Resource Estimation API

The evobench.tools.resources submodule provides public utilities for preflight estimation of memory, runtime and dimensional feasibility.

Import path

from evobench.tools import estimate_feasibility
from evobench.tools.resources import FeasibilityBudget

Main functions

`estimate_single_array_memory_mb(population_size, dim, dtype_size_bytes=8)`

Estimates the memory required by one population-like numerical array.

`estimate_algorithm_memory_mb(...)`

Estimates algorithm working memory using configurable multipliers.

Important arguments:

Argument	Purpose
`algorithm_name`	Algorithm key such as `pso`, `eda`, `abc`, or a custom key
`population_size`	Number of individuals/candidates
`dim`	Number of decision variables
`multipliers`	Optional overrides for known or custom algorithms
`quadratic_multipliers`	Optional `dim x dim` overhead model

`estimate_runtime_seconds(...)`

Estimates runtime from empirical baselines.

Important arguments:

Argument	Purpose
`algorithm_name`	Algorithm key
`dim`	Problem dimensionality
`population_size`	Population size
`generations`	Number of iterations/generations
`baselines`	Optional custom runtime baselines
`benchmark_name`	Optional label for future benchmark-specific models

`estimate_feasibility(...)`

Builds a full feasibility report by combining memory and runtime estimates with a resource budget.

Returns a FeasibilityReport object.

Data classes

`RuntimeBaseline`

Represents an empirical runtime measurement used as the basis for extrapolation.

RuntimeBaseline(
    seconds=0.31,
    dim=100,
    population_size=1000,
    generations=10,
    complexity="linear_units",
)

`FeasibilityBudget`

Defines the maximum accepted resource envelope.

FeasibilityBudget(
    max_dimension=100,
    max_population_size=1000,
    max_estimated_memory_mb=1024.0,
    max_estimated_seconds=30.0,
)

`FeasibilityReport`

Contains the final diagnosis:

report.is_feasible
report.reasons
report.estimated_memory_mb
report.estimated_seconds
report.to_dict()

Notes

These utilities are planning tools. They do not guarantee exact runtime or memory consumption because real performance depends on CPU, BLAS backend, NumPy version, objective-function cost and implementation details.