Skip to content

fix(dynamic_batching): potential hang due to duration calculation #2976

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

Open
simshi wants to merge 1 commit into
base: main
Choose a base branch
from
+7 −7
Open

fix(dynamic_batching): potential hang due to duration calculation #2976

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
14 changes: 7 additions & 7 deletions cpp/src/neighbors/detail/dynamic_batching.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -644,12 +644,12 @@ struct gpu_time_keeper {
{
if (timeout) { return false; }
update_local_remaining_time();
if (local_remaining_time_us_ <= 0) {
if (local_remaining_time_ns_ <= 0) {
timeout = true;
return false;
}
update_cpu_provided_remaining_time();
if (local_remaining_time_us_ <= 0) {
if (local_remaining_time_ns_ <= 0) {
timeout = true;
return false;
}
Expand All @@ -659,7 +659,7 @@ struct gpu_time_keeper {
private:
cuda::atomic<int32_t, cuda::thread_scope_system>* cpu_provided_remaining_time_us_;
uint64_t timestamp_ns_ = 0;
int32_t local_remaining_time_us_ = std::numeric_limits<int32_t>::max();
int64_t local_remaining_time_ns_ = std::numeric_limits<int64_t>::max();
bool timeout = false;

RAFT_DEVICE_INLINE_FUNCTION void update_timestamp() noexcept
Expand All @@ -673,14 +673,14 @@ struct gpu_time_keeper {
update_timestamp();
// subtract the time passed since the last check
// (assuming local time is updated every time timestamp is read)
local_remaining_time_us_ -= static_cast<int32_t>((timestamp_ns_ - prev_timestamp) / 1000ull);
local_remaining_time_ns_ -= static_cast<int64_t>(timestamp_ns_ - prev_timestamp);
}

RAFT_DEVICE_INLINE_FUNCTION void update_cpu_provided_remaining_time() noexcept
{
local_remaining_time_us_ =
std::min<int32_t>(local_remaining_time_us_,
cpu_provided_remaining_time_us_->load(cuda::std::memory_order_relaxed));
auto cpu_us = cpu_provided_remaining_time_us_->load(cuda::std::memory_order_relaxed);
local_remaining_time_ns_ =
std::min<int64_t>(local_remaining_time_ns_, static_cast<int64_t>(cpu_us) * 1000);
}
};

Expand Down
Loading