libretro/ppsspp
1
0
Fork 0
mirror of https://github.com/libretro/ppsspp.git synced 2025-02-13 13:13:36 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Interpolate bezier patch colors using bernstein.

Browse Source 
This closely matches hardware (some color values seem to be off by one.)

See #7525.
This commit is contained in:
Unknown W. Brackets 2016-04-09 21:24:19 -07:00
parent 59d9524308
commit 2bafba3909
2 changed files with 45 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

61
GPU/Common/SplineCommon.cpp
View File

@ -130,6 +130,16 @@ static Vec3f Bernstein3D(const Vec3f& p0, const Vec3f& p1, const Vec3f& p2, cons
return p0 * bern0(x) + p1 * bern1(x) + p2 * bern2(x) + p3 * bern3(x); return p0 * bern0(x) + p1 * bern1(x) + p2 * bern2(x) + p3 * bern3(x);
} }
static Vec4f Bernstein3D(const Vec4f& p0, const Vec4f& p1, const Vec4f& p2, const Vec4f& p3, float x) {
if (x == 0) return p0;
else if (x == 1) return p3;
return p0 * bern0(x) + p1 * bern1(x) + p2 * bern2(x) + p3 * bern3(x);
}
static Vec4f Bernstein3D(const u32& p0, const u32& p1, const u32& p2, const u32& p3, float x) {
return Bernstein3D(Vec4f::FromRGBA(p0), Vec4f::FromRGBA(p1), Vec4f::FromRGBA(p2), Vec4f::FromRGBA(p3), x);
}
static Vec3f Bernstein3DDerivative(const Vec3f& p0, const Vec3f& p1, const Vec3f& p2, const Vec3f& p3, float x) { static Vec3f Bernstein3DDerivative(const Vec3f& p0, const Vec3f& p1, const Vec3f& p2, const Vec3f& p3, float x) {
return p0 * bern0deriv(x) + p1 * bern1deriv(x) + p2 * bern2deriv(x) + p3 * bern3deriv(x); return p0 * bern0deriv(x) + p1 * bern1deriv(x) + p2 * bern2deriv(x) + p3 * bern3deriv(x);
} }
@ -624,25 +634,38 @@ static void _BezierPatchHighQuality(u8 *&dest, u16 *&indices, int &count, int te
// First compute all the vertices and put them in an array // First compute all the vertices and put them in an array
SimpleVertex *&vertices = (SimpleVertex*&)dest; SimpleVertex *&vertices = (SimpleVertex*&)dest;
Vec3f *horiz = (Vec3f *)AllocateAlignedMemory((tess_u + 1) * 4 * sizeof(Vec3f), 16); Vec3f *posHoriz1 = (Vec3f *)AllocateAlignedMemory((tess_u + 1) * 4 * sizeof(Vec3f), 16);
Vec3f *horiz2 = horiz + (tess_u + 1) * 1; Vec3f *posHoriz2 = posHoriz1 + (tess_u + 1) * 1;
Vec3f *horiz3 = horiz + (tess_u + 1) * 2; Vec3f *posHoriz3 = posHoriz1 + (tess_u + 1) * 2;
Vec3f *horiz4 = horiz + (tess_u + 1) * 3; Vec3f *posHoriz4 = posHoriz1 + (tess_u + 1) * 3;
Vec4f *colHoriz1 = (Vec4f *)AllocateAlignedMemory((tess_u + 1) * 4 * sizeof(Vec4f), 16);
Vec4f *colHoriz2 = colHoriz1 + (tess_u + 1) * 1;
Vec4f *colHoriz3 = colHoriz1 + (tess_u + 1) * 2;
Vec4f *colHoriz4 = colHoriz1 + (tess_u + 1) * 3;
Vec3f *derivU1 = (Vec3f *)AllocateAlignedMemory((tess_u + 1) * 4 * sizeof(Vec3f), 16); Vec3f *derivU1 = (Vec3f *)AllocateAlignedMemory((tess_u + 1) * 4 * sizeof(Vec3f), 16);
Vec3f *derivU2 = derivU1 + (tess_u + 1) * 1; Vec3f *derivU2 = derivU1 + (tess_u + 1) * 1;
Vec3f *derivU3 = derivU1 + (tess_u + 1) * 2; Vec3f *derivU3 = derivU1 + (tess_u + 1) * 2;
Vec3f *derivU4 = derivU1 + (tess_u + 1) * 3; Vec3f *derivU4 = derivU1 + (tess_u + 1) * 3;
bool computeNormals = patch.computeNormals; const bool computeNormals = patch.computeNormals;
const bool computeColors = origVertType & GE_VTYPE_COL_MASK;
// Precompute the horizontal curves to we only have to evaluate the vertical ones. // Precompute the horizontal curves to we only have to evaluate the vertical ones.
for (int i = 0; i < tess_u + 1; i++) { for (int i = 0; i < tess_u + 1; i++) {
float u = ((float)i / (float)tess_u); float u = ((float)i / (float)tess_u);
