joncampbell123/dosbox-x
1
0
Fork 0
mirror of https://github.com/joncampbell123/dosbox-x.git synced 2025-10-14 02:17:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

S3 YUV overlay: Streamline YUV processing, fix horizontal DDA

Browse Source
This commit is contained in:
Jonathan Campbell
2021-04-12 22:16:24 -07:00
parent d5c4cf865d
commit 5bc8e29442
3 changed files with 48 additions and 56 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
include/vga.h
View File

@@ -590,7 +590,7 @@ typedef struct {
// 1=Enable color or chroma keying for all modes other than KRGB-16
// (MM8190 Secondary Stream Control)
int16_t ssctl_dda_haccum; // [12: 0] DDA Horizontal Accumulator (signed 2's complement)
int16_t ssctl_dda_haccum; // [11: 0] DDA Horizontal Accumulator (signed 2's complement)
// Set to: 2*(W0-1) - (W1-1) [FIXME: Am I reading that right? Datasheet says 2 (W0-1) - (W1-1)
// Windows 3.1 DCI seems to compute instead: ((2*(W0-1) - (W1-1)) / 2)
// W0 = line length in pixels before scaling

96
src/hardware/vga_draw.cpp
View File

@@ -58,6 +58,13 @@ struct s3drawstream {
int32_t vaccum;
unsigned int currentline;
/* temporary line to process and filter a scanline.
* maximum scanline of an overlay is 1024. */
union tmpscanline {
uint8_t yuv[1024*3]; // Y U V packed (good for the CPU cache)
};
union tmpscanline tmpscan1; // rendering version (NTS: tmpscan2 is planned for vertical interpolation)
};
struct s3drawstream S3SSdraw = {0};
@@ -972,7 +979,7 @@ static inline bool S3_XGA_OverlayKeyMatch(const uint32_t match,const uint32_t ke
extern bool vga_8bit_dac;
void S3_XGA_RenderYUY2colorkey(uint32_t* temp2/*already adjusted to X coordinate in row*/) {
void S3_XGA_RenderYUY2colorkey(uint32_t* temp2/*already adjusted to X coordinate in row*/,unsigned char *srcyuv3,int count) {
uint32_t mask = (0xFFu << (7u - vga.s3.streams.ckctl_rgb_cc)) * 0x010101u;
// HACK: DOSBox/DOSBox-X VGA emulation, unless otherwise, maps the 6-bit RGB VGA palette to 8-bit by shifting over by 2.
@@ -980,63 +987,46 @@ void S3_XGA_RenderYUY2colorkey(uint32_t* temp2/*already adjusted to X coordinate
// Mask off the low 2 bits if not 8-bit VGA or the color key will never work.
if (!vga_8bit_dac) mask &= 0xFCFCFC;
const uint32_t key = (((uint32_t)vga.s3.streams.ckctl_b_lb) | ((uint32_t)vga.s3.streams.ckctl_g_lb << 8u) | ((uint32_t)vga.s3.streams.ckctl_r_lb << 16u)) & mask;
int count = S3SSdraw.endx - S3SSdraw.startx;
uint32_t scan = S3SSdraw.vmem_addr;
uint8_t Y,U,V;
const uint32_t key = (((uint32_t)vga.s3.streams.ckctl_b_lb) | ((uint32_t)vga.s3.streams.ckctl_g_lb << 8u) | ((uint32_t)vga.s3.streams.ckctl_r_lb << 16)) & mask;
int o = 0;
/* vmem_addr points at YUY2, interleaved Y U V samples as a series of [Y U Y V] blocks, 2 Ys, 1 U, 1 V */
while (count >= 2) {
U = S3StreamVRAMRead8(scan+1);
V = S3StreamVRAMRead8(scan+3);
if (S3_XGA_OverlayKeyMatch(temp2[0] & mask,key)) {
Y = S3StreamVRAMRead8(scan+0);
temp2[0] = YUVMPEG2RGB32(Y,U,V);
}
while (count-- > 0) {
if (S3_XGA_OverlayKeyMatch(temp2[o] & mask,key))
temp2[o] = YUVMPEG2RGB32(srcyuv3[0],srcyuv3[1],srcyuv3[2]);
if (S3_XGA_OverlayKeyMatch(temp2[1] & mask,key)) {
Y = S3StreamVRAMRead8(scan+2);
temp2[1] = YUVMPEG2RGB32(Y,U,V);
}
temp2 += 2;
count -= 2;
