piwalker/melonDS
1
0
Fork 0
mirror of https://github.com/melonDS-emu/melonDS.git synced 2024-11-14 05:17:40 -07:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

Merge 19e8774ad0 into 7c1d2a64f4

Browse Source
This commit is contained in:
Jakly 2024-11-12 12:56:34 +01:00 committed by GitHub
commit a2b30c972a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 43 additions and 55 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
src/GPU3D_Soft.cpp
View File

@ -622,7 +622,7 @@ void SoftRenderer::SetupPolygonLeftEdge(SoftRenderer::RendererPolygon* rp, s32 y
rp->XL = rp->SlopeL.Setup(polygon->Vertices[rp->CurVL]->FinalPosition[0], polygon->Vertices[rp->NextVL]->FinalPosition[0], rp->XL = rp->SlopeL.Setup(polygon->Vertices[rp->CurVL]->FinalPosition[0], polygon->Vertices[rp->NextVL]->FinalPosition[0],
polygon->Vertices[rp->CurVL]->FinalPosition[1], polygon->Vertices[rp->NextVL]->FinalPosition[1], polygon->Vertices[rp->CurVL]->FinalPosition[1], polygon->Vertices[rp->NextVL]->FinalPosition[1],
polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL], y); polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL], y, polygon->WBuffer);
} }
void SoftRenderer::SetupPolygonRightEdge(SoftRenderer::RendererPolygon* rp, s32 y) const void SoftRenderer::SetupPolygonRightEdge(SoftRenderer::RendererPolygon* rp, s32 y) const
@ -649,7 +649,7 @@ void SoftRenderer::SetupPolygonRightEdge(SoftRenderer::RendererPolygon* rp, s32
rp->XR = rp->SlopeR.Setup(polygon->Vertices[rp->CurVR]->FinalPosition[0], polygon->Vertices[rp->NextVR]->FinalPosition[0], rp->XR = rp->SlopeR.Setup(polygon->Vertices[rp->CurVR]->FinalPosition[0], polygon->Vertices[rp->NextVR]->FinalPosition[0],
polygon->Vertices[rp->CurVR]->FinalPosition[1], polygon->Vertices[rp->NextVR]->FinalPosition[1], polygon->Vertices[rp->CurVR]->FinalPosition[1], polygon->Vertices[rp->NextVR]->FinalPosition[1],
polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR], y); polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR], y, polygon->WBuffer);
} }
void SoftRenderer::SetupPolygon(SoftRenderer::RendererPolygon* rp, Polygon* polygon) const void SoftRenderer::SetupPolygon(SoftRenderer::RendererPolygon* rp, Polygon* polygon) const
@ -755,8 +755,8 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
s32 wl = rp->SlopeL.Interp.Interpolate(polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL]); s32 wl = rp->SlopeL.Interp.Interpolate(polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL]);
s32 wr = rp->SlopeR.Interp.Interpolate(polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR]); s32 wr = rp->SlopeR.Interp.Interpolate(polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR]);
s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL], polygon->WBuffer); s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL]);
s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR], polygon->WBuffer); s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR]);
// right vertical edges are pushed 1px to the left as long as either: // right vertical edges are pushed 1px to the left as long as either:
// the left edge slope is not 0, or the span is not 0 pixels wide, and it is not at the leftmost pixel of the screen // the left edge slope is not 0, or the span is not 0 pixels wide, and it is not at the leftmost pixel of the screen
@ -841,7 +841,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
int edge; int edge;
s32 x = xstart; s32 x = xstart;
Interpolator<0> interpX(xstart, xend+1, wl, wr); Interpolator<0> interpX(xstart, xend+1, wl, wr, polygon->WBuffer, zl, zr);
if (x < 0) x = 0; if (x < 0) x = 0;
s32 xlimit; s32 xlimit;
@ -863,7 +863,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
interpX.SetX(x); interpX.SetX(x);
