OpenSTA/dcalc/DelaySkewNormal.cc

294 lines
9.1 KiB
C++

// OpenSTA, Static Timing Analyzer
// Copyright (c) 2025, Parallax Software, Inc.
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with this program. If not, see <https://www.gnu.org/licenses/>.
//
// The origin of this software must not be misrepresented; you must not
// claim that you wrote the original software.
//
// Altered source versions must be plainly marked as such, and must not be
// misrepresented as being the original software.
//
// This notice may not be removed or altered from any source distribution.
#include "DelaySkewNormal.hh"
#include <cmath> // sqrt
#include "Error.hh"
#include "Fuzzy.hh"
#include "Units.hh"
#include "Format.hh"
#include "StaState.hh"
#include "Variables.hh"
namespace sta {
float
DelayOpsSkewNormal::stdDev2(const Delay &delay,
const EarlyLate *) const
{
return delay.stdDev2();
}
float
DelayOpsSkewNormal::asFloat(const Delay &delay,
const EarlyLate *early_late,
const StaState *sta) const
{
// LVF: mean + mean_shift + sigma * sigma_factor with skewness consideration.
float quantile = sta->variables()->pocvQuantile();
if (early_late == EarlyLate::early())
return delay.mean() + delay.meanShift()
- delay.stdDev() * (quantile + delay.skewness() * (square(quantile)-1.0) / 6.0);
else // (early_late == EarlyLate::late())
return delay.mean() + delay.meanShift()
+ delay.stdDev() * (quantile + delay.skewness() * (square(quantile)-1.0) / 6.0);
}
double
DelayOpsSkewNormal::asFloat(const DelayDbl &delay,
const EarlyLate *early_late,
const StaState *sta) const
{
// LVF: mean + mean_shift + sigma * sigma_factor with skewness consideration.
double quantile = sta->variables()->pocvQuantile();
if (early_late == EarlyLate::early())
return delay.mean() + delay.meanShift()
- delay.stdDev() * (quantile + delay.skewness() * (square(quantile)-1.0) / 6.0);
else // (early_late == EarlyLate::late())
return delay.mean() + delay.meanShift()
+ delay.stdDev() * (quantile + delay.skewness() * (square(quantile)-1.0) / 6.0);
}
bool
DelayOpsSkewNormal::isZero(const Delay &delay) const
{
return fuzzyZero(delay.mean())
&& fuzzyZero(delay.meanShift())
&& fuzzyZero(delay.stdDev2())
&& fuzzyZero(delay.skewness());
}
bool
DelayOpsSkewNormal::isInf(const Delay &delay) const
{
return fuzzyInf(delay.mean());
}
bool
DelayOpsSkewNormal::equal(const Delay &delay1,
const Delay &delay2,
const StaState *) const
{
return fuzzyEqual(delay1.mean(), delay2.mean())
&& fuzzyEqual(delay1.meanShift(), delay2.meanShift())
&& fuzzyEqual(delay1.stdDev2(), delay2.stdDev2())
&& fuzzyEqual(delay1.skewness(), delay2.skewness());
}
bool
DelayOpsSkewNormal::less(const Delay &delay1,
const Delay &delay2,
const StaState *sta) const
{
return fuzzyLess(delayAsFloat(delay1, EarlyLate::early(), sta),
delayAsFloat(delay2, EarlyLate::early(), sta));
}
bool
DelayOpsSkewNormal::less(const DelayDbl &delay1,
const DelayDbl &delay2,
const StaState *sta) const
{
return fuzzyLess(delayAsFloat(delay1, EarlyLate::early(), sta),
delayAsFloat(delay2, EarlyLate::early(), sta));
}
bool
DelayOpsSkewNormal::lessEqual(const Delay &delay1,
const Delay &delay2,
const StaState *sta) const
{
return fuzzyLessEqual(delayAsFloat(delay1, EarlyLate::early(), sta),
delayAsFloat(delay2, EarlyLate::early(), sta));
}
bool
DelayOpsSkewNormal::greater(const Delay &delay1,
const Delay &delay2,
const StaState *sta) const
{
return fuzzyGreater(delayAsFloat(delay1, EarlyLate::late(), sta),
delayAsFloat(delay2, EarlyLate::late(), sta));
}
bool
DelayOpsSkewNormal::greaterEqual(const Delay &delay1,
const Delay &delay2,
const StaState *sta) const
{
return fuzzyGreaterEqual(delayAsFloat(delay1, EarlyLate::late(), sta),
delayAsFloat(delay2, EarlyLate::late(), sta));
}
Delay
DelayOpsSkewNormal::sum(const Delay &delay1,
const Delay &delay2) const
{
return Delay(delay1.mean() + delay2.mean(),
delay1.meanShift() + delay2.meanShift(),
delay1.stdDev2() + delay2.stdDev2(),
skewnessSum(delay1, delay2));
}
float
DelayOpsSkewNormal::skewnessSum(const Delay &delay1,
const Delay &delay2) const
{
return skewnessSum(delay1.stdDev(), delay1.skewness(),
delay2.stdDev(), delay2.skewness());
}
// Helper function to compute combined skewness from std dev and skewness values.
