cpl bench
run on maximum-size worst-case inputs against the limits
usage: cpl bench <problem> [options]
cpl bench --zero-cost [dir] [--reps N] [--max-ratio R] [--format json]
cpl bench --corpus [--runs N] [--ports a,b] [--out PATH] [--no-write]
Runs the solution on maximum-size inputs made from the problem's `read()`
types, in the shapes that break slow or deep solutions, and compares time
and memory with the problem's limits.
options:
--zero-cost run the zero-cost abstraction pairs (abstract.cp vs direct.cp)
--corpus benchmark corpus/problems/* against original C++ (p50/p95 + perf.cp)
--fast PROG run this instead of building the problem
--time-limit MS default: the problem's `// limits:` line, or 2000
--memory-limit MB default: the problem's `// limits:` line, or 256
--seed N the seed of the random shape (default 1)
--format json print the result as JSONGenerated from cpl 0.2.2 (dev, a0d0078e).