Helper struct encapsulating most minor information for the auto tuner. More...
#include <AutoTunerInfo.h>

Public Attributes | |
| SelectorStrategyOption | selectorStrategy {SelectorStrategyOption::fastestAbs} |
| Strategy how to select the optimum from the collected samples. | |
| TuningMetricOption | tuningMetric {TuningMetricOption::time} |
| Metric used to rate configurations (time or energy). | |
| unsigned int | tuningInterval {5000} |
| Number of time steps after which the auto-tuner shall reevaluate the optimum. | |
| unsigned int | maxSamples {3} |
| Number of samples that shall be collected per combination. | |
| double | earlyStoppingFactor {std::numeric_limits<double>::infinity()} |
| EarlyStoppingFactor for the auto-tuner. | |
| EnergySensorOption | energySensor {EnergySensorOption::rapl} |
| Used energy sensor of energy metric selected. | |
| bool | useLOESSSmoothening {false} |
| Flag for whether LOESS Smoothening is used to smoothen the tuning results. | |
| size_t | aosSortingThreshold {8} |
| Default number of particles in two cells from which AoS sorting should be performed, used to seed this tuner's SortingThresholdBenchmark before/without a benchmark run. | |
| size_t | soaSortingThreshold {100} |
| Default number of particles in two SoA buffers from which SoA sorting should be performed, used to seed this tuner's SortingThresholdBenchmark before/without a benchmark run. | |
Helper struct encapsulating most minor information for the auto tuner.
| size_t autopas::AutoTunerInfo::aosSortingThreshold {8} |
Default number of particles in two cells from which AoS sorting should be performed, used to seed this tuner's SortingThresholdBenchmark before/without a benchmark run.
For details on the chosen default threshold see: https://github.com/AutoPas/AutoPas/pull/619
| size_t autopas::AutoTunerInfo::soaSortingThreshold {100} |
Default number of particles in two SoA buffers from which SoA sorting should be performed, used to seed this tuner's SortingThresholdBenchmark before/without a benchmark run.
Default comes from the LJFunctorHWY Benchmarks.