Skip to main content

Mojo function

mbarrier_arrive_expect_tx_relaxed

mbarrier_arrive_expect_tx_relaxed[type: AnyType, scope: Scope = Scope(3), space: Scope = Scope(3)](addr: UnsafePointer[type, address_space=AddressSpace(3), alignment=alignment, mut=mut, origin=origin], tx_count: SIMD[int32, 1]) -> SIMD[uint64, 1]

Configure a shared memory barrier to expect additional async transactions.

Updates the current phase of the memory barrier to track completion of additional asynchronous transactions. Only supported on NVIDIA GPUs.

Parameters:

  • type (AnyType): The type of the memory barrier.
  • scope (Scope): The scope of the memory barrier.
  • space (Scope): The space of the memory barrier.

Args:

  • addr (UnsafePointer[type, address_space=AddressSpace(3), alignment=alignment, mut=mut, origin=origin]): Pointer to the shared memory barrier.
  • tx_count (SIMD[int32, 1]): Number of expected transactions to track.

Returns:

The state.

Was this page helpful?