ericniebler
Follow
Pinned
1,091 contributions in the last year
Contribution activity
March 2023
Created 60 commits in 3 repositories
Created a pull request in NVIDIA/stdexec that received 4 comments
Opened 22 other pull requests in 6 repositories
NVIDIA/stdexec
16
merged
- avoid constraint recursion in any_sender_of __storage move ctor
- repair diagnostics broken by R5/R7 concepts support
- s/__bool/__mbool/g
-
add
exec::trampoline_schedulerto bound stack usage - hide and short-circuit disjunctions in sender and receiver concepts
- correct various macro (and formatting) related offences
-
simplify the type of
into_variant's receiver - remove defunct namespace aliases
- fix broken formatting that is driving me CRAZY
-
simplify the
__submitreceiver -
Use
__builtin_bit_castwhen available for constexpr-ness -
add default implementations for
schedule_atandschedule_after - don't use fork if it's not available
- task promise should return env by reference
-
fix bad metaprogramming in
when_all -
Conform to the R7 specification of
as_awaitable
brycelelbach/wg21_p2300_execution
2
merged
BenFrantzDale/stdexec
1
merged
maikel/stdexec
1
merged
kirkshoop/stdexec
1
merged
NVIDIA/thrust
1
merged
Reviewed 13 pull requests in 2 repositories
NVIDIA/stdexec
12 pull requests
- Implement repeat_effect_until
- [WIP] Improve diagnostics if sync_wait fails due to errors in get_completion_signatures
-
add
exec::trampoline_schedulerto bound stack usage - Start refactoring CUDA error handling:
- Improvements for io_context
-
Use
__builtin_bit_castwhen available for constexpr-ness - Implement an io_uring context
- Scheduler affine coroutine task
- Hack/tbb thread pool
- Implement variant sender
- Implement the finally algorithm
- Only use std::exception_ptr in when_all when needed
brycelelbach/wg21_p2300_execution
1 pull request
Created an issue in NVIDIA/stdexec that received 4 comments
race condition in io_uring scheduler
One of my local test runs yielded this:
[ctest] 579/579 Test #575: After -1s ......................................................................…
4
comments
Opened 17 other issues in 2 repositories
NVIDIA/stdexec
1
closed
15
open
-
Possible memory leak in
any_scheduler -
schedule_aftershould start the workdurationtime after the work is *started* - Propose a design for async streams and a few essential stream algorithms
- Propose a way to make cancellation more efficient
- Propose scheduler queries for synchronization primitives (?)
- Propose a system scheduler
- Propose sender-based async parallel algorithms
- Propose an I/O scheduler concept
- Propose async cleanup as intrinsic part of the sender concept (?)
- Propose symmetric transfer for senders (tail senders?)
-
Propose
async_scopeand/orasync_resource - Update the coroutine task paper (P1056)
- Improve compiler diagnostics for when a sender & receiver won't connect
-
Propose
timed_schedulerconcept -
Extend
with_awaitable_senderswith adisposition()member that exposes how the coroutine completed -
Extend implementor freedom to drop
set_error_t(exception_ptr)completion to all library-provided senders