Mapping std:: to hip::

hipThreads mirrors the C++ Concurrency Support Library. In most cases, porting CPU code means replacing the std:: qualifier with hip:: and adding __device__ or __host__ __device__ to the callables that run on the GPU. The tables below list the supported primitives and how they correspond to the standard library.

For behavioral differences that have no standard-library analogue, see Limitations.

Threads

C++ standard	hipThreads	Header
std::thread	hip::wthread	<hip/thread>
std::thread::id	hip::wthread::id	<hip/thread>
std::this_thread::get_id	hip::this_thread::get_id	<hip/thread>
std::this_thread::yield	hip::this_thread::pseudo_yield	<hip/thread>
std::this_thread::sleep_for	hip::this_thread::sleep_for	<hip/thread>
(no equivalent)	hip::this_thread::get_fiber_id	<hip/thread>

Mutexes and locks

C++ standard	hipThreads	Header
std::mutex	hip::spin_mutex (also hip::pseudo_mutex)	<hip/mutex>, <hip/pseudo_mutex>
std::lock_guard	hip::lock_guard	<hip/mutex>
std::unique_lock	hip::unique_lock	<hip/mutex>

Condition variables

C++ standard	hipThreads	Header
std::condition_variable_any	hip::condition_variable_any (also hip::spin_condition_variable, hip::pseudo_condition_variable)	<hip/condition_variable>, <hip/pseudo_condition_variable>

Note

The spin_ variants busy-wait. The pseudo_ variants busy-wait but periodically call pseudo_yield to let other GPU work progress. Choose a pseudo_ primitive only when you can guarantee no yield-loops occur; see Limitations.