Vector Addition: CPU vs. GPU Implementation¶

1. In CPU vector addition, how are elements of the vectors processed?

• A. Sequentially, one by one
• B. In parallel, using threads
• C. By blocks of elements
• D. Randomly

GPU Vector Addition with CUDA¶

2. Which CUDA function is used to allocate memory on the GPU?

• A. malloc
• B. free
• C. cudaMalloc
• D. cudaMemcpy

3. What does cudaMemcpy(d_a, h_a, sizeof(float) * N, cudaMemcpyHostToDevice); do?

• A. Copies data from GPU to CPU
• B. Allocates memory on the GPU
• C. Copies data from CPU to GPU
• D. Frees memory on the GPU

Thread and Block Configuration¶

4. Which formula calculates the required number of blocks per grid?

• A. N / threadsPerBlock
• B. (N + threadsPerBlock - 1) / threadsPerBlock
• C. N * threadsPerBlock
• D. (N - threadsPerBlock) / threadsPerBlock

5. True or False: Using dim3 threadsPerBlock(16, 16) configures a 2D block with 256 threads.

Memory Management¶

6. Why is cudaFree() used in GPU programs?

• A. To allocate device memory
• B. To transfer data from host to device
• C. To deallocate device memory
• D. To initialize device memory

Kernel Execution and Indexing¶

7. What does blockIdx.x * blockDim.x + threadIdx.x calculate in CUDA?

• A. The global thread index
• B. The block size
• C. The grid size
• D. The shared memory size