blob: d325d86a83c99a1cedbe3ad82c93c362c1d73444 [file] [log] [blame]
/*
* Copyright 2012 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "src/pathops/SkPathOpsTypes.h"
#include "include/private/base/SkFloatBits.h"
#include "include/private/base/SkFloatingPoint.h"
#include "include/private/base/SkMath.h"
#include "include/private/base/SkTemplates.h"
#include <algorithm>
#include <cstdint>
static bool arguments_denormalized(float a, float b, int epsilon) {
float denormalizedCheck = FLT_EPSILON * epsilon / 2;
return fabsf(a) <= denormalizedCheck && fabsf(b) <= denormalizedCheck;
}
// from http://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/
// FIXME: move to SkFloatBits.h
static bool equal_ulps(float a, float b, int epsilon, int depsilon) {
if (arguments_denormalized(a, b, depsilon)) {
return true;
}
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits < bBits + epsilon && bBits < aBits + epsilon;
}
static bool equal_ulps_no_normal_check(float a, float b, int epsilon, int depsilon) {
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits < bBits + epsilon && bBits < aBits + epsilon;
}
static bool equal_ulps_pin(float a, float b, int epsilon, int depsilon) {
if (!SkScalarIsFinite(a) || !SkScalarIsFinite(b)) {
return false;
}
if (arguments_denormalized(a, b, depsilon)) {
return true;
}
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits < bBits + epsilon && bBits < aBits + epsilon;
}
static bool d_equal_ulps(float a, float b, int epsilon) {
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits < bBits + epsilon && bBits < aBits + epsilon;
}
static bool not_equal_ulps(float a, float b, int epsilon) {
if (arguments_denormalized(a, b, epsilon)) {
return false;
}
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits >= bBits + epsilon || bBits >= aBits + epsilon;
}
static bool not_equal_ulps_pin(float a, float b, int epsilon) {
if (!SkScalarIsFinite(a) || !SkScalarIsFinite(b)) {
return false;
}
if (arguments_denormalized(a, b, epsilon)) {
return false;
}
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits >= bBits + epsilon || bBits >= aBits + epsilon;
}
static bool d_not_equal_ulps(float a, float b, int epsilon) {
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits >= bBits + epsilon || bBits >= aBits + epsilon;
}
static bool less_ulps(float a, float b, int epsilon) {
if (arguments_denormalized(a, b, epsilon)) {
return a <= b - FLT_EPSILON * epsilon;
}
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits <= bBits - epsilon;
}
static bool less_or_equal_ulps(float a, float b, int epsilon) {
if (arguments_denormalized(a, b, epsilon)) {
return a < b + FLT_EPSILON * epsilon;
}
int aBits = SkFloatAs2sCompliment(a);
int bBits = SkFloatAs2sCompliment(b);
// Find the difference in ULPs.
return aBits < bBits + epsilon;
}
// equality using the same error term as between
bool AlmostBequalUlps(float a, float b) {
const int UlpsEpsilon = 2;
return equal_ulps(a, b, UlpsEpsilon, UlpsEpsilon);
}
bool AlmostPequalUlps(float a, float b) {
const int UlpsEpsilon = 8;
return equal_ulps(a, b, UlpsEpsilon, UlpsEpsilon);
}
bool AlmostDequalUlps(float a, float b) {
const int UlpsEpsilon = 16;
return d_equal_ulps(a, b, UlpsEpsilon);
}
bool AlmostDequalUlps(double a, double b) {
if (fabs(a) < SK_ScalarMax && fabs(b) < SK_ScalarMax) {
return AlmostDequalUlps(SkDoubleToScalar(a), SkDoubleToScalar(b));
}
// We allow divide-by-zero here. It only happens if one of a,b is zero, and the other is NaN.
