Skip to content

Commit

Permalink
- cleanup of m_fixed.h.
Browse files Browse the repository at this point in the history 
With fixed point math barely being used anywhere with the vast majority of calls in the software voxel drawer it makes sense to reduce the function interface to the minimum possible and move the shift value into the function call as an argument.
  • Loading branch information
@coelckers
coelckers committed Nov 22, 2020
1 parent a2b8ad7 commit 58d8ce1
Show file tree
Hide file tree
Showing 9 changed files with 58 additions and 79 deletions.
10 changes: 10 additions & 0 deletions src/common/utility/cmdlib.h
View file
Expand Up @@ -90,4 +90,14 @@ struct MD5Context;
void md5Update(FileReader& file, MD5Context& md5, unsigned len);
void uppercopy(char* to, const char* from);

inline void fillshort(void* buff, size_t count, uint16_t clear)
{
int16_t* b2 = (int16_t*)buff;
for (size_t i = 0; i < count; ++i)
{
b2[i] = clear;
}
}


#endif
42 changes: 3 additions & 39 deletions src/common/utility/m_fixed.h
View file
Expand Up @@ -6,48 +6,12 @@
#include "basics.h"


// Modern compilers are smart enough to do these multiplications intelligently.
__forceinline int32_t MulScale14(int32_t a, int32_t b) { return (int32_t)(((int64_t)a * b) >> 14); } // only used by R_DrawVoxel
__forceinline int32_t MulScale30(int32_t a, int32_t b) { return (int32_t)(((int64_t)a * b) >> 30); } // only used once in the node builder
__forceinline int32_t MulScale32(int32_t a, int32_t b) { return (int32_t)(((int64_t)a * b) >> 32); } // only used by R_DrawVoxel

__forceinline uint32_t UMulScale16(uint32_t a, uint32_t b) { return (uint32_t)(((uint64_t)a * b) >> 16); } // used for sky drawing

__forceinline int32_t DMulScale3(int32_t a, int32_t b, int32_t c, int32_t d) { return (int32_t)(((int64_t)a*b + (int64_t)c*d) >> 3); } // used for setting up slopes for Build maps
__forceinline int32_t DMulScale6(int32_t a, int32_t b, int32_t c, int32_t d) { return (int32_t)(((int64_t)a*b + (int64_t)c*d) >> 6); } // only used by R_DrawVoxel
__forceinline int32_t DMulScale10(int32_t a, int32_t b, int32_t c, int32_t d) { return (int32_t)(((int64_t)a*b + (int64_t)c*d) >> 10); } // only used by R_DrawVoxel
__forceinline int32_t DMulScale18(int32_t a, int32_t b, int32_t c, int32_t d) { return (int32_t)(((int64_t)a*b + (int64_t)c*d) >> 18); } // only used by R_DrawVoxel
__forceinline int32_t DMulScale32(int32_t a, int32_t b, int32_t c, int32_t d) { return (int32_t)(((int64_t)a*b + (int64_t)c*d) >> 32); } // used by R_PointOnSide.

// Sadly, for divisions this is not true but these are so infrequently used that the C versions are just fine, despite not being fully optimal.
__forceinline int32_t DivScale6(int32_t a, int32_t b) { return (int32_t)(((int64_t)a << 6) / b); } // only used by R_DrawVoxel
__forceinline int32_t DivScale21(int32_t a, int32_t b) { return (int32_t)(((int64_t)a << 21) / b); } // only used by R_DrawVoxel
__forceinline int32_t DivScale30(int32_t a, int32_t b) { return (int32_t)(((int64_t)a << 30) / b); } // only used once in the node builder

__forceinline void fillshort(void *buff, unsigned int count, uint16_t clear)
{
int16_t *b2 = (int16_t *)buff;
for (unsigned int i = 0; i != count; ++i)
{
b2[i] = clear;
}
}
__forceinline constexpr int32_t MulScale(int32_t a, int32_t b, int32_t shift) { return (int32_t)(((int64_t)a * b) >> shift); }
__forceinline constexpr int32_t DMulScale(int32_t a, int32_t b, int32_t c, int32_t d, int32_t shift) { return (int32_t)(((int64_t)a * b + (int64_t)c * d) >> shift); }
__forceinline constexpr int32_t DivScale(int32_t a, int32_t b, int shift) { return (int32_t)(((int64_t)a << shift) / b); }

