Eigen::TensorCostModel< Device > Class Template Reference

#include <contrib/Eigen/src/Tensor/TensorCostModel.h>

Detailed Description

template<typename Device>
class Eigen::TensorCostModel< Device >

A cost model used to limit the number of threads used for evaluating tensor expressions.

Uses a roofline model: cost = max(memory_time, compute_time) instead of summing them. This avoids overestimating cost for balanced workloads. Memory-bound operations are capped at a limited number of threads to avoid wasting cycles competing for shared memory bandwidth.

---

The documentation for this class was generated from the following file:

• /builds/libeigen/eigen/contrib/Eigen/src/Tensor/TensorCostModel.h