// (Otherwise, we'd hit the condition above). Thus, if std::max returns 0, we compute NaN / 0,
// which will produce NaN. The comparison will return false, which is the correct answer.
return sk_ieee_double_divide(fabs(a - b), std::max(fabs(a), fabs(b))) < FLT_EPSILON * 16;
}
bool AlmostEqualUlps(float a, float b) {
const int UlpsEpsilon = 16;
return equal_ulps(a, b, UlpsEpsilon, UlpsEpsilon);
}
bool AlmostEqualUlpsNoNormalCheck(float a, float b) {
const int UlpsEpsilon = 16;
return equal_ulps_no_normal_check(a, b, UlpsEpsilon, UlpsEpsilon);
}
bool AlmostEqualUlps_Pin(float a, float b) {
const int UlpsEpsilon = 16;
return equal_ulps_pin(a, b, UlpsEpsilon, UlpsEpsilon);
}
bool NotAlmostEqualUlps(float a, float b) {
const int UlpsEpsilon = 16;
return not_equal_ulps(a, b, UlpsEpsilon);
}
bool NotAlmostEqualUlps_Pin(float a, float b) {
const int UlpsEpsilon = 16;
return not_equal_ulps_pin(a, b, UlpsEpsilon);
}
bool NotAlmostDequalUlps(float a, float b) {
const int UlpsEpsilon = 16;
return d_not_equal_ulps(a, b, UlpsEpsilon);
}
bool RoughlyEqualUlps(float a, float b) {
const int UlpsEpsilon = 256;
const int DUlpsEpsilon = 1024;
return equal_ulps(a, b, UlpsEpsilon, DUlpsEpsilon);
}
bool AlmostBetweenUlps(float a, float b, float c) {
const int UlpsEpsilon = 2;
return a <= c ? less_or_equal_ulps(a, b, UlpsEpsilon) && less_or_equal_ulps(b, c, UlpsEpsilon)
: less_or_equal_ulps(b, a, UlpsEpsilon) && less_or_equal_ulps(c, b, UlpsEpsilon);
}
bool AlmostLessUlps(float a, float b) {
const int UlpsEpsilon = 16;
return less_ulps(a, b, UlpsEpsilon);
}
bool AlmostLessOrEqualUlps(float a, float b) {
const int UlpsEpsilon = 16;
return less_or_equal_ulps(a, b, UlpsEpsilon);
}
int UlpsDistance(float a, float b) {
SkFloatIntUnion floatIntA, floatIntB;
floatIntA.fFloat = a;
floatIntB.fFloat = b;
// Different signs means they do not match.
if ((floatIntA.fSignBitInt < 0) != (floatIntB.fSignBitInt < 0)) {
// Check for equality to make sure +0 == -0
return a == b ? 0 : SK_MaxS32;
}
// Find the difference in ULPs.
return SkTAbs(floatIntA.fSignBitInt - floatIntB.fSignBitInt);
}
SkOpGlobalState::SkOpGlobalState(SkOpContourHead* head,
SkArenaAlloc* allocator
SkDEBUGPARAMS(bool debugSkipAssert)
SkDEBUGPARAMS(const char* testName))
: fAllocator(allocator)
, fCoincidence(nullptr)
, fContourHead(head)
, fNested(0)
, fWindingFailed(false)
, fPhase(SkOpPhase::kIntersecting)
SkDEBUGPARAMS(fDebugTestName(testName))
SkDEBUGPARAMS(fAngleID(0))
SkDEBUGPARAMS(fCoinID(0))
SkDEBUGPARAMS(fContourID(0))
SkDEBUGPARAMS(fPtTID(0))
SkDEBUGPARAMS(fSegmentID(0))
SkDEBUGPARAMS(fSpanID(0))
SkDEBUGPARAMS(fDebugSkipAssert(debugSkipAssert)) {
#if DEBUG_T_SECT_LOOP_COUNT
debugResetLoopCounts();
#endif
#if DEBUG_COIN
fPreviousFuncName = nullptr;
#endif
}