#include "xs_Float.h"

inline int32_t FixedDiv (int32_t a, int32_t b)
{
if ((uint32_t)abs(a) >> (31-16) >= (uint32_t)abs (b))
return (a^b)<0 ? FIXED_MIN : FIXED_MAX;

return (int32_t)(((int64_t)a << 16) / b);
}

__forceinline constexpr int32_t FixedMul(int32_t a, int32_t b)
{
return (int32_t)(((int64_t)a * b) >> 16);
}

inline fixed_t FloatToFixed(double f)
{
return xs_Fix<16>::ToFix(f);
Expand Down
10 changes: 7 additions & 3 deletions src/playsim/p_acs.cpp
View file
Expand Up @@ -9050,15 +9050,19 @@ int DLevelScript::RunScript()
break;

case PCD_FIXEDMUL:
STACK(2) = FixedMul (STACK(2), STACK(1));
STACK(2) = MulScale(STACK(2), STACK(1), 16);
sp--;
break;

case PCD_FIXEDDIV:
STACK(2) = FixedDiv (STACK(2), STACK(1));
{
int a = STACK(2), b = STACK(1);
// Overflow check.
if ((uint32_t)abs(a) >> (31 - 16) >= (uint32_t)abs(b)) STACK(2) = (a ^ b) < 0 ? FIXED_MIN : FIXED_MAX;
else STACK(2) = DivScale(STACK(2), STACK(1), 16);
sp--;
break;

}
case PCD_SETGRAVITY:
Level->gravity = ACSToDouble(STACK(1));
sp--;
Expand Down
4 changes: 2 additions & 2 deletions src/playsim/p_maputl.cpp
View file
Expand Up @@ -1929,8 +1929,8 @@ int P_VanillaPointOnLineSide(double x, double y, const line_t* line)
auto dx = FloatToFixed(x - line->v1->fX());
auto dy = FloatToFixed(y - line->v1->fY());

auto left = FixedMul( int(delta.Y * 256) , dx );
auto right = FixedMul( dy , int(delta.X * 256) );
auto left = MulScale( int(delta.Y * 256) , dx, 16 );
auto right = MulScale( dy , int(delta.X * 256), 16 );

if (right < left)
return 0; // front side
Expand Down
8 changes: 4 additions & 4 deletions src/rendering/r_utility.h
View file
Expand Up @@ -91,17 +91,17 @@ const double r_Yaspect = 200.0; // Why did I make this a variable? It's never
//
//==========================================================================

inline int R_PointOnSide (fixed_t x, fixed_t y, const node_t *node)
inline constexpr int R_PointOnSide (fixed_t x, fixed_t y, const node_t *node)
{
return DMulScale32 (y-node->y, node->dx, node->x-x, node->dy) > 0;
return DMulScale (y-node->y, node->dx, node->x-x, node->dy, 32) > 0;
}
inline int R_PointOnSide(double x, double y, const node_t *node)
{
return DMulScale32(FLOAT2FIXED(y) - node->y, node->dx, node->x - FLOAT2FIXED(x), node->dy) > 0;
return DMulScale(FLOAT2FIXED(y) - node->y, node->dx, node->x - FLOAT2FIXED(x), node->dy, 32) > 0;
}
inline int R_PointOnSide(const DVector2 &pos, const node_t *node)
{
return DMulScale32(FLOAT2FIXED(pos.Y) - node->y, node->dx, node->x - FLOAT2FIXED(pos.X), node->dy) > 0;
return DMulScale(FLOAT2FIXED(pos.Y) - node->y, node->dx, node->x - FLOAT2FIXED(pos.X), node->dy, 32) > 0;
}

// Used for interpolation waypoints.
Expand Down
2 changes: 2 additions & 0 deletions src/rendering/swrenderer/plane/r_skyplane.cpp
View file
Expand Up @@ -222,6 +222,8 @@ namespace swrenderer
DrawSky(pl);
}

static uint32_t UMulScale16(uint32_t a, uint32_t b) { return (uint32_t)(((uint64_t)a * b) >> 16); }

void RenderSkyPlane::DrawSkyColumnStripe(int start_x, int y1, int y2, double scale, double texturemid, double yrepeat)
{
RenderPortal *renderportal = Thread->Portal.get();
Expand Down
1 change: 0 additions & 1 deletion src/rendering/swrenderer/things/r_visiblesprite.cpp
View file
Expand Up @@ -239,7 +239,6 @@ namespace swrenderer
// killough 3/27/98: end special clipping for deep water / fake ceilings
else if (!spr->IsVoxel() && spr->floorclip)
{ // [RH] Move floorclip stuff from R_DrawVisSprite to here
//int clip = ((FLOAT2FIXED(CenterY) - FixedMul (spr->texturemid - (spr->pic->GetHeight() << FRACBITS) + spr->floorclip, spr->yscale)) >> FRACBITS);
int clip = xs_RoundToInt(viewport->CenterY - (spr->texturemid - spr->pic->GetHeight() + spr->floorclip) * spr->yscale);
if (clip < botclip)
{
Expand Down
56 changes: 28 additions & 28 deletions src/rendering/swrenderer/things/r_voxel.cpp
View file
Expand Up @@ -323,8 +323,8 @@ namespace swrenderer
sprsinang = FLOAT2FIXED(-dasprang.Sin()) >> 2;

