This example demonstrates how to execute HIP operations and kernels asynchronously with regard to the host. For more information on this topic, please refer to the HIP documentation.

This example should be compared to the sequential kernel execution example and the event-based synchronization example.

1. Two data vectors are created both on the host and on the device.
2. Two streams are created.
3. Loop for a fixed number of iterations:
   1. Copy the vector contents from the host to the device. The first vector is copied in the first stream and the second vector in the second stream.
   2. Launch the first GPU kernel on the first stream which operates on the first vector and writes the result back to the same vector.
   3. Launch the second GPU kernel on the second stream which operates on both vectors and writes the result back to the second vector.
   4. Copy the results back to the corresponding host vectors. The first vector is copied in the first stream and the second vector in the second stream.
4. The streams and the host are synchronized.
5. The results are verified.
6. The streams are destroyed and the device memory is freed.

• hipStreamCreate creates a stream. A stream executes HIP operations asynchronously with regard to the host.
• hipStreamDestroy destroys a stream.
• hipMemcpyAsync performs a copy operation in a stream. The call returns immediately to the host and the copy operation is performed asynchronously.
• hipStreamSynchronize blocks the host until all operations in the given stream have finished.

• blockDim
• blockIdx
• threadIdx

• hipFree
• hipGetErrorString
• hipMalloc
• hipMemcpyAsync
• hipStreamCreate
• hipStreamDestroy
• hipStreamSynchronize