AMDGOptimizing compilers can inline across translation units, these days.
francesco biscani wrote:
- header-only library: if used with small operands, GMP can suffer some
performance penalty related to function call overhead. Maybe it would be
possible to minimize it using aggressive inlining?