incidentalstoat/blankart
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

More optimized code path for different pow2 sizes

Browse source
This commit is contained in:
NepDisk 2025-10-19 16:54:06 -04:00
parent 3fa76c409c
commit fbf10dbf9e
1 changed files with 100 additions and 89 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

189
src/r_draw_column.cpp
View file

@ -107,7 +107,6 @@ template<DrawColumnType Type>
static void R_DrawColumnTemplate(drawcolumndata_t *dc)
{
INT32 count;
UINT8 *dest;
const INT32 vidheight = vid.height;
// leban 1/17/99:
@ -220,13 +219,22 @@ static void R_DrawColumnTemplate(drawcolumndata_t *dc)
}
else
{
fixed_t fracstep;
fixed_t frac;
INT32 heightmask;
INT32 npow2min;
INT32 npow2max;
// Inner loop that does the actual texture mapping,
// e.g. a DDA-lile scaling.
// This is as fast as it gets. (Yeah, right!!! -- killough)
//
// killough 2/1/98: more performance tuning
intptr_t frac;
// Looks familiar.
const intptr_t fracstep = dc->iscale;
const intptr_t heightmask = dc->sourcelength-1; // CPhipps - specify type
constexpr INT32 npow2min = -1;
const INT32 npow2max = dc->sourcelength;
// Framebuffer destination address.
// SoM: MAGIC
UINT8 * restrict dest;
if constexpr (Type & DrawColumnType::DC_DIRECT)
dest = R_Address(dc->x, dc->yl);
@ -248,109 +256,112 @@ static void R_DrawColumnTemplate(drawcolumndata_t *dc)
count++;
// Determine scaling, which is the only mapping to be done.
fracstep = dc->iscale;
//frac = dc_texturemid + (dc_yl - centery)*fracstep;
frac = dc->texturemid + FixedMul((dc->yl << FRACBITS) - centeryfrac, fracstep);
frac = (dc->texturemid + FixedMul((dc->yl << FRACBITS) - centeryfrac, fracstep));
// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
// This is as fast as it gets.
heightmask = dc->sourcelength-1;
npow2min = -1;
npow2max = dc->sourcelength;
if (heightmask == -1)
switch (heightmask)
{
if (frac < 0)
// adjust in case we underread
frac += fracstep;
// texture has no height, so just go
while (--count > 0)
case 255:
case 127:
{
*dest = R_DrawColumnPixel<Type>(dc, dest, frac>>FRACBITS);
dest += stride;
frac += fracstep;
}
}
else if (dc->sourcelength & heightmask) // not a power of 2 -- killough
{
heightmask = dc->texheight << FRACBITS;
if (frac < 0)
{
while ((frac += heightmask) < 0)
while (count--)
{
;
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
dest += stride;
frac += fracstep;
}
}
else
break;
case npow2min:
{
while (frac >= heightmask)
if (frac < 0)
// adjust in case we underread
frac += fracstep;
// texture has no height, so just go
while (--count >= 0)
{
frac -= heightmask;
*dest = R_DrawColumnPixel<Type>(dc, dest, frac>>FRACBITS);
dest += stride;
frac += fracstep;
}
}
do
break;
default:
{
// Re-map color indices from wall texture column
// using a lighting/special effects LUT.
// heightmask is the Tutti-Frutti fix
// -1 is the lower clamp bound because column posts have a "safe" byte before the real data
// and a few bytes after as well
//*dest = R_DrawColumnPixel<Type>(dc, dest, std::clamp(frac >> FRACBITS, npow2min, npow2max));
if (!(dc->sourcelength & heightmask)) // power of 2 -- killough
{
// jartha: faster on my AMD FX-6300 CPU.
// Faster than ternaries, faster than std::min/std::max. Don't ask me why.
// I tested by viewing a non-PO2 texture from a consistent distance so it covered the entire screen.
// The framerate difference was about 50 frames at 640x400.
INT32 n = frac >> FRACBITS;
if (n < npow2min)
n = npow2min;
if (n > npow2max)
n = npow2max;
*dest = R_DrawColumnPixel<Type>(dc, dest, n);
}
while ((count -= 2) >= 0) // texture height is a power of 2 -- killough
{
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
dest += stride;
dest += stride;
frac += fracstep;
// Avoid overflow.
if (fracstep > 0x7FFFFFFF - frac)
{
frac += fracstep - heightmask;
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
dest += stride;
frac += fracstep;
}
if (count & 1)
{
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
}
}
else
{
frac += fracstep;
}
const intptr_t fixed_heightmask = dc->texheight << FRACBITS;
while (frac >= heightmask)
{
frac -= heightmask;
if (frac < 0)
{
while ((frac += fixed_heightmask) < 0)
{
;
}
}
else
{
while (frac >= fixed_heightmask)
{
frac -= fixed_heightmask;
}
}
do
{
// Re-map color indices from wall texture column
// using a lighting/special effects LUT.
// heightmask is the Tutti-Frutti fix -- killough
// -1 is the lower clamp bound because column posts have a "safe" byte before the real data
// and a few bytes after as well
*dest = R_DrawColumnPixel<Type>(dc, dest, CLAMP((frac >> FRACBITS), npow2min, npow2max));
dest += stride;
#if __SIZEOF_POINTER__ < 8 // 64-bit systems have large enough numbers for this to be a non-issue
// Avoid overflow.
if (fracstep > 0x7FFFFFFF - frac)
{
frac += fracstep - fixed_heightmask;
}
else
#endif
{
frac += fracstep;
}
while (frac >= fixed_heightmask)
{
frac -= fixed_heightmask;
}
}
while (--count);
}
}
while (--count);
}
else
{
while ((count -= 2) >= 0) // texture height is a power of 2
{
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
dest += stride;
frac += fracstep;
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
dest += stride;
frac += fracstep;
}
if (count & 1)
{
*dest = R_DrawColumnPixel<Type>(dc, dest, (frac>>FRACBITS) & heightmask);
}
break;
}
}
}