Retro/pcem
Watch
1
0
Fork
You've already forked pcem
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Add fast path in Banshee overlay render; if overlay is unfiltered and

Browse source 
unscaled then just write straight to destination buffer.
This commit is contained in:
SarahW 2020-09-28 20:02:38 +01:00
commit f6406c77e5
1 changed files with 77 additions and 68 deletions
Download patch file Download diff file Expand all files Collapse all files

147
src/vid_voodoo_banshee.c
View file

@ -1535,7 +1535,7 @@ void banshee_hwcursor_draw(svga_t *svga, int displine)
int g = (data >> 5) & 0x3f; \ int g = (data >> 5) & 0x3f; \
int b = data >> 11; \ int b = data >> 11; \
\ \
banshee->overlay_buffer[buf][wp++] = (r << 3) | (g << 10) | (b << 19); \ buf[wp++] = (r << 3) | (g << 10) | (b << 19); \
src += 2; \ src += 2; \
} \ } \
} while (0) } while (0)
@ -1553,7 +1553,7 @@ void banshee_hwcursor_draw(svga_t *svga, int displine)
int g = (data >> 5) & 0x3f; \ int g = (data >> 5) & 0x3f; \
int b = data >> 11; \ int b = data >> 11; \
\ \
banshee->overlay_buffer[buf][wp++] = (r << 3) | (g << 10) | (b << 19); \ buf[wp++] = (r << 3) | (g << 10) | (b << 19); \
} \ } \
} while (0) } while (0)
@ -1586,7 +1586,7 @@ void banshee_hwcursor_draw(svga_t *svga, int displine)
CLAMP(g); \ CLAMP(g); \
b = y1 + dB; \ b = y1 + dB; \
CLAMP(b); \ CLAMP(b); \
banshee->overlay_buffer[buf][wp++] = r | (g << 8) | (b << 16); \ buf[wp++] = r | (g << 8) | (b << 16); \
\ \
r = y2 + dR; \ r = y2 + dR; \
CLAMP(r); \ CLAMP(r); \
@ -1594,7 +1594,7 @@ void banshee_hwcursor_draw(svga_t *svga, int displine)
CLAMP(g); \ CLAMP(g); \
b = y2 + dB; \ b = y2 + dB; \
CLAMP(b); \ CLAMP(b); \
banshee->overlay_buffer[buf][wp++] = r | (g << 8) | (b << 16); \ buf[wp++] = r | (g << 8) | (b << 16); \
} \ } \
} while (0) } while (0)
@ -1627,7 +1627,7 @@ void banshee_hwcursor_draw(svga_t *svga, int displine)
CLAMP(g); \ CLAMP(g); \
b = y1 + dB; \ b = y1 + dB; \
CLAMP(b); \ CLAMP(b); \
banshee->overlay_buffer[buf][wp++] = r | (g << 8) | (b << 16); \ buf[wp++] = r | (g << 8) | (b << 16); \
\ \
r = y2 + dR; \ r = y2 + dR; \
CLAMP(r); \ CLAMP(r); \
@ -1635,7 +1635,7 @@ void banshee_hwcursor_draw(svga_t *svga, int displine)
CLAMP(g); \ CLAMP(g); \
b = y2 + dB; \ b = y2 + dB; \
CLAMP(b); \ CLAMP(b); \
banshee->overlay_buffer[buf][wp++] = r | (g << 8) | (b << 16); \ buf[wp++] = r | (g << 8) | (b << 16); \
} \ } \
} while (0) } while (0)
@ -1691,79 +1691,88 @@ static void banshee_overlay_draw(svga_t *svga, int displine)
// fatal("overlay out of range!\n"); // fatal("overlay out of range!\n");
p = &((uint32_t *)buffer32->line[displine])[svga->overlay_latch.x + 32]; p = &((uint32_t *)buffer32->line[displine])[svga->overlay_latch.x + 32];
OVERLAY_SAMPLE(0); if ((banshee->vidProcCfg & VIDPROCCFG_FILTER_MODE_MASK) != VIDPROCCFG_FILTER_MODE_BILINEAR &&
!(banshee->vidProcCfg & VIDPROCCFG_H_SCALE_ENABLE))
switch (banshee->vidProcCfg & VIDPROCCFG_FILTER_MODE_MASK)
{ {
case VIDPROCCFG_FILTER_MODE_BILINEAR: /*No scaling or filtering required, just write straight to output buffer*/
src = &svga->vram[src_addr2 & svga->vram_mask]; OVERLAY_SAMPLE(p);
OVERLAY_SAMPLE(1); }
if (banshee->vidProcCfg & VIDPROCCFG_H_SCALE_ENABLE) else
{ {
for (x = 0; x < svga->overlay_latch.xsize; x++) OVERLAY_SAMPLE(banshee->overlay_buffer[0]);
{
unsigned int x_coeff = (src_x & 0xfffff) >> 4;
unsigned int coeffs[4] = {
((0x10000 - x_coeff) * (0x10000 - y_coeff)) >> 16,
( x_coeff * (0x10000 - y_coeff)) >> 16,
((0x10000 - x_coeff) * y_coeff) >> 16,
( x_coeff * y_coeff) >> 16
};
uint32_t samp0 = banshee->overlay_buffer[0][src_x >> 20];
uint32_t samp1 = banshee->overlay_buffer[0][(src_x >> 20) + 1];
uint32_t samp2 = banshee->overlay_buffer[1][src_x >> 20];