s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer); s32 z = interpX.InterpolateZ(zl, zr);
u32 dstattr = AttrBuffer[pixeladdr]; u32 dstattr = AttrBuffer[pixeladdr];
if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr)) if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr))
@ -889,7 +889,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
interpX.SetX(x); interpX.SetX(x);
s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer); s32 z = interpX.InterpolateZ(zl, zr);
u32 dstattr = AttrBuffer[pixeladdr]; u32 dstattr = AttrBuffer[pixeladdr];
if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr)) if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr))
@ -915,7 +915,7 @@ void SoftRenderer::RenderShadowMaskScanline(const GPU3D& gpu3d, RendererPolygon*
interpX.SetX(x); interpX.SetX(x);
s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer); s32 z = interpX.InterpolateZ(zl, zr);
u32 dstattr = AttrBuffer[pixeladdr]; u32 dstattr = AttrBuffer[pixeladdr];
if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr)) if (!fnDepthTest(DepthBuffer[pixeladdr], z, dstattr))
@ -980,8 +980,8 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
s32 wl = rp->SlopeL.Interp.Interpolate(polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL]); s32 wl = rp->SlopeL.Interp.Interpolate(polygon->FinalW[rp->CurVL], polygon->FinalW[rp->NextVL]);
s32 wr = rp->SlopeR.Interp.Interpolate(polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR]); s32 wr = rp->SlopeR.Interp.Interpolate(polygon->FinalW[rp->CurVR], polygon->FinalW[rp->NextVR]);
s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL], polygon->WBuffer); s32 zl = rp->SlopeL.Interp.InterpolateZ(polygon->FinalZ[rp->CurVL], polygon->FinalZ[rp->NextVL]);
s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR], polygon->WBuffer); s32 zr = rp->SlopeR.Interp.InterpolateZ(polygon->FinalZ[rp->CurVR], polygon->FinalZ[rp->NextVR]);
// right vertical edges are pushed 1px to the left as long as either: // right vertical edges are pushed 1px to the left as long as either:
// the left edge slope is not 0, or the span is not 0 pixels wide, and it is not at the leftmost pixel of the screen // the left edge slope is not 0, or the span is not 0 pixels wide, and it is not at the leftmost pixel of the screen
@ -1091,7 +1091,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
int edge; int edge;
s32 x = xstart; s32 x = xstart;
Interpolator<0> interpX(xstart, xend+1, wl, wr); Interpolator<0> interpX(xstart, xend+1, wl, wr, polygon->WBuffer, zl, zr);
if (x < 0) x = 0; if (x < 0) x = 0;
s32 xlimit; s32 xlimit;
@ -1130,7 +1130,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
interpX.SetX(x); interpX.SetX(x);
s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer); s32 z = interpX.InterpolateZ(zl, zr);
// if depth test against the topmost pixel fails, test // if depth test against the topmost pixel fails, test
// against the pixel underneath // against the pixel underneath
@ -1226,7 +1226,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
interpX.SetX(x); interpX.SetX(x);
s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer); s32 z = interpX.InterpolateZ(zl, zr);
// if depth test against the topmost pixel fails, test // if depth test against the topmost pixel fails, test
// against the pixel underneath // against the pixel underneath
@ -1318,7 +1318,7 @@ void SoftRenderer::RenderPolygonScanline(const GPU& gpu, RendererPolygon* rp, s3
interpX.SetX(x); interpX.SetX(x);
s32 z = interpX.InterpolateZ(zl, zr, polygon->WBuffer); s32 z = interpX.InterpolateZ(zl, zr);
// if depth test against the topmost pixel fails, test // if depth test against the topmost pixel fails, test
// against the pixel underneath // against the pixel underneath

