Hello World from GPU with CUDA and HIP¶

1. Which keyword in CUDA and HIP marks a function to execute on the GPU?

• A) __host__
• B) __global__
• C) __device__
• D) __shared__

Kernel Launch Configuration¶

2. What does the syntax <<<1, 1>>> specify in CUDA kernel launches?

• A) It sets one GPU block with one thread.
• B) It defines one CPU thread to run the GPU function.
• C) It sets the kernel to run on one GPU core.
• D) It specifies that the function will run on a single CPU core.

Synchronization¶

3. True or False: cudaDeviceSynchronize() in CUDA and hipDeviceSynchronize() in HIP ensure the CPU waits until the GPU completes all tasks.

HIP Portability¶

4. Which of the following is true about HIP?

• A) HIP code only runs on AMD GPUs.
• B) HIP code can be compiled to run on both AMD and NVIDIA GPUs.
• C) HIP requires major syntax changes from CUDA.
• D) HIP is only compatible with OpenCL devices.

Thread Synchronization in Kernels¶

5. In both CUDA and HIP, which command synchronizes all threads within a GPU block?

• A) cudaDeviceSynchronize() or hipDeviceSynchronize()
• B) __device_sync()
• C) __syncthreads()
• D) gpu_synchronize()