Deck 12: Cuda Model, Parallelism, Memory System, and Communication Models

Question 1

the BlockPerGrid and ThreadPerBlock parameters are related to the ________ model supported by CUDA.

A)host
B)kernel
C)thread abstraction
D)none of above

Answer 1

thread abstraction

Question 2

_________ is Callable from the device only

A)_host_
B)__global__ 
C)_device_
D)none of above

Answer 2

_device_

Question 3

______ is Callable from the host

A)_host_
B)__global__ 
C)_device_
D)none of above

Answer 3

__global__ 

Question 4

______ is Callable from the host

A)_host_
B)__global__ 
C)_device_
D)none of above

Question 5

Cache memory works on the principle of

A)locality of data  
B)locality of memory
C)locality of reference  
D)locality of reference & memory

Question 6

SIMD represents an organization that ______________.

A)refers to a computer system capable of processing several programs at the same time.
B)represents organization of single computer containing a control unit, processor unit and a memory unit.
C) includes many processing units under the supervision of a common control unit
D)none of the above.

Question 7

Select different aspects of parallelism

A)server applications utilize high aggregate network bandwidth
B)scientific applications typically utilize high processing and memory system performance
C)all of the above
D)data intensive applications utilize high aggregate throughput

Question 8

Select correct answer: DRAM access times have only improved at the rate of roughly___________ % per year over this interval.

A)20
B)40
C)50
D)10

Question 9

Analyze, if the second instruction has data dependencies with the first, but the third instruction does not, the first

A)out-of-order
B)both of the above
C)none of the above
D)in-order

Question 10

Select the parameters which captures Memory system performance

A)bandwidth
B)both of the above
C)none of the above
D)latency

Question 11

Consider the example of a fire- hose. If the water comes out of the hose five seconds after the hydrant is turned on. Once the water starts flowing, if the hydrant delivers water at the rate of 15 gallons/second. Analyze the bandwidth and latency.

A)bandwidth: 5*15 gallons/second and latency: 15 seconds
B)bandwidth: 15 gallons/second and latency: 5 seconds
C)bandwidth: 3 gallons/second and latency: 5 seconds
D)bandwidth: 5 gallons/second and latency: 15 seconds

Question 12

Select alternate approaches for Hiding Memory Latency

A)multithreading
B)spatial locality
C)all of the above
D)prefeching

Question 13

Select which clause in OpenMP is similar to the private, except values of variables are initialized to corresponding values before the

A)firstprivate
B)shared
C)all of the above
D)private

Question 14

Which of the following projects of Blue Gene is not in development?

A)blue gene / m
B)blue gene / p
C)blue gene / q
D)blue gene / l

Question 15

A decomposition can be illustrated in the form of a directed graph with nodes corresponding to tasks and edges indicating that the result of one task is required for processing the next. Such graph is called as

A)task dependency graph
B)task interaction graph
C)process interaction graph
D)process dependency graph

Question 16

In which case, the owner computes rule implies that the output is computed by the process to which the output data is assigned?

A)output data decomposition
B)both of the above
C)none of the above
D)input data decomposition

Question 17

Select relevant task characteristics from the options given below:

A)task sizes
B)size of data associated with tasks
C)all of the above
D)task generation

Question 18

A classic example of game playing - each 15 puzzle board is the example of

A)dynamic task generation
B)none of the above
C)all of the above
D)static task generation

Question 19

Which model is equally suitable to shared-address- space or message- passing paradigms, since the interaction is naturally two ways.

A)master slave model
B)data parallel model
C)producer consumer or pipeline model
D)work pool model

Question 20

In which type of the model, tasks are dynamically assigned to the processes for balancing the load?

A)master slave model
B)data parallel model
C)producer consumer or pipeline model
D)work pool model

Question 21

Select the appropriate stage of GPU Pipeline which receives commands from CPU and also pulls geometry information from system memory.

A)vertex processing
B)memory interface
C)host interface
D)pixel processing

Question 22

In all-to-one reduction, data items must be combined piece-wise and the result made available at a___________             processor.

A)last
B)target
C)n-1
D)first

Question 23

Analyze the Cost of Scatter and Gather .

A)t=ts log p + tw m (p-1)
B)t=ts log p - tw m (p-1)
C)t=tw log p - ts m (p-1)
D)t=tw log p + ts m (p-1)

Question 24

All-to-all personalized communication is also known as___________                       .

A)total exchange
B)both of the above
C)none of the above
D)partial exchange

Question 25

All-to-all personalized communication is performed independently in each row with clustered messages of size on a mesh.

A)p
B)m?p
C)p?m
D)m