Parallel Computing Fundamentals (2)

First, fill in the blank

1. improve the structure of the computer system, primarily around the same time increasing the number of the operating interval, i.e. _ __ parallel processing _________ .

2. Parallel Algorithm issues and algorithm means for processing data in parallel on a variety of computer, its essence is to map a plurality of tasks to a plurality of processors for execution. Implement parallel algorithm depends strongly on the computer _ ___ hardware environment ________ and _ ____ _ soft pieces of environment _ _____ .

3. parallel algorithms in order relation between the process of program execution can be divided into synchronization algorithm, asynchronous algorithms , and _ _ independent stand parallel algorithms (distributed arithmetic) __________ .

4. In the process of performing parallel algorithm in order to ensure proper operation of the processors, and the correct access to shared data which may be necessary to properly set the synchronization point in the algorithm the position of the program. In time to force the implementation process at some point of time waiting for each other to cause the overhead called _ __ synchronization overhead time _________ .

The problems may be roughly divided into the size of the input and output size of the computations, _ _ memory requirements scale __________ and communication (synchronous) scale.

6. The algorithm performance metrics, in addition to the process-related operating time, parallel machine size and scale of the problem, the parallel algorithm metrics further comprises parallelism, size, cost, _ __ speedup _____ and _ ___ scalability ____ .

7. The best serial algorithm execution time on the order of the fastest machines with parallel algorithm than the running time on a parallel machine called _ ____ absolute speedup _______ .

8. The three fixed load speedup performance model speedup performance model, comprising a fixed time than the acceleration performance models and _ ____ memory-constrained _______ speedup model.

9. In a given run a parallel system of a given application of (parallel algorithms), if the size of the system increases with a parallel increase in the size of the problem properly, their performance scale parallel system is linearly proportional growth, the system is called a parallel _ _ extensible __________ of.

Second, the short answer

1. The main factors affecting the speedup brief narrative.

answer:

1) the ratio occupied by the serial part of the problem;

2) a large number of communication and synchronization

3). Unbalanced load

4) The resource conflict

5) The effectiveness of the algorithm

 

2, a brief description of defects than a fixed load acceleration performance model.

answer:

Lower than the acceleration curve with α sharp rise portion Ws of order due to the presence, not the number of processors used to solve the system increased.

(Unparallelizable serial Representative of partial load Ws,

Serial factor α ratio occupied by the serial portion)

 

3, a brief description of the purpose of scalability Parallel Algorithm.

answer:

1) to determine what type of questions to the parallel algorithm with which parallel system combination, the system can effectively utilize a large number of processors;

2) performance of the algorithm running on a small scale in parallel to predict the machine from running performance after transplant to massively parallel machine;

3) the problem of a fixed size can be determined from certain run on a parallel system, the most desired and the number of processors available maximum speedup.

4) Improved Algorithm parallel guide and parallel architecture of the machine

 

Published 152 original articles · won praise 124 · views 20000 +

Guess you like

Origin blog.csdn.net/qq_44762986/article/details/104760822