Change from boost::math to std

src/AdaptiveImportanceSampler.cc

@@ -20,12 +20,12 @@
  * the GNU General Public License along with this program.  If not,
  * see <http://www.lsstcorp.org/LegalNotices/>.
  */
+#include <cmath>
 
 #include "ndarray/eigen.h"
 
 #define LSST_MAX_DEBUG 10
 #include "lsst/pex/logging/Debug.h"
-#include "lsst/utils/ieee.h"
 #include "lsst/afw/table/BaseRecord.h"
 #include "lsst/afw/table/Catalog.h"
 #include "lsst/meas/modelfit/AdaptiveImportanceSampler.h"
@@ -133,7 +133,7 @@ void AdaptiveImportanceSampler::run(
             for (int k = 0; k < ctrl.nSamples; ++k) {
                 PTR(afw::table::BaseRecord) record = samples.addNew();
                 double objectiveValue = objective(parameters[k], *record);
-                if (utils::isfinite(objectiveValue)) {
+                if (std::isfinite(objectiveValue)) {
                     subSamples.push_back(record);
                     record->set(_parametersKey, parameters[k]);
                     record->set(_objectiveKey, objectiveValue);
@@ -157,7 +157,7 @@ void AdaptiveImportanceSampler::run(
             }
             computeRobustWeights(subSamples, _weightKey);
             perplexity = computeNormalizedPerplexity(subSamples);
-            if (!lsst::utils::isfinite(perplexity)) {
+            if (!std::isfinite(perplexity)) {
                 throw LSST_EXCEPT(
                     pex::exceptions::LogicError,
                     "Normalized perplexity is non-finite."

src/CModel.cc

@@ -987,7 +987,7 @@ class CModelAlgorithm::Impl {
         keys->checkBadReferenceFlag(record);
         // Check for unflagged NaNs.  Warn if we see any so we can fix the underlying problem, and
         // then flag them anyway.
-        if (lsst::utils::isnan(record.get(keys->flux)) && !record.get(keys->flags[CModelResult::FAILED])) {
+        if (std::isnan(record.get(keys->flux)) && !record.get(keys->flags[CModelResult::FAILED])) {
             // We throw a non-MeasurementError exception so the measurement error *will* log a warning.
             throw LSST_EXCEPT(
                 pex::exceptions::LogicError,

src/MarginalSamplingInterpreter.cc

@@ -20,8 +20,7 @@
  * the GNU General Public License along with this program.  If not,
  * see <http://www.lsstcorp.org/LegalNotices/>.
  */
-
-#include "boost/math/special_functions/round.hpp"
+#include <cmath>
 
 #include "ndarray/eigen.h"
 
@@ -255,7 +254,7 @@ void UnnestMarginalSamples::apply(
         marginalIter != marginalRecord.getSamples().end();
         ++marginalIter
     ) {
-        nNewSamplesTotal += boost::math::iround(
+        nNewSamplesTotal += std::lround(
             _multiplier
             * marginalRecord.getSamples().size()
             * marginalIter->get(_marginalInterpreter->getWeightKey())
@@ -276,7 +275,7 @@ void UnnestMarginalSamples::apply(
         ++marginalIter
     ) {
         Scalar marginalWeight = marginalIter->get(_marginalInterpreter->getWeightKey());
-        int nNewSamples = boost::math::iround(
+        int nNewSamples = std::lround(
             _multiplier * marginalRecord.getSamples().size() * marginalWeight
         );
         if (nNewSamples == 0) continue;

src/PixelFitRegion.cc

@@ -20,9 +20,9 @@
  * the GNU General Public License along with this program.  If not,
  * see <http://www.lsstcorp.org/LegalNotices/>.
  */
+#include <cmath>
 
 #include "lsst/meas/modelfit/PixelFitRegion.h"
-#include "lsst/utils/ieee.h"
 
 namespace lsst { namespace meas { namespace modelfit {
 

src/optimizer.cc

@@ -20,6 +20,7 @@
  * the GNU General Public License along with this program.  If not,
  * see <http://www.lsstcorp.org/LegalNotices/>.
  */
+#include <cmath>
 
 #include "Eigen/Eigenvalues"
 #include "boost/math/special_functions/erf.hpp"
@@ -28,7 +29,6 @@
 
 #define LSST_MAX_DEBUG 10
 #include "lsst/pex/logging/Debug.h"
-#include "lsst/utils/ieee.h"
 #include "lsst/afw/table/Catalog.h"
 #include "lsst/afw/table/BaseTable.h"
 #include "lsst/afw/table/BaseRecord.h"
@@ -196,7 +196,7 @@ void OptimizerObjective::fillObjectiveValueGrid(
         if (hasPrior()) {
             Scalar prior = computePrior(grid[i]);
             output[i] -= std::log(prior);
-            if (utils::isnan(output[i])) {
+            if (std::isnan(output[i])) {
                 output[i] = std::numeric_limits<Scalar>::infinity();
             }
         }
@@ -577,7 +577,7 @@ bool Optimizer::_stepImpl(
         );
         _next.parameters.asEigen() = _current.parameters.asEigen() + _step.asEigen();
         double stepLength = _step.asEigen().norm();
-        if (utils::isnan(stepLength)) {
+        if (std::isnan(stepLength)) {
             log.debug<7>("NaN encountered in step length");
             _state |= FAILED_NAN;
             return false;
@@ -586,7 +586,7 @@ bool Optimizer::_stepImpl(
         if (_objective->hasPrior()) {
             _next.priorValue = _objective->computePrior(_next.parameters);
             _next.objectiveValue = -std::log(_next.priorValue);
-            if (_next.priorValue <= 0.0 || utils::isnan(_next.objectiveValue)) {
+            if (_next.priorValue <= 0.0 || std::isnan(_next.objectiveValue)) {
                 _next.objectiveValue = std::numeric_limits<Scalar>::infinity();
                 log.debug<10>("Rejecting step due to zero prior");
                 if (stepLength < _trustRadius) {
@@ -617,7 +617,7 @@ bool Optimizer::_stepImpl(
             _gradient.asEigen() + 0.5*_hessian.asEigen()*_step.asEigen()
         );
         double rho = actualChange / predictedChange;
-        if (utils::isnan(rho)) {
+        if (std::isnan(rho)) {
             log.debug<10>("NaN encountered in rho");
             _state |= FAILED_NAN;
             return false;