[hkaiser]

This builds on #7926 and should be merged after it.

A transparent function object such as std::less<> names no operand type, so its task size can be known only from the arguments it is called with. This adds hpx::traits::get_task_size_for<F, Args...>, which gives the size of a call of f with arguments of the types Args. Unless it is specialized, it is the size of f itself (get_task_size), so a size given to f by nano_task(f), with_task_size<Size>(f), etc. always wins.

The function objects of the standard library (std::plus, std::minus, std::multiplies, std::divides, std::modulus, std::negate, the comparisons, the logical and the bitwise operations) and the comparisons in std::ranges are registered as builtin operations. So are the predicates in hpx::parallel::detail (equal_to, less, plus, ..., min_of, max_of). A transparent operation called with arithmetic, enum or pointer arguments is a nano task. A typed operation (std::less<int>, min_of<int>, ...) is a nano task if the type it names is such a type. In all other cases the size stays unknown, for instance std::plus<> on std::complex<double>.

detail::with_task_size_of takes the argument types. reduce, transform_reduce and merge now pass them, merge only if both projections are hpx::identity, since a projection may do any amount of work.

transform gives its iterations the size of the transformation of one or two elements, sized the same way through the new task_size_argument_t<Iter, Proj>: the reference of the element if it is not projected, a type of unknown size otherwise. merge uses it too.

Other algorithms are to follow, as are the sizes for datapar execution policies, where the function is called once per pack.

Tests: new compile-time and run-time checks in sized_function.cpp. The checks of the sizes the algorithms pass to the scheduler move from task_size_hint.cpp to the new libs/core/algorithms/tests/unit/task_sizes, with the counting scheduler in task_size_checks.hpp, so that the following PRs for the other algorithms can add their checks in files of their own.

[coderabbitai]

Caution

Review failed

The pull request is closed.

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}