/*
	2004.02.01
		first released source code for IOMP
*/
// Generic alpha renderers for all YUV modes and RGB depths.
// These are "reference implementations", should be optimized later (MMX, etc)
// Templating Code from Michael Niedermayer (michaelni@gmx.at) is under GPL

//#define FAST_OSD
//#define FAST_OSD_TABLE

#include "config.h"
#include "osd.h"
#include "mp_msg.h"
//#define ENABLE_PROFILE
#include "../my_profile.h"
#include <inttypes.h>
#include "../cpudetect.h"
#include "../mangle.h"

extern int verbose; // defined in mplayer.c

#ifdef ARCH_X86
#define CAN_COMPILE_X86_ASM
#endif

#ifdef CAN_COMPILE_X86_ASM
static const uint64_t bFF  __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL;
static const unsigned long long mask24lh  __attribute__((aligned(8))) = 0xFFFF000000000000ULL;
static const unsigned long long mask24hl  __attribute__((aligned(8))) = 0x0000FFFFFFFFFFFFULL;
#endif

//Note: we have C, X86-nommx, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
//Plain C versions
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT)
#define COMPILE_C
#endif

#ifdef CAN_COMPILE_X86_ASM

#if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_MMX
#endif

#if defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)
#define COMPILE_MMX2
#endif

#if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_3DNOW
#endif
#endif //CAN_COMPILE_X86_ASM

#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef ARCH_X86

#ifdef COMPILE_C
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef ARCH_X86
#define RENAME(a) a ## _C
#include "osd_template.c"
#endif

#ifdef CAN_COMPILE_X86_ASM

//X86 noMMX versions
#ifdef COMPILE_C
#undef RENAME
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _X86
#include "osd_template.c"
#endif

//MMX versions
#ifdef COMPILE_MMX
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _MMX
#include "osd_template.c"
#endif

//MMX2 versions
#ifdef COMPILE_MMX2
#undef RENAME
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _MMX2
#include "osd_template.c"
#endif

//3DNOW versions
#ifdef COMPILE_3DNOW
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _3DNow
#include "osd_template.c"
#endif

#endif //CAN_COMPILE_X86_ASM

void vo_draw_alpha_yv12(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
#ifdef RUNTIME_CPUDETECT
#ifdef CAN_COMPILE_X86_ASM
	// ordered per speed fasterst first
	if(gCpuCaps.hasMMX2)
		vo_draw_alpha_yv12_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.has3DNow)
		vo_draw_alpha_yv12_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.hasMMX)
		vo_draw_alpha_yv12_MMX(w, h, src, srca, srcstride, dstbase, dststride);
	else
		vo_draw_alpha_yv12_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_yv12_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
		vo_draw_alpha_yv12_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_3DNOW)
		vo_draw_alpha_yv12_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_MMX)
		vo_draw_alpha_yv12_MMX(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (ARCH_X86)
		vo_draw_alpha_yv12_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_yv12_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#endif //!RUNTIME_CPUDETECT
}

void vo_draw_alpha_yuy2(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
#ifdef RUNTIME_CPUDETECT
#ifdef CAN_COMPILE_X86_ASM
	// ordered per speed fasterst first
	if(gCpuCaps.hasMMX2)
		vo_draw_alpha_yuy2_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.has3DNow)
		vo_draw_alpha_yuy2_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.hasMMX)
		vo_draw_alpha_yuy2_MMX(w, h, src, srca, srcstride, dstbase, dststride);
	else
		vo_draw_alpha_yuy2_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_yuy2_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
		vo_draw_alpha_yuy2_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_3DNOW)
		vo_draw_alpha_yuy2_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_MMX)
		vo_draw_alpha_yuy2_MMX(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (ARCH_X86)
		vo_draw_alpha_yuy2_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_yuy2_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#endif //!RUNTIME_CPUDETECT
}

void vo_draw_alpha_rgb24(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
#ifdef RUNTIME_CPUDETECT
#ifdef CAN_COMPILE_X86_ASM
	// ordered per speed fasterst first
	if(gCpuCaps.hasMMX2)
		vo_draw_alpha_rgb24_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.has3DNow)
		vo_draw_alpha_rgb24_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.hasMMX)
		vo_draw_alpha_rgb24_MMX(w, h, src, srca, srcstride, dstbase, dststride);
	else
		vo_draw_alpha_rgb24_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_rgb24_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
		vo_draw_alpha_rgb24_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_3DNOW)
		vo_draw_alpha_rgb24_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_MMX)
		vo_draw_alpha_rgb24_MMX(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (ARCH_X86)
		vo_draw_alpha_rgb24_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_rgb24_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#endif //!RUNTIME_CPUDETECT
}

