm@1: 
rlm@1: 		    "decl %2\n"
rlm@1: 		    "jnz 0b\n"
rlm@1: 		    "pop %2\n"
rlm@1: 		    "emms\n"
rlm@1: 			: "+r" (src0), "+r" (src1)
rlm@1: 			: "r" (count), "r" (qRGB_COLOR_MASK)
rlm@1: 		    );
rlm@1: #else
rlm@1: 		__asm {
rlm@1: 			movq mm7, qword ptr [qRGB_COLOR_MASK];
rlm@1: 			mov	 eax, src0;
rlm@1: 			mov	 ebx, src1;
rlm@1: 			mov	 edi, count;
rlm@1: label0:
rlm@1: 			movq mm0, qword ptr [eax]; // src0
rlm@1: 			movq	   mm1, qword ptr [ebx]; // src1
rlm@1: 			movq qword ptr [ebx], mm0; // src1 = src0
rlm@1: 			pand	   mm0, mm7; // color & colorMask
rlm@1: 			pand	   mm1, mm7; // src1 & colorMask
rlm@1: 			psrld	   mm0, 1; // (color & colorMask) >> 1 (E)
rlm@1: 			psrld	   mm1, 1; // (src & colorMask) >> 1 (F)
rlm@1: 			paddd	   mm0, mm1; // E+F
rlm@1: 
rlm@1: 			movq qword ptr [eax], mm0; // src0 = res
rlm@1: 
rlm@1: 			add eax, 8;
rlm@1: 			add ebx, 8;
rlm@1: 
rlm@1: 			dec edi;
rlm@1: 			jnz label0;
rlm@1: 			mov src0, eax;
rlm@1: 			mov src1, ebx;
rlm@1: 			emms;
rlm@1: 		}
rlm@1: #endif
rlm@1: 		src0++;
rlm@1: 		src1++;
rlm@1: 	}
rlm@1: }
rlm@1: 
rlm@1: #endif
rlm@1: 
rlm@1: void MotionBlurIB32(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	if (frm1 == NULL)
rlm@1: 	{
rlm@1: 		Init();
rlm@1: 	}
rlm@1: 
rlm@1: #ifdef MMX
rlm@1: 	if (cpu_mmx)
rlm@1: 	{
rlm@1: 		MotionBlurIB32_MMX(srcPtr, srcPitch, width, height);
rlm@1: 		return;
rlm@1: 	}
rlm@1: #endif
rlm@1: 
rlm@1: 	u32 *src0 = (u32 *)srcPtr;
rlm@1: 	u32 *src1 = (u32 *)frm1;
rlm@1: 
rlm@1: 	u32 colorMask = 0xfefefe;
rlm@1: 
rlm@1: 	int sPitch = srcPitch >> 2;
rlm@1: 	int pos	   = 0;
rlm@1: 
rlm@1: 	for (int j = 0; j < height; j++)
rlm@1: 		for (int i = 0; i < sPitch; i++)
rlm@1: 		{
rlm@1: 			u32 color = src0[pos];
rlm@1: 			src0[pos] = (((color & colorMask) >> 1) +
rlm@1: 			             ((src1[pos] & colorMask) >> 1));
rlm@1: 			src1[pos] = color;
rlm@1: 			pos++;
rlm@1: 		}
rlm@1: }
rlm@1: 
rlm@1: static int count = 0;
rlm@1: 
rlm@1: void InterlaceIB(u8 *srcPtr, u32 srcPitch, int width, int height)
rlm@1: {
rlm@1: 	if (frm1 == NULL)
rlm@1: 	{
rlm@1: 		Init();
rlm@1: 	}
rlm@1: 
rlm@1: 	u16 colorMask = ~RGB_LOW_BITS_MASK;
rlm@1: 
rlm@1: 	u16 *src0 = (u16 *)srcPtr;
rlm@1: 	u16 *src1 = (u16 *)frm1;
rlm@1: 
rlm@1: 	int sPitch = srcPitch >> 1;
rlm@1: 
rlm@1: 	int pos = 0;
rlm@1: 	for (int j = 0; j < height; j++)
rlm@1: 	{
rlm@1: 		bool render = count ? (j & 1) != 0 : (j & 1) == 0;
rlm@1: 		if (render)
rlm@1: 		{
rlm@1: 			for (int i = 0; i < sPitch; i++)
rlm@1: 			{
rlm@1: 				u16 color = src0[pos];
rlm@1: 				src0[pos] =
rlm@1: 				    (((color & colorMask) >> 1) + ((((src1[pos] & colorMask) >> 1) & colorMask) >> 1));
rlm@1: 				src1[pos] = color;
rlm@1: 				pos++;
rlm@1: 			}
rlm@1: 		}
rlm@1: 		else
rlm@1: 		{
rlm@1: 			for (int i = 0; i < sPitch; i++)
rlm@1: 			{
rlm@1: 				u16 color = src0[pos];
rlm@1: 				src0[pos] =
rlm@1: 				    (((((color & colorMask) >> 1) & colorMask) >> 1) + ((src1[pos] & colorMask) >> 1));
rlm@1: 				src1[pos] = color;
rlm@1: 				pos++;
rlm@1: 			}
rlm@1: 		}
rlm@1: 	}
rlm@1: 	count = count ^ 1;
rlm@1: }
rlm@1: