Hello all,
I don't have any RGB overlay, and YUV2 orvelays seem broken, so I set out
to improve YV12 support for xmame.
The patch in this mail does three things:
* It adds a -yv12 switch to force yv12 overlays.
* It **hugely** improves the aspect of yv12 rendering, and, for 16bits
games, the speed of yv12 rendering.
* it adds support for -scale 2 in yv12 rendering (only -scale 2. It has
the effect of rendering perfect colors. Other values of -scale are
useless and unsupported.)
Note however that yv12 functions are still only in x11_window.c. Hacking
blit.h is way too complicated for me, I won't touch that thing.
The patch is against pre8, but as x11_window.c was not modified by pre9,
it should work as well in the latest version.
Regards,
�ric Brunet
diff -ru xmame-0.61.1-pr8-vanilla/src/unix/video-drivers/x11_window.c
xmame-0.61.1-pr8/src/unix/video-drivers/x11_window.c
--- xmame-0.61.1-pr8-vanilla/src/unix/video-drivers/x11_window.c Thu Oct 17
16:56:55 2002
+++ xmame-0.61.1-pr8/src/unix/video-drivers/x11_window.c Thu Oct 31 10:33:48
+2002
@@ -32,8 +32,10 @@
#ifdef USE_XV
static void x11_window_update_16_to_YUY2 (struct mame_bitmap *bitmap);
static void x11_window_update_16_to_YV12 (struct mame_bitmap *bitmap);
+static void x11_window_update_16_to_YV12_perfect (struct mame_bitmap *bitmap);
static void x11_window_update_32_to_YUY2_direct (struct mame_bitmap *bitmap);
static void x11_window_update_32_to_YV12_direct (struct mame_bitmap *bitmap);
+static void x11_window_update_32_to_YV12_direct_perfect (struct mame_bitmap *bitmap);
static void x11_window_make_yuv_lookup();
#endif
static void x11_window_update_8_to_8bpp (struct mame_bitmap *bitmap);
@@ -62,6 +64,7 @@
static int xv_bpp=0;
static long xv_format=0;
static long xv_yuv=0;
+static long xv_yv12=0;
static int *xv_yuvlookup=NULL;
#define XV_YUY2 0x32595559
#define XV_YV12 0x32315659
@@ -122,6 +125,9 @@
{ "yuv", NULL, rc_bool, &xv_yuv,
"0", 0, 0, NULL,
"Force YUV mode (for video cards with broken RGB overlays)" },
+ { "yv12", NULL, rc_bool, &xv_yv12,
+ "0", 0, 0, NULL,
+ "Force YV12 mode (for video cards with broken RGB overlays)" },
#endif
{ "xsync", "xs", rc_bool, &use_xsync,
"1", 0, 0, NULL,
@@ -773,6 +779,7 @@
/* Create an XV MITSHM image. */
{
fprintf (stderr_file, "MIT-SHM & XV Extensions Available. trying to
use... ");
+ if (xv_yv12) xv_yuv=1;
XSetErrorHandler (test_mit_shm);
if(xv_yuv==0)
{
@@ -789,9 +796,9 @@
xv_bluemask=0xff;
xv_bpp=32;
xv_format=XV_YUY2;
- if(!(FindXvPort(display, xv_format, &xv_port)))
+ if(xv_yv12 || !(FindXvPort(display, xv_format, &xv_port)))
{
- fprintf(stderr,"\nYUY2 not available - trying YV12... ");
+ if (!xv_yv12) fprintf(stderr,"\nYUY2 not available - trying YV12...
+");
xv_format=XV_YV12;
if(!(FindXvPort(display, xv_format, &xv_port)))
{
@@ -932,7 +939,19 @@
break;
case XV_YV12:
ClearYV12(xvimage);
- x11_window_update_display_func = x11_window_update_32_to_YV12_direct;
+ if (widthscale == 1 && heightscale == 1)
+ x11_window_update_display_func
+ = x11_window_update_32_to_YV12_direct;
+ else if (widthscale ==2 && heightscale == 2)
+ x11_window_update_display_func
+ = x11_window_update_32_to_YV12_direct_perfect;
+ else {
+ fprintf(stderr_file, "\nScaling different from 1 or 2"
+ " is useless and unsupported\n");
+ return OSD_NOT_OK;
+ }
+
+
break;
}
}
@@ -954,7 +973,17 @@
break;
case XV_YV12:
ClearYV12(xvimage);
- x11_window_update_display_func = x11_window_update_16_to_YV12;
+ if (widthscale == 1 && heightscale == 1)
+ x11_window_update_display_func
+ = x11_window_update_16_to_YV12;
+ else if (widthscale ==2 && heightscale == 2)
+ x11_window_update_display_func
+ = x11_window_update_16_to_YV12_perfect;
+ else {
+ fprintf(stderr_file, "\nScaling different from 1 or 2"
+ " is useless and unsupported\n");
+ return OSD_NOT_OK;
+ }
break;
}
}
@@ -1393,6 +1422,15 @@
#define RMASK 0xff0000
#define GMASK 0x00ff00
#define BMASK 0x0000ff
+
+#define RGB2YUV(r,g,b,y,u,v) \
+ (y) = ( 9797*(r) + 19237*(g) + 3734*(b) ) >> 15;\
+ (u) = (18492*((b)-(y)) >> 15) + 128;\
+ (v) = (23372*((r)-(y)) >> 15) + 128;
+ /* (v) = (( 16385*(r) - 13721*(g) - 2664*(b) ) >> 15) + 128
+ (u) = (( -5528*(r) - 10856*(g) + 16385*(b) ) >> 15) + 128;\
+ */
+
static void x11_window_make_yuv_lookup()
{
int i,r,g,b,y,u,v,n;
@@ -1413,9 +1451,7 @@
g=(g&GMASK)>>8;
b=(b&BMASK);
- y = (( 9897*r + 19235*g + 3736*b ) >> 15) & 255;
- u = ((( -5537*r - 10878*g + 16384*b ) >> 15) + 128) & 255;
- v = ((( 16384*r - 13730*g -2664*b ) >> 15 ) + 128) & 255;
+ RGB2YUV(r,g,b,y,u,v);
/* Storing this data in YUYV order simplifies using the data for
YUY2, both with and without smoothing... */
@@ -1427,36 +1463,140 @@
/* Hacked into place, until I integrate YV12 support into the blit core... */
static void x11_window_update_16_to_YV12(struct mame_bitmap *bitmap)
{
- int _x,_y,r,g,b;
+ int _x,_y;
unsigned char *dest_y;
unsigned char *dest_u;
unsigned char *dest_v;
unsigned short *src;
- int *indirect=current_palette->lookup;
+ unsigned short *src2;
+ int u,v,y,u2,v2,y2,u3,v3,y3,u4,v4,y4; /* 12 */
+ int *indirect=current_palette->lookup; /* 34 */
+
+ for(_y=visual.min_y;_y<=visual.max_y;_y+=2)
+ {
+ src=bitmap->line[_y] ;
+ src+= visual.min_x;
+ src2=bitmap->line[_y+1];
+ src2+= visual.min_x;
- for(_y=visual.min_y;_y<=visual.max_y;++_y)
- {
- src=bitmap->line[_y];
- src+=visual.min_x;
dest_y=xvimage->data+xvimage->offsets[0]+(xvimage->width*(_y-visual.min_y));