uint32_t samp3 = banshee->overlay_buffer[1][(src_x >> 20) + 1];
int r = (((samp0 >> 16) & 0xff) * coeffs[0] +
((samp1 >> 16) & 0xff) * coeffs[1] +
((samp2 >> 16) & 0xff) * coeffs[2] +
((samp3 >> 16) & 0xff) * coeffs[3]) >> 16;
int g = (((samp0 >> 8) & 0xff) * coeffs[0] +
((samp1 >> 8) & 0xff) * coeffs[1] +
((samp2 >> 8) & 0xff) * coeffs[2] +
((samp3 >> 8) & 0xff) * coeffs[3]) >> 16;
int b = ((samp0 & 0xff) * coeffs[0] +
(samp1 & 0xff) * coeffs[1] +
(samp2 & 0xff) * coeffs[2] +
(samp3 & 0xff) * coeffs[3]) >> 16;
p[x] = (r << 16) | (g << 8) | b;
src_x += voodoo->overlay.vidOverlayDudx; switch (banshee->vidProcCfg & VIDPROCCFG_FILTER_MODE_MASK)
}
}
else
{ {
for (x = 0; x < svga->overlay_latch.xsize; x++) case VIDPROCCFG_FILTER_MODE_BILINEAR:
src = &svga->vram[src_addr2 & svga->vram_mask];
OVERLAY_SAMPLE(banshee->overlay_buffer[1]);
if (banshee->vidProcCfg & VIDPROCCFG_H_SCALE_ENABLE)
{ {
uint32_t samp0 = banshee->overlay_buffer[0][src_x >> 20]; for (x = 0; x < svga->overlay_latch.xsize; x++)
uint32_t samp1 = banshee->overlay_buffer[1][src_x >> 20]; {
int r = (((samp0 >> 16) & 0xff) * (0x10000 - y_coeff) + unsigned int x_coeff = (src_x & 0xfffff) >> 4;
((samp1 >> 16) & 0xff) * y_coeff) >> 16; unsigned int coeffs[4] = {
int g = (((samp0 >> 8) & 0xff) * (0x10000 - y_coeff) + ((0x10000 - x_coeff) * (0x10000 - y_coeff)) >> 16,
((samp1 >> 8) & 0xff) * y_coeff) >> 16; ( x_coeff * (0x10000 - y_coeff)) >> 16,
int b = ((samp0 & 0xff) * (0x10000 - y_coeff) + ((0x10000 - x_coeff) * y_coeff) >> 16,
(samp1 & 0xff) * y_coeff) >> 16; ( x_coeff * y_coeff) >> 16
p[x] = (r << 16) | (g << 8) | b; };
uint32_t samp0 = banshee->overlay_buffer[0][src_x >> 20];
uint32_t samp1 = banshee->overlay_buffer[0][(src_x >> 20) + 1];
uint32_t samp2 = banshee->overlay_buffer[1][src_x >> 20];
uint32_t samp3 = banshee->overlay_buffer[1][(src_x >> 20) + 1];
int r = (((samp0 >> 16) & 0xff) * coeffs[0] +
((samp1 >> 16) & 0xff) * coeffs[1] +
((samp2 >> 16) & 0xff) * coeffs[2] +
((samp3 >> 16) & 0xff) * coeffs[3]) >> 16;
int g = (((samp0 >> 8) & 0xff) * coeffs[0] +
((samp1 >> 8) & 0xff) * coeffs[1] +
((samp2 >> 8) & 0xff) * coeffs[2] +
((samp3 >> 8) & 0xff) * coeffs[3]) >> 16;
int b = ((samp0 & 0xff) * coeffs[0] +
(samp1 & 0xff) * coeffs[1] +
(samp2 & 0xff) * coeffs[2] +
(samp3 & 0xff) * coeffs[3]) >> 16;
p[x] = (r << 16) | (g << 8) | b;
src_x += voodoo->overlay.vidOverlayDudx;
}
} }
} else
break; {
for (x = 0; x < svga->overlay_latch.xsize; x++)
{
uint32_t samp0 = banshee->overlay_buffer[0][src_x >> 20];
uint32_t samp1 = banshee->overlay_buffer[1][src_x >> 20];
int r = (((samp0 >> 16) & 0xff) * (0x10000 - y_coeff) +
((samp1 >> 16) & 0xff) * y_coeff) >> 16;
int g = (((samp0 >> 8) & 0xff) * (0x10000 - y_coeff) +
((samp1 >> 8) & 0xff) * y_coeff) >> 16;
int b = ((samp0 & 0xff) * (0x10000 - y_coeff) +
(samp1 & 0xff) * y_coeff) >> 16;
p[x] = (r << 16) | (g << 8) | b;
}
}
break;
case VIDPROCCFG_FILTER_MODE_POINT: case VIDPROCCFG_FILTER_MODE_POINT:
default: default:
if (banshee->vidProcCfg & VIDPROCCFG_H_SCALE_ENABLE) if (banshee->vidProcCfg & VIDPROCCFG_H_SCALE_ENABLE)
{
for (x = 0; x < svga->overlay_latch.xsize; x++)
{ {
p[x] = banshee->overlay_buffer[0][src_x >> 20]; for (x = 0; x < svga->overlay_latch.xsize; x++)
{
p[x] = banshee->overlay_buffer[0][src_x >> 20];
src_x += voodoo->overlay.vidOverlayDudx; src_x += voodoo->overlay.vidOverlayDudx;
}
} }
else
{
for (x = 0; x < svga->overlay_latch.xsize; x++)
p[x] = banshee->overlay_buffer[0][x];
}
break;
} }
else
{
for (x = 0; x < svga->overlay_latch.xsize; x++)
p[x] = banshee->overlay_buffer[0][x];
}
break;
} }
if (banshee->vidProcCfg & VIDPROCCFG_V_SCALE_ENABLE) if (banshee->vidProcCfg & VIDPROCCFG_V_SCALE_ENABLE)