70
src/GPU3D_Soft.h
View File

@ -69,23 +69,24 @@ private:
{ {
public: public:
constexpr Interpolator() {} constexpr Interpolator() {}
constexpr Interpolator(s32 x0, s32 x1, s32 w0, s32 w1) constexpr Interpolator(s32 x0, s32 x1, s32 w0, s32 w1, bool wbuffer, s32 z0, s32 z1)
{ {
Setup(x0, x1, w0, w1); Setup(x0, x1, w0, w1, wbuffer, z0, z1);
} }
constexpr void Setup(s32 x0, s32 x1, s32 w0, s32 w1) constexpr void Setup(s32 x0, s32 x1, s32 w0, s32 w1, bool wbuffer, s32 z0 = 0, s32 z1 = 0)
{ {
this->x0 = x0; this->x0 = x0;
this->x1 = x1; this->x1 = x1;
this->xdiff = x1 - x0; this->xdiff = x1 - x0;
this->wbuffer = wbuffer;
// calculate reciprocal for Z interpolation // calculate increment and init counter for Z interpolation
// TODO eventually: use a faster reciprocal function? if (!dir && !wbuffer && xdiff != 0)
if (this->xdiff != 0) {
this->xrecip_z = (1<<22) / this->xdiff; this->zincr = ((z1 - z0) >> 1) / xdiff << 1;
else this->zcounter = z0;
this->xrecip_z = 0; }
// linear mode is used if both W values are equal and have // linear mode is used if both W values are equal and have
// low-order bits cleared (0-6 along X, 1-6 along Y) // low-order bits cleared (0-6 along X, 1-6 along Y)
@ -144,7 +145,7 @@ private:
constexpr s32 Interpolate(s32 y0, s32 y1) const constexpr s32 Interpolate(s32 y0, s32 y1) const
{ {
if (xdiff == 0 || y0 == y1) return y0; if (x == 0 || xdiff == 0 || y0 == y1) return y0;
if (!linear) if (!linear)
{ {
@ -164,9 +165,9 @@ private:
} }
} }
constexpr s32 InterpolateZ(s32 z0, s32 z1, bool wbuffer) const constexpr s32 InterpolateZ(s32 z0, s32 z1)
{ {
if (xdiff == 0 || z0 == z1) return z0; if (x == 0 || xdiff == 0 || z0 == z1) return z0;
if (wbuffer) if (wbuffer)
{ {
@ -179,37 +180,22 @@ private:
else else
{ {
// Z-buffering: linear interpolation // Z-buffering: linear interpolation
// still doesn't quite match hardware... // not perfect, but close
s32 base = 0, disp = 0, factor = 0;
if (z0 < z1)
{
base = z0;
disp = z1 - z0;
factor = x;
}
else
{
base = z1;
disp = z0 - z1,
factor = xdiff - x;
}
if (dir) if (dir)
{ {
int shift = 0; // interpolating along y uses a different algorithm than x
while (disp > 0x3FF) // this algo probably isn't quite right though...
{ if (z0 < z1)
disp >>= 1; return z0 + (s64)(z1-z0) * x / xdiff;
shift++; else
} return z1 + (s64)(z0-z1) * (xdiff-x) / xdiff;
return base + ((((s64)disp * factor * xrecip_z) >> 22) << shift);
} }
else else
{ {
disp >>= 9; // unoptimized algorithm is: z0 + ((z1-z0 >> 1) / xdiff * x << 1);
return base + (((s64)disp * factor * xrecip_z) >> 13); // or alternatively there's: z0 + (z1-z0) / (xdiff<<1) * (x<<1);
return zcounter += zincr;
} }
} }
} }
@ -219,8 +205,10 @@ private:
int shift; int shift;
bool linear; bool linear;
bool wbuffer;
s32 xrecip_z; s32 zincr;
s32 zcounter;
s32 w0n, w0d, w1d; s32 w0n, w0d, w1d;
u32 yfactor; u32 yfactor;
@ -244,7 +232,7 @@ private:
Increment = 0; Increment = 0;
XMajor = false; XMajor = false;
Interp.Setup(0, 0, 0, 0); Interp.Setup(0, 0, 0, 0, false);
Interp.SetX(0); Interp.SetX(0);
xcov_incr = 0; xcov_incr = 0;
@ -252,7 +240,7 @@ private:
return x0; return x0;
} }
constexpr s32 Setup(s32 x0, s32 x1, s32 y0, s32 y1, s32 w0, s32 w1, s32 y) constexpr s32 Setup(s32 x0, s32 x1, s32 y0, s32 y1, s32 w0, s32 w1, s32 y, bool wbuffer)
{ {
this->x0 = x0; this->x0 = x0;
this->y = y; this->y = y;
@ -318,7 +306,7 @@ private:
s32 x = XVal(); s32 x = XVal();
int interpoffset = (Increment >= 0x40000) && (side ^ Negative); int interpoffset = (Increment >= 0x40000) && (side ^ Negative);
Interp.Setup(y0-interpoffset, y1-interpoffset, w0, w1); Interp.Setup(y0-interpoffset, y1-interpoffset, w0, w1, wbuffer);
Interp.SetX(y); Interp.SetX(y);
// used for calculating AA coverage // used for calculating AA coverage