double
DelayOpsSkewNormal::skewnessSum(double std_dev1,
double skewness1,
double std_dev2,
double skewness2) const
{
double std_dev_sum = square(std_dev1) + square(std_dev2);
if (std_dev_sum == 0.0)
return 0.0;
else {
// Un-normalize the skews so they are third moments so they can be added.
double skew = (skewness1 * cube(std_dev1) + skewness2 * cube(std_dev2))
// std_dev_sum^(3/2)
/ (std_dev_sum * std::sqrt(std_dev_sum));
return skew;
}
}
Delay
DelayOpsSkewNormal::sum(const Delay &delay1,
float delay2) const
{
return Delay(delay1.mean() + delay2,
delay1.meanShift(),
delay1.stdDev2(),
delay1.skewness());
}
Delay
DelayOpsSkewNormal::diff(const Delay &delay1,
const Delay &delay2) const
{
return Delay(delay1.mean() - delay2.mean(),
delay1.meanShift() - delay2.meanShift(),
delay1.stdDev2() + delay2.stdDev2(),
skewnessSum(delay1, delay2));
}
Delay
DelayOpsSkewNormal::diff(const Delay &delay1,
float delay2) const
{
return Delay(delay1.mean() - delay2,
delay1.meanShift(),
delay1.stdDev2(),
delay1.skewness());
}
Delay
DelayOpsSkewNormal::diff(float delay1,
const Delay &delay2) const
{
return Delay(delay1 - delay2.mean(),
delay2.meanShift(),
delay2.stdDev2(),
delay2.skewness());
}
void
DelayOpsSkewNormal::incr(Delay &delay1,
const Delay &delay2) const
{
delay1.setValues(delay1.mean() + delay2.mean(),
delay1.meanShift() + delay2.meanShift(),
delay1.stdDev2() + delay2.stdDev2(),
skewnessSum(delay1, delay2));
}
void
DelayOpsSkewNormal::incr(DelayDbl &delay1,
const Delay &delay2) const
{
delay1.setValues(delay1.mean() + delay2.mean(),
delay1.meanShift() + delay2.meanShift(),
delay1.stdDev2() + delay2.stdDev2(),
skewnessSum(delay1.stdDev(), delay1.skewness()));
}
void
DelayOpsSkewNormal::decr(Delay &delay1,
const Delay &delay2) const
{
delay1.setValues(delay1.mean() - delay2.mean(),
delay1.meanShift() + delay2.meanShift(),
delay1.stdDev2() + delay2.stdDev2(),
skewnessSum(delay1, delay2));
}
void
DelayOpsSkewNormal::decr(DelayDbl &delay1,
const Delay &delay2) const
{
delay1.setValues(delay1.mean() - delay2.mean(),
delay1.meanShift() + delay2.meanShift(),
delay1.stdDev2() + delay2.stdDev2(),
skewnessSum(delay1.stdDev(), delay1.skewness()));
}
Delay
DelayOpsSkewNormal::product(const Delay &delay1,
float delay2) const
{
return Delay(delay1.mean() * delay2,
delay1.meanShift() * delay2,
delay1.stdDev2() * square(delay2),
delay1.skewness() * cube(delay2));
}
Delay
DelayOpsSkewNormal::div(float delay1,
const Delay &delay2) const
{
return Delay(delay1 / delay2.mean());
}
std::string
DelayOpsSkewNormal::asStringVariance(const Delay &delay,
int digits,
const StaState *sta) const
{
const Unit *unit = sta->units()->timeUnit();
return sta::format("{}[{},{},{}]",
unit->asString(delay.mean(), digits),
unit->asString(delay.meanShift(), digits),
unit->asString(delay.stdDev(), digits),
sta->units()->scalarUnit()->asString(delay.skewness(), digits));
}
} // namespace