// Select mip level
i = abs(DMulScale6(dasprx - globalposx, cosang, daspry - globalposy, sinang));
i = DivScale6(i, MIN(daxscale, dayscale));
i = abs(DMulScale(dasprx - globalposx, cosang, daspry - globalposy, sinang, 6));
i = DivScale(i, MIN(daxscale, dayscale), 6);
j = xs_Fix<13>::ToFix(viewport->FocalLengthX);
for (k = 0; i >= j && k < voxobj->NumMips; ++k)
{
Expand All @@ -338,10 +338,10 @@ namespace swrenderer
maxslabz >>= k;

daxscale <<= (k + 8); dayscale <<= (k + 8);
dazscale = FixedDiv(dayscale, FLOAT2FIXED(viewport->BaseYaspectMul));
dazscale = DivScale(dayscale, FLOAT2FIXED(viewport->BaseYaspectMul), 16);
daxscale = fixed_t(daxscale / viewport->YaspectMul);
daxscale = Scale(daxscale, xdimenscale, viewport->viewwindow.centerxwide << 9);
dayscale = Scale(dayscale, FixedMul(xdimenscale, viewport->viewingrangerecip), viewport->viewwindow.centerxwide << 9);
dayscale = Scale(dayscale, MulScale(xdimenscale, viewport->viewingrangerecip, 16), viewport->viewwindow.centerxwide << 9);

daxscalerecip = (1 << 30) / daxscale;
dayscalerecip = (1 << 30) / dayscale;
Expand All @@ -350,26 +350,26 @@ namespace swrenderer
fixed_t piv_y = fixed_t(mip->Pivot.Y*256.);
fixed_t piv_z = fixed_t(mip->Pivot.Z*256.);

x = FixedMul(globalposx - dasprx, daxscalerecip);
y = FixedMul(globalposy - daspry, daxscalerecip);
backx = (DMulScale10(x, sprcosang, y, sprsinang) + piv_x) >> 8;
backy = (DMulScale10(y, sprcosang, x, -sprsinang) + piv_y) >> 8;
x = MulScale(globalposx - dasprx, daxscalerecip, 16);
y = MulScale(globalposy - daspry, daxscalerecip, 16);
backx = (DMulScale(x, sprcosang, y, sprsinang, 10) + piv_x) >> 8;
backy = (DMulScale(y, sprcosang, x, -sprsinang, 10) + piv_y) >> 8;
cbackx = clamp(backx, 0, mip->SizeX - 1);
cbacky = clamp(backy, 0, mip->SizeY - 1);

sprcosang = MulScale14(daxscale, sprcosang);
sprsinang = MulScale14(daxscale, sprsinang);
sprcosang = MulScale(daxscale, sprcosang, 14);
sprsinang = MulScale(daxscale, sprsinang, 14);

x = (dasprx - globalposx) - DMulScale18(piv_x, sprcosang, piv_y, -sprsinang);
y = (daspry - globalposy) - DMulScale18(piv_y, sprcosang, piv_x, sprsinang);
x = (dasprx - globalposx) - DMulScale(piv_x, sprcosang, piv_y, -sprsinang, 18);
y = (daspry - globalposy) - DMulScale(piv_y, sprcosang, piv_x, sprsinang, 18);

cosang = FixedMul(cosang, dayscalerecip);
sinang = FixedMul(sinang, dayscalerecip);
cosang = MulScale(cosang, dayscalerecip, 16);
sinang = MulScale(sinang, dayscalerecip, 16);

gxstart = y*cosang - x*sinang;
gystart = x*cosang + y*sinang;
gxinc = DMulScale10(sprsinang, cosang, sprcosang, -sinang);
gyinc = DMulScale10(sprcosang, cosang, sprsinang, sinang);
gxinc = DMulScale(sprsinang, cosang, sprcosang, -sinang, 10);
gyinc = DMulScale(sprcosang, cosang, sprsinang, sinang, 10);
if ((abs(globalposz - dasprz) >> 10) >= abs(dazscale)) return;

x = 0; y = 0; j = MAX(mip->SizeX, mip->SizeY);
Expand All @@ -381,7 +381,7 @@ namespace swrenderer
ggyinc[i] = y; y += gyinc;
}

