AuroraMiddleware/skia2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Avoid catastrophic cancellation when computing det for QuadUVMatrix

Browse Source 
With data with large values we can get into situations where clearly
degenerate quads are not giving a zero determinant, because of
errors when computing the additions and subtractions in sequence.
The solution is to compute the subtractions first, and then add them
together.

Bug: oss-fuzz:41445
Change-Id: I7febead4b020e11a137f1703b761af70c14e4e47
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/511796
Reviewed-by: Michael Ludwig <michaelludwig@google.com>
Commit-Queue: Jim Van Verth <jvanverth@google.com>
This commit is contained in:
Jim Van Verth 2022-02-23 12:47:43 -05:00 committed by SkCQ
parent a8689e3d4d
commit 8b53da2b37
1 changed files with 13 additions and 38 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

51
src/gpu/geometry/GrPathUtils.cpp
View File

@ -129,7 +129,6 @@ uint32_t GrPathUtils::generateCubicPoints(const SkPoint& p0,
}
void GrPathUtils::QuadUVMatrix::set(const SkPoint qPts[3]) {
SkMatrix m;
// We want M such that M * xy_pt = uv_pt
// We know M * control_pts = [0 1/2 1]
// [0 0 1]
@ -147,7 +146,12 @@ void GrPathUtils::QuadUVMatrix::set(const SkPoint qPts[3]) {
double y1 = qPts[1].fY;
double x2 = qPts[2].fX;
double y2 = qPts[2].fY;
double det = x0*y1 - y0*x1 + x2*y0 - y2*x0 + x1*y2 - y1*x2;
// pre-calculate some adjugate matrix factors for determinant
double a2 = x1*y2-x2*y1;
double a5 = x2*y0-x0*y2;
double a8 = x0*y1-x1*y0;
double det = a2 + a5 + a8;
if (!sk_float_isfinite(det)
|| SkScalarNearlyZero((float)det, SK_ScalarNearlyZero * SK_ScalarNearlyZero)) {
@ -191,50 +195,21 @@ void GrPathUtils::QuadUVMatrix::set(const SkPoint qPts[3]) {
double scale = 1.0/det;
// compute adjugate matrix
double a2, a3, a4, a5, a6, a7, a8;
a2 = x1*y2-x2*y1;
double a3, a4, a6, a7;
a3 = y2-y0;
a4 = x0-x2;
a5 = x2*y0-x0*y2;
a6 = y0-y1;
a7 = x1-x0;
a8 = x0*y1-x1*y0;
// this performs the uv_pts*adjugate(control_pts) multiply,
// then does the scale by 1/det afterwards to improve precision
m[SkMatrix::kMScaleX] = (float)((0.5*a3 + a6)*scale);
m[SkMatrix::kMSkewX] = (float)((0.5*a4 + a7)*scale);
m[SkMatrix::kMTransX] = (float)((0.5*a5 + a8)*scale);
m[SkMatrix::kMSkewY] = (float)(a6*scale);
m[SkMatrix::kMScaleY] = (float)(a7*scale);
m[SkMatrix::kMTransY] = (float)(a8*scale);
// kMPersp0 & kMPersp1 should algebraically be zero
m[SkMatrix::kMPersp0] = 0.0f;
m[SkMatrix::kMPersp1] = 0.0f;
m[SkMatrix::kMPersp2] = (float)((a2 + a5 + a8)*scale);
// It may not be normalized to have 1.0 in the bottom right
float m33 = m.get(SkMatrix::kMPersp2);
if (1.f != m33) {
m33 = 1.f / m33;
fM[0] = m33 * m.get(SkMatrix::kMScaleX);
fM[1] = m33 * m.get(SkMatrix::kMSkewX);
fM[2] = m33 * m.get(SkMatrix::kMTransX);
fM[3] = m33 * m.get(SkMatrix::kMSkewY);
fM[4] = m33 * m.get(SkMatrix::kMScaleY);
fM[5] = m33 * m.get(SkMatrix::kMTransY);
} else {
fM[0] = m.get(SkMatrix::kMScaleX);
fM[1] = m.get(SkMatrix::kMSkewX);
fM[2] = m.get(SkMatrix::kMTransX);
fM[3] = m.get(SkMatrix::kMSkewY);
fM[4] = m.get(SkMatrix::kMScaleY);
fM[5] = m.get(SkMatrix::kMTransY);
}
fM[0] = (float)((0.5*a3 + a6)*scale);
fM[1] = (float)((0.5*a4 + a7)*scale);
fM[2] = (float)((0.5*a5 + a8)*scale);
fM[3] = (float)(a6*scale);
fM[4] = (float)(a7*scale);
fM[5] = (float)(a8*scale);
}
}