Skip to main content

/

Docs

Nightly

v25.1

Log in

Mojo function

load_acquire

load_acquire[type: DType, //, *, scope: Scope = Scope(6), memory: Bool = True](ptr: UnsafePointer[SIMD[type, 1], address_space=address_space, alignment=alignment, mut=mut, origin=origin]) -> SIMD[type, 1]

Performs an atomic load operation with acquire memory ordering semantics.

This function provides a memory barrier that ensures no subsequent memory operations from the calling thread are executed until after this load completes.

Note: - Only supported on NVIDIA GPUs. - Maps directly to PTX ld.acquire instruction. - Ensures subsequent memory operations don't execute until after load. - Critical for implementing synchronization primitives.

Parameters:

  • ​type (DType): The data type to load.
  • ​scope (Scope): Memory scope for the operation (default: Scope.SYSTEM).
  • ​memory (Bool): Whether to include memory side effects in constraints (default: True).

Args:

  • ​ptr (UnsafePointer[SIMD[type, 1], address_space=address_space, alignment=alignment, mut=mut, origin=origin]): Pointer to the memory location to load from.

Returns:

The loaded value.