40 most important Parallel Computing mcq with answers HPC

Parallel Programming (HPC) multiple choice questions

1. Task dependency graph is ——————

1. directed
2. undirected
3. directed acyclic
4. undirected acyclic

directed acyclic

2. In task dependency graph longest directed path between any pair of start and finish node is called as ————–

1. total work
2. critical path

critical path

3. which of the following is not a granularity type

1. course grain
2. large grain
3. medium grain
4. fine grain

large grain

4. which of the following is a an example of data decomposition

1. matrix multiplication
2. merge sort
3. quick sort
4. 15 puzzal

matrix multiplication

5. which problems can be handled by recursive decomposition

1. backtracking
2. greedy method
3. divide and conquer problem
4. branch and bound

divide and conquer problem

6. In this decomposition problem decomposition goes hand in hand with its execution

1. recursive decomposition
2. data decomposition
3. exploratory decomposition
4. speculative decomposition

recursive decomposition

7. which of the following is not an example of explorative decomposition

1. queens problem
2. 15 puzzal problem
3. tic tac toe
4. quick sort

quick sort

8. Topological sort can be applied to which of the following graphs?

1. Undirected Cyclic Graphs
2. Directed Cyclic Graphs
3. Undirected Acyclic Graphs
4. Directed Acyclic Graphs

Directed Acyclic Graphs

9. In most of the cases, topological sort starts from a node which has __________

1. Maximum Degree
2. Minimum Degree
3. Any degree
4. Zero Degree

Zero Degree

10. Which of the following is not an application of topological sorting?

1. Finding prerequisite of a task
2. Finding Deadlock in an Operating System
3. Finding Cycle in a graph
4. Ordered Statistics

Ordered Statistics

11. In ————task are defined before starting the execution of the algorithm

12. What is GPU?

1. Grouped Processing Unit
2. Graphics Processing Unit
3. Graphical Performance Utility
4. Graphical Portable Unit

Graphics Processing Unit

13. which of the following is not the array distribution method of data partitioning

1. block
2. cyclic
3. block cyclic
4. chunk

chunk

14. blocking optimization is used to improve temporal locality for reduce

1. hit miss
2. misses
3. hit rate
4. cache misses

misses

15. CUDA thought that ‘unifying theme’ of every form of parallelism is

3. Size of data associated with tasks
4. both A and B

both A and B

17. Topological sort of a Directed Acyclic graph is?

1. Always unique
2. Always Not unique
3. Sometimes unique and sometimes not unique
4. Always unique if graph has even number of vertices

Sometimes unique and sometimes not unique

18. threads being block altogether  and being executed in the sets of 32 threads called a

3. 32 block
4. unit block

19. The threads in a thread block are distributed across SM units so that each thread is executed by one SM unit.

1. TRUE
2. FALSE

TRUE

20. If there is dependency between tasks it implies their is no need of interaction between them.

1. TRUE
2. FALSE

FALSE

21. When the topological sort of a graph is unique?

1. When there exists a hamiltonian path in the graph
2. In the presence of multiple nodes with indegree 0
3. In the presence of single node with indegree 0
4. In the presence of single node with outdegree 0

When there exists a hamiltonian path in the graph

22. A good mapping does not depends on which following  factor

2. the size of data associated with tasks

23. CUDA is a parallel computing platform and programming model

1. TRUE
2. FALSE

TRUE

24. Which of the following is not a form of parallelism supported by CUDA

1. Vector parallelism – Floating point computations are executed in parallel on wide vector units
3. Block and grid level parallelism – Different blocks or grids execute different tasks
4. Data parallelism – Different threads and blocks process different parts of data in memory

Vector parallelism – Floating point computations are executed in parallel on wide vector units

25. The style of parallelism supported on GPUs is best described as

1. MISD – Multiple Instruction Single Data
2. SIMT – Single Instruction Multiple Thread
3. SISD – Single Instruction Single Data
4. MIMD

SIMT – Single Instruction Multiple Thread

26. Functions annotated with the __global__ qualifier may be executed on the host or the device

1. TRUE
2. FALSE

TRUE

27. Which of the following correctly describes a GPU kernel

1. A kernel may contain a mix of host and GPU code
2. All thread blocks involved in the same computation use the same kernel
3. A kernel is part of the GPU’s internal micro-operating system, allowing it to act as in independent host
4. kernel may contain only host code

All thread blocks involved in the same computation use the same kernel

28. a code known as grid which runs on GPU consisting of a set of

2. unit block
3. 32 block

29. which of the following is not an parallel algorithm model

1. data parallel model
4. work pool model

30. Having load before the store in a running program order, then interchanging this order, results in a

1. WAW hazards
2. Destination registers
3. WAR hazards
4. Registers

WAR hazards

high performance computing mcq – Parallel Computing

31. model based on the passing of stream of data through process arranged in a succession is called as

1. producer consumer model
2. hybrid model
4. work pool model

producer consumer model

32. When instruction i and instruction j are tends to write the same register or the memory location, it is called

1. Input dependence
2. Output dependence
3. Ideal pipeline
4. Digital call

Output dependence

1. Multiple processor
2. Single processor
3. Dual core
4. Corei5

Single processor

34. Allowing multiple instructions for issuing in a clock cycle, is the goal of

1. Single-issue processors
2. Dual-issue processors
3. Multiple-issue processors
4. No-issue processors

Multiple-issue processors

35. OpenGL stands for:

1. Open General Liability
2. Open Graphics Library
3. Open Guide Line
4. Open Graphics Layer

Open Graphics Library

high performance computing mcq sppu

36. which of the following is not an advantage of OpenGL

1. There is more detailed documentation for OpenGL
2. while other API’s don’t have such detailed documentation.
3. OpenGL is portable.
4. OpenGL is more functional than any other API.
5. It is not a  cross-platform API,

It is not a cross-platform API,

37. work pool  model uses —————- approach for task assignment

1. static
2. dynamic
3. centralized
4. decentralized

dynamic

38. which of the following is false regarding data parallel model

1. all task perform same computations
2. degree of parallelism increase with size of problem
3. matrix multiplication is example of data parallel computations
4. dynamic mapping is done

dynamic mapping is done

39. which of the following are methods for containing interaction overheads

1. maximizing data locality
2. minimize volumn of data exchange
3. min frequency of interactions
4. all the above

all the above

40. which of the following are classes of dynamic mapping centralized method

1. self scheduling
2. chunk scheduling
3. both a and b
4. none of the above

both a and b

41. which of the following is not scheme for static mapping

1. block distribution
2. block cyclic distributions
3. cyclic distributions
4. self scheduling