Skip to content

nvexec sync_wait uses correct env when querying for completion scheduler #2393

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ericniebler merged 1 commit into from
Oct 31, 2026
+11 −16
Merged

nvexec sync_wait uses correct env when querying for completion scheduler #2393

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion include/nvexec/stream/common.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,7 +104,7 @@ namespace nv::execution

template <class Sender, class Env>
concept stream_completing_sender =
sender<Sender>
sender_in<Sender, Env>
&& gpu_stream_scheduler<
__result_of<get_completion_scheduler<set_value_t>, env_of_t<Sender>, Env>,
Env>;
Expand Down
2 changes: 1 addition & 1 deletion include/nvexec/stream/schedule_from.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,7 +184,7 @@ namespace nv::execution
struct transform_sender_for<STDEXEC::schedule_from_t>
{
template <class Env, class Sender>
auto operator()(Env const & env, __ignore, __ignore, Sender&& sndr) const
constexpr auto operator()(Env const & env, __ignore, __ignore, Sender&& sndr) const
{
if constexpr (stream_completing_sender<Sender, Env>)
{
Expand Down
23 changes: 9 additions & 14 deletions include/nvexec/stream/sync_wait.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -179,24 +179,26 @@ namespace nv::execution::_strm

struct sync_wait_t
{
template <sender_in<env> Sender>
template <stream_completing_sender<env> Sender>
requires(__count_of<set_value_t, Sender, env>::value == 1)
auto operator()(context ctx, Sender&& sndr) const //
-> std::optional<sync_wait_result_t<Sender>>
auto operator()(Sender&& sndr) const -> std::optional<sync_wait_result_t<Sender>>
{
using state_t = _sync_wait::state<Sender>;
state_t state{};
run_loop loop;

auto sched = STDEXEC::get_completion_scheduler<set_value_t>(STDEXEC::get_env(sndr),
env{loop.get_scheduler()});

cudaError_t status = cudaSuccess;
auto opstate = host_allocate<exit_opstate_t<Sender, receiver<Sender>>>(
status,
ctx.pinned_resource_,
sched.ctx_.pinned_resource_,
__emplace_from{[&]
{
return exit_opstate(static_cast<Sender&&>(sndr),
receiver<Sender>{{}, &state, &loop},
ctx);
sched.ctx_);
}});
if (status != cudaSuccess)
{
Expand All @@ -222,15 +224,8 @@ namespace nv::execution::_strm
} // namespace _sync_wait

template <>
struct apply_sender_for<sync_wait_t>
{
template <stream_completing_sender<STDEXEC::env<>> Sender>
auto operator()(Sender&& sndr) const
{
auto sched = get_completion_scheduler<set_value_t>(get_env(sndr), STDEXEC::env{});
return _sync_wait::sync_wait_t{}(sched.ctx_, static_cast<Sender&&>(sndr));
}
};
struct apply_sender_for<sync_wait_t> : _sync_wait::sync_wait_t
{};
} // namespace nv::execution::_strm

namespace nvexec = nv::execution;
Loading