scan += 4;
}
if (count > 0 && S3_XGA_OverlayKeyMatch(temp2[0] & mask,key)) {
Y = S3StreamVRAMRead8(scan+0);
U = S3StreamVRAMRead8(scan+1);
V = S3StreamVRAMRead8(scan+3);
temp2[0] = YUVMPEG2RGB32(Y,U,V);
srcyuv3 += 3;
o++;
}
}
void S3_XGA_RenderYUY2(uint32_t* temp2/*already adjusted to X coordinate in row*/) {
int count = S3SSdraw.endx - S3SSdraw.startx;
uint32_t scan = S3SSdraw.vmem_addr;
uint8_t Y,U,V;
void S3_XGA_RenderYUY2(uint32_t* temp2/*already adjusted to X coordinate in row*/,unsigned char *srcyuv3,int count) {
int o = 0;
/* vmem_addr points at YUY2, interleaved Y U V samples as a series of [Y U Y V] blocks, 2 Ys, 1 U, 1 V */
while (count >= 2) {
Y = S3StreamVRAMRead8(scan+0);
U = S3StreamVRAMRead8(scan+1);
V = S3StreamVRAMRead8(scan+3);
temp2[0] = YUVMPEG2RGB32(Y,U,V);
Y = S3StreamVRAMRead8(scan+2);
temp2[1] = YUVMPEG2RGB32(Y,U,V);
temp2 += 2;
count -= 2;
scan += 4;
while (count-- > 0) {
temp2[o] = YUVMPEG2RGB32(srcyuv3[0],srcyuv3[1],srcyuv3[2]);
srcyuv3 += 3;
o++;
}
}
if (count > 0) {
Y = S3StreamVRAMRead8(scan+0);
U = S3StreamVRAMRead8(scan+1);
V = S3StreamVRAMRead8(scan+3);
temp2[0] = YUVMPEG2RGB32(Y,U,V);
void S3_XGA_YUY2HProc(unsigned char *dst3yuv,uint32_t vram,int count) {
/* nearest neighbor */
{
int32_t haccum = vga.s3.streams.ssctl_dda_haccum;
while (count-- > 0) {
dst3yuv[0] = S3StreamVRAMRead8(vram+0); // Y
dst3yuv[1] = S3StreamVRAMRead8((vram&(~3u))+1); // U
dst3yuv[2] = S3StreamVRAMRead8((vram&(~3u))+3); // V
dst3yuv += 3;
haccum += vga.s3.streams.ssctl_k1_hscale;
if (haccum >= 0) {
haccum -= vga.s3.streams.ssctl_k1_hscale;
haccum += vga.s3.streams.ssctl_k2_hscale;
vram += 2;
}
}
}
}
@@ -1044,10 +1034,12 @@ void S3_XGA_SecondaryStreamRender(uint32_t* temp2) {
if (S3SSdraw.draw) {
if (S3SSdraw.currentline >= S3SSdraw.starty && S3SSdraw.currentline < S3SSdraw.endy) {
// FIXME: This assumes YUY2 16-240 range (MPEG-style), check format code.
S3_XGA_YUY2HProc(S3SSdraw.tmpscan1.yuv,S3SSdraw.vmem_addr,S3SSdraw.endx - S3SSdraw.startx);
if (vga.s3.streams.blendctl_composemode == 5/*color key on primary stream, secondary overlay on primary*/)
S3_XGA_RenderYUY2colorkey(temp2+S3SSdraw.startx);
S3_XGA_RenderYUY2colorkey(temp2+S3SSdraw.startx,S3SSdraw.tmpscan1.yuv,S3SSdraw.endx - S3SSdraw.startx);
else
S3_XGA_RenderYUY2(temp2+S3SSdraw.startx);
S3_XGA_RenderYUY2(temp2+S3SSdraw.startx,S3SSdraw.tmpscan1.yuv,S3SSdraw.endx - S3SSdraw.startx);
/* it's not clear from the datasheet, but I think what the card is doing is a
* DDA to vertically scale the image, and K1/K2 are just terms to add/subtract

6
src/hardware/vga_xga.cpp
View File

@@ -1179,9 +1179,9 @@ void XGA_Write(Bitu port, Bitu val, Bitu len) {
break;
case 0x8190: // S3 Trio64V+ streams processor, Secondary Stream Control (MMIO only)
if (s3Card == S3_Trio64V || s3Card >= S3_ViRGE) {
vga.s3.streams.ssctl_dda_haccum = val & 0x1FFFu; // signed 13-bit value
if (vga.s3.streams.ssctl_dda_haccum & 0x1000u)
vga.s3.streams.ssctl_dda_haccum -= 0x2000u;
vga.s3.streams.ssctl_dda_haccum = val & 0xFFFu; // signed 12-bit value
if (vga.s3.streams.ssctl_dda_haccum & 0x0800u)
vga.s3.streams.ssctl_dda_haccum -= 0x1000u;
vga.s3.streams.ssctl_sdif = (val >> 24u) & 7u;
vga.s3.streams.ssctl_sfc = (val >> 28u) & 7u;
}