void vo_draw_alpha_rgb32(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
#ifdef RUNTIME_CPUDETECT
#ifdef CAN_COMPILE_X86_ASM
	// ordered per speed fasterst first
	if(gCpuCaps.hasMMX2)
		vo_draw_alpha_rgb32_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.has3DNow)
		vo_draw_alpha_rgb32_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
	else if(gCpuCaps.hasMMX)
		vo_draw_alpha_rgb32_MMX(w, h, src, srca, srcstride, dstbase, dststride);
	else
		vo_draw_alpha_rgb32_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_rgb32_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
		vo_draw_alpha_rgb32_MMX2(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_3DNOW)
		vo_draw_alpha_rgb32_3DNow(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (HAVE_MMX)
		vo_draw_alpha_rgb32_MMX(w, h, src, srca, srcstride, dstbase, dststride);
#elif defined (ARCH_X86)
		vo_draw_alpha_rgb32_X86(w, h, src, srca, srcstride, dstbase, dststride);
#else
		vo_draw_alpha_rgb32_C(w, h, src, srca, srcstride, dstbase, dststride);
#endif
#endif //!RUNTIME_CPUDETECT
}

#ifdef FAST_OSD_TABLE
static unsigned short fast_osd_15bpp_table[256];
static unsigned short fast_osd_16bpp_table[256];
#endif

void vo_draw_alpha_init(){
#ifdef FAST_OSD_TABLE
    int i;
    for(i=0;i<256;i++){
        fast_osd_15bpp_table[i]=((i>>3)<<10)|((i>>3)<<5)|(i>>3);
        fast_osd_16bpp_table[i]=((i>>3)<<11)|((i>>2)<<5)|(i>>3);
    }
#endif
//FIXME the optimized stuff is a lie for 15/16bpp as they arent optimized yet
	if(verbose)
	{
#ifdef RUNTIME_CPUDETECT
#ifdef CAN_COMPILE_X86_ASM
		// ordered per speed fasterst first
		if(gCpuCaps.hasMMX2)
			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay\n");
		else if(gCpuCaps.has3DNow)
			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit 3DNow) Optimized OnScreenDisplay\n");
		else if(gCpuCaps.hasMMX)
			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX Optimized OnScreenDisplay\n");
		else
			mp_msg(MSGT_OSD,MSGL_INFO,"Using X86 Optimized OnScreenDisplay\n");
#else
			mp_msg(MSGT_OSD,MSGL_INFO,"Using Unoptimized OnScreenDisplay\n");
#endif
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay\n");
#elif defined (HAVE_3DNOW)
			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit 3DNow) Optimized OnScreenDisplay\n");
#elif defined (HAVE_MMX)
			mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX Optimized OnScreenDisplay\n");
#elif defined (ARCH_X86)
			mp_msg(MSGT_OSD,MSGL_INFO,"Using X86 Optimized OnScreenDisplay\n");
#else
			mp_msg(MSGT_OSD,MSGL_INFO,"Using Unoptimized OnScreenDisplay\n");
#endif
#endif //!RUNTIME_CPUDETECT
	}
}

void vo_draw_alpha_rgb15(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
    int y;
    for(y=0;y<h;y++){
        register unsigned short *dst = (unsigned short*) dstbase;
        register int x;
        for(x=0;x<w;x++){
            if(srca[x]){
#ifdef FAST_OSD
#ifdef FAST_OSD_TABLE
                dst[x]=fast_osd_15bpp_table[src[x]];
#else
		register unsigned int a=src[x]>>3;
                dst[x]=(a<<10)|(a<<5)|a;
#endif
#else
                unsigned char r=dst[x]&0x1F;
                unsigned char g=(dst[x]>>5)&0x1F;
                unsigned char b=(dst[x]>>10)&0x1F;
                r=(((r*srca[x])>>5)+src[x])>>3;
                g=(((g*srca[x])>>5)+src[x])>>3;
                b=(((b*srca[x])>>5)+src[x])>>3;
                dst[x]=(b<<10)|(g<<5)|r;
#endif
            }
        }
        src+=srcstride;
        srca+=srcstride;
        dstbase+=dststride;
    }
    return;
}

