let_value, _error, & _stopped: Destroy Child Operation State After Completion #1715
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…mpletion Previously the operation states for let_value, let_error, and let_stopped directly contained the operation states of one or two child operations. When first connected this number was one, and after the completion of that first operation and the connection of the sender returned by the wrapped invocable the number was two. Notably when the second operation state is connected the first operation has completed. P3373 proposes changing the above behavior such that after the tuple of values sent by the first operation is populated the operation state for the first operation is destroyed. Thereafter that storage can be reused for the operation state of the second operation thereby: - Release resources held by the first operation state sooner, and - Reducing the size of the operation states of let_value, let_error, and let_stopped This commit implements the above-described change.
ericniebler
requested changes
Dec 16, 2025
Comment on lines
+306
to
+325
| const auto __impl = [&]() noexcept(__nothrow) { | ||
| auto& __tuple = | ||
| __state.__args_.emplace_from(__tup::__mktuple, static_cast<_Args&&>(__args)...); | ||
| auto&& __sender = __tuple.apply(static_cast<_Fun&&>(__state.__fun_), __tuple); | ||
| __state.__storage_.template emplace<__monostate>(); | ||
| __second_rcvr_t __r{__rcvr, static_cast<__env2_t&&>(__state.__env2_)}; | ||
| auto& __op = __state.__storage_.template emplace<__submit_t>( | ||
| static_cast<__sender_t&&>(__sender), static_cast<__second_rcvr_t&&>(__r)); | ||
| __op.submit(static_cast<__sender_t&&>(__sender), static_cast<__second_rcvr_t&&>(__r)); | ||
| }; | ||
| if constexpr (__nothrow) { | ||
| __impl(); | ||
| } else { | ||
| STDEXEC_TRY { | ||
| __impl(); | ||
| } | ||
| STDEXEC_CATCH_ALL { | ||
| ::stdexec::set_error(static_cast<_Receiver&&>(__rcvr), std::current_exception()); | ||
| } | ||
| } |
Collaborator
There was a problem hiding this comment.
i learned from lewis that it is simpler to use try unconditionally, and only guard the set_error call in the catch block. the optimized codegen is identical.
Suggested change
| const auto __impl = [&]() noexcept(__nothrow) { | |
| auto& __tuple = | |
| __state.__args_.emplace_from(__tup::__mktuple, static_cast<_Args&&>(__args)...); | |
| auto&& __sender = __tuple.apply(static_cast<_Fun&&>(__state.__fun_), __tuple); | |
| __state.__storage_.template emplace<__monostate>(); | |
| __second_rcvr_t __r{__rcvr, static_cast<__env2_t&&>(__state.__env2_)}; | |
| auto& __op = __state.__storage_.template emplace<__submit_t>( | |
| static_cast<__sender_t&&>(__sender), static_cast<__second_rcvr_t&&>(__r)); | |
| __op.submit(static_cast<__sender_t&&>(__sender), static_cast<__second_rcvr_t&&>(__r)); | |
| }; | |
| if constexpr (__nothrow) { | |
| __impl(); | |
| } else { | |
| STDEXEC_TRY { | |
| __impl(); | |
| } | |
| STDEXEC_CATCH_ALL { | |
| ::stdexec::set_error(static_cast<_Receiver&&>(__rcvr), std::current_exception()); | |
| } | |
| } | |
| STDEXEC_TRY { | |
| auto& __tuple = | |
| __state.__args_.emplace_from(__tup::__mktuple, static_cast<_Args&&>(__args)...); | |
| auto&& __sender = __tuple.apply(static_cast<_Fun&&>(__state.__fun_), __tuple); | |
| __state.__storage_.template emplace<__monostate>(); | |
| __second_rcvr_t __r{__rcvr, static_cast<__env2_t&&>(__state.__env2_)}; | |
| auto& __op = __state.__storage_.template emplace<__submit_t>( | |
| static_cast<__sender_t&&>(__sender), static_cast<__second_rcvr_t&&>(__r)); | |
| __op.submit(static_cast<__sender_t&&>(__sender), static_cast<__second_rcvr_t&&>(__r)); | |
| } | |
| STDEXEC_CATCH_ALL { | |
| if constexpr (!__nothrow) { | |
| ::stdexec::set_error(static_cast<_Receiver&&>(__rcvr), std::current_exception()); | |
| } | |
| } |
Comment on lines
+573
to
+576
| template <typename _Sender> | ||
| using __sender_of = decltype((std::declval<__data_of<_Sender>>().__sndr)); | ||
| template <typename _Sender> | ||
| using __fun_of = decltype((std::declval<__data_of<_Sender>>().__fun)); |
Collaborator
There was a problem hiding this comment.
microbenchmarks showed that this somewhat bizarre implementation of declval compiles about 2x faster than the standard one:
template <class _Tp, bool _Noexcept = true>
using __declfn_t = auto (*)() noexcept(_Noexcept) -> _Tp;
template <class _Tp>
extern __declfn_t<_Tp &&> __declval;
Suggested change
| template <typename _Sender> | |
| using __sender_of = decltype((std::declval<__data_of<_Sender>>().__sndr)); | |
| template <typename _Sender> | |
| using __fun_of = decltype((std::declval<__data_of<_Sender>>().__fun)); | |
| template <typename _Sender> | |
| using __sender_of = decltype((__declval<__data_of<_Sender>>().__sndr)); | |
| template <typename _Sender> | |
| using __fun_of = decltype((__declval<__data_of<_Sender>>().__fun)); |
Comment on lines
513
to
+586
| auto operator()(_Sender&& __sndr, _Fun __fun) const -> __well_formed_sender auto { | ||
| return __make_sexpr<__let_t<_SetTag>>( | ||
| static_cast<_Fun&&>(__fun), static_cast<_Sender&&>(__sndr)); | ||
| __data_t{static_cast<_Sender&&>(__sndr), static_cast<_Fun&&>(__fun)}); |
Collaborator
There was a problem hiding this comment.
this is a problem. the structured binding of let_value is part of its interface. this should compile:
auto [tag, fn_copy, sndr_copy] = stdexec::let_value(sndr, fn);you can use transform_sender to swivel the fn and the sndr into a __data_t object. exec::repeat_n does this, so you can crib from that.
| return __sndr.apply( | ||
| static_cast<_Sender&&>(__sndr), | ||
| [&]<class _Fn, class _Child>(__ignore, _Fn&& __fn, _Child&& __child) { | ||
| // TODO(ericniebler): this needs a fallback |
Collaborator
There was a problem hiding this comment.
please don't loose this comment. i need to come back to this.
Collaborator
|
/ok to test 9f66f97 |
ericniebler
reviewed
Dec 16, 2025
Collaborator
|
/ok to test db1b0bb |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Previously the operation states for let_value, let_error, and let_stopped directly contained the operation states of one or two child operations. When first connected this number was one, and after the completion of that first operation and the connection of the sender returned by the wrapped invocable the number was two. Notably when the second operation state is connected the first operation has completed.
P3373 proposes changing the above behavior such that after the tuple of values sent by the first operation is populated the operation state for the first operation is destroyed. Thereafter that storage can be reused for the operation state of the second operation thereby:
This commit implements the above-described change.