syoff = DivScale21(globalposz - dasprz, FixedMul(dazscale, 0xE800)) + (piv_z << 7);
syoff = DivScale(globalposz - dasprz, MulScale(dazscale, 0xE800, 16), 21) + (piv_z << 7);
yoff = (abs(gxinc) + abs(gyinc)) >> 1;

bool useSlabDataBgra = !drawerargs.DrawerNeedsPalInput() && viewport->RenderTarget->IsBgra();
Expand Down Expand Up @@ -446,12 +446,12 @@ namespace swrenderer
uint8_t oand16 = oand + 16;
uint8_t oand32 = oand + 32;

if (yi > 0) { dagxinc = gxinc; dagyinc = FixedMul(gyinc, viewport->viewingrangerecip); }
else { dagxinc = -gxinc; dagyinc = -FixedMul(gyinc, viewport->viewingrangerecip); }
if (yi > 0) { dagxinc = gxinc; dagyinc = MulScale(gyinc, viewport->viewingrangerecip, 16); }
else { dagxinc = -gxinc; dagyinc = -MulScale(gyinc, viewport->viewingrangerecip, 16); }

/* Fix for non 90 degree viewing ranges */
nxoff = FixedMul(x2 - x1, viewport->viewingrangerecip);
x1 = FixedMul(x1, viewport->viewingrangerecip);
nxoff = MulScale(x2 - x1, viewport->viewingrangerecip, 16);
x1 = MulScale(x1, viewport->viewingrangerecip, 16);

ggxstart = gxstart + ggyinc[ys];
ggystart = gystart - ggxinc[ys];
Expand All @@ -462,7 +462,7 @@ namespace swrenderer
uint8_t *slabxoffs = &SlabData[mip->OffsetX[x]];
short *xyoffs = &mip->OffsetXY[x * (mip->SizeY + 1)];

nx = FixedMul(ggxstart + ggxinc[x], viewport->viewingrangerecip) + x1;
nx = MulScale(ggxstart + ggxinc[x], viewport->viewingrangerecip, 16) + x1;
ny = ggystart + ggyinc[x];
for (y = ys; y != ye; y += yi, nx += dagyinc, ny -= dagxinc)
{
Expand Down Expand Up @@ -522,20 +522,20 @@ namespace swrenderer
if (k < 0)
{
if ((voxptr->backfacecull & oand32) == 0) continue;
z2 = MulScale32(l2, k) + viewport->viewwindow.centery; /* Below slab */
z2 = MulScale(l2, k, 32) + viewport->viewwindow.centery; /* Below slab */
}
else
{
if ((voxptr->backfacecull & oand) == 0) continue; /* Middle of slab */
z2 = MulScale32(l1, k) + viewport->viewwindow.centery;
z2 = MulScale(l1, k, 32) + viewport->viewwindow.centery;
}
z1 = MulScale32(l1, j) + viewport->viewwindow.centery;
z1 = MulScale(l1, j, 32) + viewport->viewwindow.centery;
}
else
{
if ((voxptr->backfacecull & oand16) == 0) continue;
z1 = MulScale32(l2, j) + viewport->viewwindow.centery; /* Above slab */
z2 = MulScale32(l1, j + (zleng << 15)) + viewport->viewwindow.centery;
z1 = MulScale(l2, j, 32) + viewport->viewwindow.centery; /* Above slab */
z2 = MulScale(l1, j + (zleng << 15), 32) + viewport->viewwindow.centery;
}

if (z2 <= z1) continue;
Expand All @@ -546,7 +546,7 @@ namespace swrenderer
}
else
{
if (z2 - z1 >= 1024) yinc = FixedDiv(zleng, z2 - z1);
if (z2 - z1 >= 1024) yinc = DivScale(zleng, z2 - z1, 16);
else yinc = (((1 << 24) - 1) / (z2 - z1)) * zleng >> 8;
}
// [RH] Clip each column separately, not just by the first one.
Expand Down
4 changes: 2 additions & 2 deletions src/utility/nodebuilder/nodebuild_utility.cpp
View file
Expand Up @@ -453,8 +453,8 @@ void FNodeBuilder::FindPolyContainers (TArray<FPolyStart> &spots, TArray<FPolySt

if (PointOnSide (center.fixX(), center.fixY(), v1->x, v1->y, dx, dy) <= 0)
{
fixed_t t = DivScale30 (center.fixY() - v1->y, dy);
fixed_t sx = v1->x + MulScale30 (dx, t);
fixed_t t = DivScale (center.fixY() - v1->y, dy, 30);
fixed_t sx = v1->x + MulScale(dx, t, 30);
fixed_t dist = sx - spot->x;

if (dist < closestdist && dist >= 0)
Expand Down

0 comments on commit 58d8ce1

Please sign in to comment.