Introduction to Parallel Programming¶

1. What is the primary benefit of parallel programming in high-performance computing (HPC)?

• A. Reducing code complexity
• B. Accelerating computations and solving larger problems
• C. Simplifying debugging processes
• D. Improving data security

2. True or False: GPGPU programming is based on a SIMT architecture, which allows for massive parallel processing with many threads.

3. Parallel programming models are generally classified based on which criteria?

• A. Memory architecture they target
• B. Programming language used
• C. Processor speed
• D. Compiler type

4. Which of the following is NOT a parallel programming model?

• A. Shared Memory Programming
• B. Distributed Memory Programming
• C. Object-Oriented Programming
• D. GPGPU Programming

5. True or False: Shared memory programming is similar to GPGPU programming, as both involve multiple threads accessing a shared memory space.

Shared Memory Programming - OpenMP¶

1. OpenMP primarily uses which type of parallelism?

• A. Data-level parallelism
• B. Thread-level parallelism
• C. Task-level parallelism
• D. Instruction-level parallelism

2. True or False: In OpenMP, the fork-join model involves creating a team of threads at the beginning of a parallel section and then returning to a single thread after the section completes.

3. Which of the following are components of OpenMP?

• A. Compiler directives, runtime libraries, and environmental variables
• B. Data storage, cache management, and memory optimization
• C. Communication protocols, data encryption, and security protocols
• D. Memory mapping, kernel routines, and GPU scheduling

4. In a Uniform Memory Access (UMA) model:

• A. Memory access times vary across processors
• B. Memory access time is consistent for all processors
• C. Each processor has exclusive access to its own memory
• D. Memory access is limited to one processor at a time

5. True or False: Non-Uniform Memory Access (NUMA) allows each processor to access all memory regions with the same latency.

Distributed Memory Programming Model - Message Passing Interface (MPI)¶

1. Which memory model is commonly used in supercomputers and clusters where each processor has its own memory?

• A. Shared Memory Model
• B. Distributed Memory Model
• C. Hybrid Memory Model
• D. Direct Access Model

2. True or False: MPI is suitable for Single Program Multiple Data (SPMD) applications.

3. Which MPI function is used for sending data directly from one process to another?

• A. MPI_Bcast
• B. MPI_Send
• C. MPI_Allreduce
• D. MPI_Scatter

4. In distributed memory programming, processes communicate with each other through:

• A. Shared variables
• B. Network-based message passing
• C. Direct cache access
• D. Thread synchronization

5. True or False: In MPI, the memory of each process is shared across all nodes, allowing direct access by other processes.

GPU Programming - General-Purpose Computing on GPUs (GPGPU)¶

1. Which of the following programming models is commonly used for GPGPU programming on NVIDIA GPUs?

• A. OpenMP
• B. MPI
• C. CUDA
• D. POSIX

2. True or False: GPGPU programming requires efficient management of data transfers between CPU and GPU to avoid bottlenecks.

3. In GPGPU programming, what is a kernel?

• A. A function that runs in parallel on the GPU
• B. A data structure for memory management
• C. A debugging tool for GPU code
• D. A process management routine

4. What is the main advantage of GPGPU programming?

• A. Simplified coding
• B. Reduced power consumption
• C. Massive parallel processing capabilities
• D. Enhanced security

5. True or False: In GPGPU programming, the CPU and GPU share the same memory space, eliminating the need for data transfer.