Contraintes et vérifications statiques

A constraint is a rule the compiler verifies about your parameters, stopping a bad build before the program ever runs. ⚙️

Call constrained with a compile-time condition; if it is false, compilation fails with the message you supply.

The condition must use parameters or aliases, so the check is settled while building, not while the program executes.

The string you pass becomes the error shown at the call site, so misuse is explained instead of failing in some confusing later way.

Constraints shine for SIMD code: require a power-of-two width so the kernel never gets an impossible vector size.

A clear constraint tells future readers exactly what a parameter expects, acting as machine-checked documentation that can never drift.

Sprinkle checks through a parametric algorithm so every specialization is validated, catching one bad combination without testing every input.

Because the test runs at compile time, the shipped code carries no if-statement to verify the rule, so there is zero runtime cost.

Combine constraints with @parameter if: branch on what is supported, and reject the unsupported cases outright with a clear error.

Constraints turn flexible parameters into a contract: powerful specialization stays safe because the compiler enforces every rule for you. 🚀

Place a constrained check near the top of a function so an invalid parameter is rejected immediately, with the reason stated up front.

Recall what constrained does in Mojo.

You used constraints: constrained validates parameters at compile time with clear errors, making powerful metaprogramming safe at zero runtime cost. 🎯