Basics in OpenACC¶

1. What is the main purpose of adding the kernels directive to a loop in OpenACC?

• A. To initialize memory on the GPU
• B. To instruct the compiler to attempt parallelization of the loop
• C. To run the loop sequentially on the CPU
• D. To copy data from the GPU to the CPU

Parallel Execution in OpenACC¶

2. True or False: In OpenACC, the term vector refers to a group of threads within a gang, similar to CUDA’s threads within a block.

3. When compiling the serial version of the Print_Hello_World function in C/C++, the compiler output indicated:

• A. The loop was fully parallelized by default
• B. A parallelization error due to missing directives
• C. “Loop not vectorized/parallelized: contains call”
• D. The code was not executable on the GPU

Key Concepts and Considerations¶

4. In OpenACC, what does gang typically refer to in GPU execution?

• A. A single thread
• B. A memory allocation block
• C. A group of threads, similar to CUDA blocks
• D. A function that must run sequentially

5. True or False: Fortran has limited I/O capabilities on the GPU, making it challenging to use print statements directly on the device.