unsigned char osd_color_r = 0;
unsigned char osd_color_g = 0;
unsigned char osd_color_b = 0;
static unsigned char t3=0,t4,BR,BG,BB,DR,DG,DB;
#define rgb24torgb16(r,g,b) \
	((((short)b & 0x00f8) >> 3) | \
	(((short)g & 0x00fc) << 3) | \
	(((short)r & 0x00f8) << 8))
int start_c_x;
int end_c_x;
int start_c_y;
int end_c_y;
int col_color;

void vo_draw_alpha_rgb16(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){
    int y;
	//printf("draw_alpha %d:%d\n",start_c_x,start_c_y);
	if (avamax_var->my_spu_draw) {
    for(y=0;y<h;y++){
        register unsigned short *dst = (unsigned short*) dstbase;
        register int x;
        if ((y > start_c_y) && (y < end_c_y)) {
            unsigned char BC,GC,RC;
            BC = (col_color >> 8) & 0xf0;
            GC = (col_color >> 4) & 0xf;
            RC = (col_color >> 0) & 0xf0;
            for (x = 0;((x < end_c_x) && (x < w)); x++) {
                if(srca[x]){
                    unsigned char b=dst[x]&0x1F;
                    unsigned char g=(dst[x]>>5)&0x3F;
                    unsigned char r=(dst[x]>>11)&0x1F;
                    b=(((r*srca[x])>>5)+(src[x]*BC)>>8)>>3;
                    g=(((g*srca[x])>>6)+(src[x]*GC)>>8)>>2;
                    r=(((b*srca[x])>>5)+(src[x]*RC)>>8)>>3;
                    dst[x]=(r<<11)|(g<<5)|b;
                }
            }
            for (;x < w; x++) {
                if(srca[x]){
                    unsigned char r=dst[x]&0x1F;
                    unsigned char g=(dst[x]>>5)&0x3F;
                    unsigned char b=(dst[x]>>11)&0x1F;
                    r=(((r*srca[x])>>5)+src[x])>>3;
                    g=(((g*srca[x])>>6)+src[x])>>2;
                    b=(((b*srca[x])>>5)+src[x])>>3;
                    dst[x]=(b<<11)|(g<<5)|r;
                }
            }
        } else {
            for(x=0;x<w;x++){
                if(srca[x]){
                    unsigned char r=dst[x]&0x1F;
                    unsigned char g=(dst[x]>>5)&0x3F;
                    unsigned char b=(dst[x]>>11)&0x1F;
                    r=(((r*srca[x])>>5)+src[x])>>3;
                    g=(((g*srca[x])>>6)+src[x])>>2;
                    b=(((b*srca[x])>>5)+src[x])>>3;
                    dst[x]=(b<<11)|(g<<5)|r;
                }
            }
        }
        src+=srcstride;
        srca+=srcstride;
        dstbase+=dststride;
    }
	} else { // spu_draw
		if (osd_color_g) {
			for(y=0;y<h;y++){
            	register unsigned short *dst = (unsigned short*) dstbase;
            	register int x;
            	for(x=0;x<w;x++){
                	if(srca[x]){
                   		unsigned char r=dst[x]&0x1F;
                    	unsigned char g=(dst[x]>>5)&0x3F;
                    	unsigned char b=(dst[x]>>11)&0x1F;
                    	r=(((r*srca[x])>>5)+0x0f)>>3;
                    	g=(((g*srca[x])>>6)+src[x])>>2;
                    	b=(((b*srca[x])>>5)+0x0f)>>3;
                    	dst[x]=(b<<11)|(g<<5)|r;
                	}
            	}
            	src+=srcstride;
           	 	srca+=srcstride;
            	dstbase+=dststride;
			}
		} else {
    	for(y=0;y<h;y++){
        	register unsigned short *dst = (unsigned short*) dstbase;
        	register int x;
            for(x=0;x<w;x++){
                if(srca[x]){
                    unsigned char r=dst[x]&0x1F;
                    unsigned char g=(dst[x]>>5)&0x3F;
                    unsigned char b=(dst[x]>>11)&0x1F;
                    r=(((r*srca[x])>>5)+src[x])>>3;
                    g=(((g*srca[x])>>6)+src[x])>>2;
                    b=(((b*srca[x])>>5)+src[x])>>3;
					// Green
                    //r=(((r*srca[x])>>5)+0x0f)>>3;
                    //g=(((g*srca[x])>>6)+src[x])>>2;
                    //b=(((b*srca[x])>>5)+0x0f)>>3;
                    dst[x]=(b<<11)|(g<<5)|r;
                }
            }
        	src+=srcstride;
        	srca+=srcstride;
        	dstbase+=dststride;
		}
		}
	}
    return;
}