horiz[i] = Bernstein3D(patch.points[0]->pos, patch.points[1]->pos, patch.points[2]->pos, patch.points[3]->pos, u); posHoriz1[i] = Bernstein3D(patch.points[0]->pos, patch.points[1]->pos, patch.points[2]->pos, patch.points[3]->pos, u);
horiz2[i] = Bernstein3D(patch.points[4]->pos, patch.points[5]->pos, patch.points[6]->pos, patch.points[7]->pos, u); posHoriz2[i] = Bernstein3D(patch.points[4]->pos, patch.points[5]->pos, patch.points[6]->pos, patch.points[7]->pos, u);
horiz3[i] = Bernstein3D(patch.points[8]->pos, patch.points[9]->pos, patch.points[10]->pos, patch.points[11]->pos, u); posHoriz3[i] = Bernstein3D(patch.points[8]->pos, patch.points[9]->pos, patch.points[10]->pos, patch.points[11]->pos, u);
horiz4[i] = Bernstein3D(patch.points[12]->pos, patch.points[13]->pos, patch.points[14]->pos, patch.points[15]->pos, u); posHoriz4[i] = Bernstein3D(patch.points[12]->pos, patch.points[13]->pos, patch.points[14]->pos, patch.points[15]->pos, u);
if (computeColors) {
colHoriz1[i] = Bernstein3D(patch.points[0]->color_32, patch.points[1]->color_32, patch.points[2]->color_32, patch.points[3]->color_32, u);
colHoriz2[i] = Bernstein3D(patch.points[4]->color_32, patch.points[5]->color_32, patch.points[6]->color_32, patch.points[7]->color_32, u);
colHoriz3[i] = Bernstein3D(patch.points[8]->color_32, patch.points[9]->color_32, patch.points[10]->color_32, patch.points[11]->color_32, u);
colHoriz4[i] = Bernstein3D(patch.points[12]->color_32, patch.points[13]->color_32, patch.points[14]->color_32, patch.points[15]->color_32, u);
}
if (computeNormals) { if (computeNormals) {
derivU1[i] = Bernstein3DDerivative(patch.points[0]->pos, patch.points[1]->pos, patch.points[2]->pos, patch.points[3]->pos, u); derivU1[i] = Bernstein3DDerivative(patch.points[0]->pos, patch.points[1]->pos, patch.points[2]->pos, patch.points[3]->pos, u);
@ -659,10 +682,10 @@ static void _BezierPatchHighQuality(u8 *&dest, u16 *&indices, int &count, int te
float v = ((float)tile_v / (float)tess_v); float v = ((float)tile_v / (float)tess_v);
float bv = v; float bv = v;
const Vec3f &pos1 = horiz[tile_u]; const Vec3f &pos1 = posHoriz1[tile_u];
const Vec3f &pos2 = horiz2[tile_u]; const Vec3f &pos2 = posHoriz2[tile_u];
const Vec3f &pos3 = horiz3[tile_u]; const Vec3f &pos3 = posHoriz3[tile_u];
const Vec3f &pos4 = horiz4[tile_u]; const Vec3f &pos4 = posHoriz4[tile_u];
SimpleVertex &vert = vertices[tile_v * (tess_u + 1) + tile_u]; SimpleVertex &vert = vertices[tile_v * (tess_u + 1) + tile_u];
@ -691,18 +714,24 @@ static void _BezierPatchHighQuality(u8 *&dest, u16 *&indices, int &count, int te
vert.uv[1] = v + patch.v_index * third; vert.uv[1] = v + patch.v_index * third;
} else { } else {
// Sample UV from control points // Sample UV from control points
// TODO: We probably need to use bernstein here too?
patch.sampleTexUV(u, v, vert.uv[0], vert.uv[1]); patch.sampleTexUV(u, v, vert.uv[0], vert.uv[1]);
} }
if (origVertType & GE_VTYPE_COL_MASK) { if (computeColors) {
patch.sampleColor(u, v, vert.color); const Vec4f &col1 = colHoriz1[tile_u];
const Vec4f &col2 = colHoriz2[tile_u];
const Vec4f &col3 = colHoriz3[tile_u];
const Vec4f &col4 = colHoriz4[tile_u];
vert.color_32 = Bernstein3D(col1, col2, col3, col4, bv).ToRGBA();
} else { } else {
memcpy(vert.color, patch.points[0]->color, 4); memcpy(vert.color, patch.points[0]->color, 4);
} }
} }
} }
FreeAlignedMemory(derivU1); FreeAlignedMemory(derivU1);
FreeAlignedMemory(horiz); FreeAlignedMemory(colHoriz1);
FreeAlignedMemory(posHoriz1);
GEPatchPrimType prim_type = patch.primType; GEPatchPrimType prim_type = patch.primType;
// Combine the vertices into triangles. // Combine the vertices into triangles.

