A parameterized (generic) algorithm is written once in terms of parameters — the data type, a comparison rule, a size or a strategy — instead of being hard-coded for one case. Templates in C++ and generics in Java are the usual tools.
Reusability: one tested implementation serves many types and behaviours, for example a single sort for integers, strings or employee records, ordered by any comparator.
static <T> void sort(List<T> items, Comparator<T> order) { ... }
sort(names, Comparator.naturalOrder());
sort(employees, Comparator.comparing(Employee::salary));Efficiency: less duplicated code to maintain and fewer bugs; with C++ templates the compiler generates specialised code for each type, so there is no run-time cost; tuning parameters (block size, threshold) lets one algorithm adapt to the input.
(In complexity theory the term also means fixed-parameter algorithms, running in f(k)·n^c time for a small parameter k, which makes some NP-hard problems practical.)
— rhs