Skip to main content

/

Docs

Log in

Mojo function

parallel_memcpy

parallel_memcpy[type: DType](dest: UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1], src: UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1], count: Int, count_per_task: Int, num_tasks: Int)

Copies count elements from a memory buffer src to dest in parallel by spawning num_tasks tasks each copying count_per_task elements.

Parameters:

  • ​type (DType): The element dtype.

Args:

  • ​dest (UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1]): The destination buffer.
  • ​src (UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1]): The source buffer.
  • ​count (Int): Number of elements in the buffer.
  • ​count_per_task (Int): Task size.
  • ​num_tasks (Int): Number of tasks to run in parallel.

parallel_memcpy[type: DType](dest: UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1], src: UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1], count: Int)

Copies count elements from a memory buffer src to dest in parallel.

Parameters:

  • ​type (DType): The element type.

Args:

  • ​dest (UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1]): The destination pointer.
  • ​src (UnsafePointer[SIMD[type, 1], 0, 0, alignof[::AnyType,__mlir_type.!kgen.target]() if triple_is_nvidia_cuda() else 1]): The source pointer.
  • ​count (Int): The number of elements to copy.

Was this page helpful?