dest_v=xvimage->data+xvimage->offsets[1]+((xvimage->width/2)*((_y-visual.min_y)/2));
dest_u=xvimage->data+xvimage->offsets[2]+((xvimage->width/2)*((_y-visual.min_y)/2));
- for(_x=visual.min_x;_x<=visual.max_x;++_x)
+ for(_x=visual.min_x;_x<=visual.max_x;_x+=2)
{
- r=g=b=*src++;
- if(indirect)
- r=g=b=indirect[r];
-
- r&=RMASK; r>>=16; /* red is << 8 */
- g&=GMASK; g>>=8; /* green is << 3 */
- b&=BMASK; b>>=0; /* blue is >> 3 */
+ if (indirect) {
+ v = xv_yuvlookup[*src++];
+ y = (v>>8) & 0xff;
+ u = (v>>16) & 0xff;
+ v = (v) & 0xff;
+
+ v2 = xv_yuvlookup[*src++];
+ y2 = (v2>>8) & 0xff;
+ u2 = (v2>>16) & 0xff;
+ v2 = (v2) & 0xff;
+
+ v3 = xv_yuvlookup[*src2++];
+ y3 = (v3>>8) & 0xff;
+ u3 = (v3>>16) & 0xff;
+ v3 = (v3) & 0xff;
+
+ v4 = xv_yuvlookup[*src2++];
+ y4 = (v4>>8) & 0xff;
+ u4 = (v4>>16) & 0xff;
+ v4 = (v4) & 0xff;
+ } else { /* Can this really happen ? */
+ int r,g,b;
+ b = *src++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y,u,v);
+
+ b = *src++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y2,u2,v2);
+
+ b = *src2++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y3,u3,v3);
+
+ b = *src2++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y4,u4,v4);
+ }
+
+ *dest_y = y;
+ *(dest_y++ + xvimage->width) = y3;
+ *dest_y = y2;
+ *(dest_y++ + xvimage->width) = y4;
+
+ *dest_u++ = (u+u2+u3+u4)/4;
+ *dest_v++ = (v+v2+v3+v4)/4;
+
+ /* I thought that the following would be better, but it is not
+ * the case. The color gets blurred
+ if (y || y2 || y3 || y4) {
+ *dest_u++ = (u*y+u2*y2+u3*y3+u4*y4)/(y+y2+y3+y4);
+ *dest_v++ = (v*y+v2*y2+v3*y3+v4*y4)/(y+y2+y3+y4);
+ } else {
+ *dest_u++ =128;
+ *dest_v++ =128;
+ }
+ */
+ }
+ }
+}
- *dest_y++ = (( 9897*r + 19235*g + 3736*b ) >> 15);
- if(((_x&1)==1) && ((_y&1)==1))
- {
- *dest_u++ = (( -5537*r - 10878*g + 16384*b ) >> 15) + 128;
- *dest_v++ = (( 16384*r - 13730*g -2664*b ) >> 15 ) + 128;
- }
+static void x11_window_update_16_to_YV12_perfect(struct mame_bitmap *bitmap)
+{ /* this one is used when scale==2 */
+ unsigned int _x,_y,r;
+ unsigned char *dest_y;
+ unsigned char *dest_u;
+ unsigned char *dest_v;
+ unsigned short *src;
+ unsigned short *src2;
+ int u,v,y;
+ int *indirect=current_palette->lookup;
+
+ for(_y=visual.min_y;_y<=visual.max_y;_y++)
+ {
+ src=bitmap->line[_y];
+ src += visual.min_x;
+ src2=bitmap->line[_y+1];
+ src2 += visual.min_x;
+
+ dest_y=xvimage->data+xvimage->offsets[0]+2*(xvimage->width*(_y-visual.min_y));
+ dest_v=xvimage->data+xvimage->offsets[1]+((xvimage->width/2)*(_y-visual.min_y));
+ dest_u=xvimage->data+xvimage->offsets[2]+((xvimage->width/2)*(_y-visual.min_y));
+ for(_x=visual.min_x;_x<=visual.max_x;_x++)
+ {
+ if (indirect) {
+ v= xv_yuvlookup[*src++];
+ y = (v>>8) & 0xff;
+ u = (v>>16) & 0xff;