21
GPU/Common/SplineCommon.h
View File

@ -36,17 +36,6 @@ inline float lerp(float a, float b, float x) {
return a + x * (b - a); return a + x * (b - a);
} }
// SLOW!
inline void lerpColor(const Vec4f &a, const Vec4f &b, float x, Vec4f &out) {
for (int i = 0; i < 4; i++) {
out[i] = a[i] + x * (b[i] - a[i]);
}
}
inline void lerpColor(const u8 *a, const u8 *b, float x, Vec4f &out) {
lerpColor(Vec4f::FromRGBA(a), Vec4f::FromRGBA(b), x, out);
}
// We decode all vertices into a common format for easy interpolation and stuff. // We decode all vertices into a common format for easy interpolation and stuff.
// Not fast but can be optimized later. // Not fast but can be optimized later.
struct BezierPatch { struct BezierPatch {
@ -90,16 +79,6 @@ struct BezierPatch {
} }
}; };
// Interpolate colors between control points (bilinear, should be good enough).
void sampleColor(float u, float v, u8 color[4]) const {
const SamplingParams params(u, v);
Vec4f upperColor, lowerColor, resultColor;
lerpColor(points[params.tl]->color, points[params.tr]->color, params.fracU, upperColor);
lerpColor(points[params.bl]->color, points[params.br]->color, params.fracU, lowerColor);
lerpColor(upperColor, lowerColor, params.fracV, resultColor);
resultColor.ToRGBA(color);
}
void sampleTexUV(float u, float v, float &tu, float &tv) const { void sampleTexUV(float u, float v, float &tu, float &tv) const {
const SamplingParams params(u, v); const SamplingParams params(u, v);
float upperTU = lerp(points[params.tl]->uv[0], points[params.tr]->uv[0], params.fracU); float upperTU = lerp(points[params.tl]->uv[0], points[params.tr]->uv[0], params.fracU);