+ v = (v) & 0xff;
+ } else { /* Can this really happen ? */
+ int r,g,b;
+ b = *src++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y,u,v);
+ }
+
+
+ *(dest_y+xvimage->width)=y;
+ *dest_y++=y;
+ *(dest_y+xvimage->width)=y;
+ *dest_y++=y;
+ *dest_u++ = u;
+ *dest_v++ = v;
}
}
}
@@ -1468,28 +1608,91 @@
unsigned char *dest_u;
unsigned char *dest_v;
unsigned int *src;
+ unsigned int *src2;
+ int u,v,y,u2,v2,y2,u3,v3,y3,u4,v4,y4; /* 12 */
+ /* 34 */
- for(_y=visual.min_y;_y<=visual.max_y;++_y)
+ for(_y=visual.min_y;_y<=visual.max_y;_y+=2)
{
src=bitmap->line[_y];
src+=visual.min_x;
+ src2=bitmap->line[_y+1];
+ src2 += visual.min_x;
+
dest_y=xvimage->data+xvimage->offsets[0]+(xvimage->width*(_y-visual.min_y));
dest_v=xvimage->data+xvimage->offsets[1]+((xvimage->width/2)*((_y-visual.min_y)/2));
dest_u=xvimage->data+xvimage->offsets[2]+((xvimage->width/2)*((_y-visual.min_y)/2));
- for(_x=visual.min_x;_x<=visual.max_x;++_x)
+ for(_x=visual.min_x;_x<=visual.max_x;_x+=2)
{
- r=g=b=*src++;
+ b = *src++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y,u,v);
+
+ b = *src++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y2,u2,v2);
+
+ b = *src2++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y3,u3,v3);
+
+ b = *src2++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y4,u4,v4);
+
+ *dest_y = y;
+ *(dest_y++ + xvimage->width) = y3;
+ *dest_y = y2;
+ *(dest_y++ + xvimage->width) = y4;
- r&=RMASK; r>>=16;
- g&=GMASK; g>>=8;
- b&=BMASK; b>>=0;
+ *dest_u++ = (u+u2+u3+u4)/4;
+ *dest_v++ = (v+v2+v3+v4)/4;
+ }
+ }
+}
- *dest_y++ = (( 9897*r + 19235*g + 3736*b ) >> 15);
- if(((_x&1)==1) && ((_y&1)==1))
- {
- *dest_u++ = (( -5537*r - 10878*g + 16384*b ) >> 15) + 128;
- *dest_v++ = (( 16384*r - 13730*g -2664*b ) >> 15 ) + 128;
- }
+static void x11_window_update_32_to_YV12_direct_perfect(struct mame_bitmap *bitmap)
+{ /* This one is used when scale == 2 */
+ int _x,_y,r,g,b;
+ unsigned char *dest_y;
+ unsigned char *dest_u;
+ unsigned char *dest_v;
+ unsigned int *src;
+ unsigned int *src2;
+ int u,v,y;
+
+ for(_y=visual.min_y;_y<=visual.max_y;_y++)
+ {
+ src = bitmap->line[_y];
+ src += visual.min_x;
+ src2 = bitmap->line[_y+1];
+ src2 += visual.min_x;
+
+ dest_y=xvimage->data+xvimage->offsets[0]+2*(xvimage->width*(_y-visual.min_y));
+ dest_v=xvimage->data+xvimage->offsets[1]+((xvimage->width/2)*(_y-visual.min_y));
+ dest_u=xvimage->data+xvimage->offsets[2]+((xvimage->width/2)*(_y-visual.min_y));
+ for(_x=visual.min_x;_x<=visual.max_x;_x++)
+ {
+ b = *src++;
+ r = (b>>16) & 0xFF;
+ g = (b>>8) & 0xFF;
+ b = (b) & 0xFF;
+ RGB2YUV(r,g,b,y,u,v);
+
+ *(dest_y+xvimage->width) = y;
+ *dest_y++ = y;
+ *(dest_y+xvimage->width) = y;
+ *dest_y++ = y;
+ *dest_u++ = u;
+ *dest_v++ = v;
}
}
}
