1974 views|0 replies

1140

Posts

0

Resources
The OP
 

Computer System Architecture Q&A [Copy link]

1. If the actual flow rate of the channel required by the peripheral is very close to or equal to the maximum flow rate of the channel, local data loss may occur. How can we solve it? (1) Increase the maximum flow rate of the channel. (2) Dynamically change the device priority. (3) Add a certain number of buffers, especially for devices with lower priority. 2. What are the best ways to solve software migration? (1) Use series machines (2) Use simulation and emulation (3) Use a unified high-level language 3. What are the four key technologies of RISC processors? (1) Delayed transfer technology (2) Instruction cancellation technology (3) Overlapping register window technology (4) Instruction flow adjustment technology 4. What are the basic characteristics of multiprocessors? What are the main purposes of developing such a system? What technical problems do multiprocessors focus on solving? Features: (1) Flexible structure (2) Program parallelism (3) Parallel task derivation (4) Process synchronization (5) When multiple processors are working, they need to call resources according to the number of tasks. Therefore, the required resources vary in complexity. Purpose: Use multiple processors to execute a job concurrently, so that the execution speed is faster than that of a single processor. Key issues to be solved Key technical issues to be solved: (1) In terms of hardware structure, multiple processors, main memory, and I/O subsystems should have high bandwidth, low price, flexible and irregular interconnection, and avoid information transmission conflicts as much as possible. (2) From the perspective of parallel language parallel algorithm compilation, maximize the parallelism of program development. (3) How to divide a large task into multiple subtasks. (4) Solve the parallel task allocation, scheduling and resource allocation from the operating system; solve the synchronization between tasks or processes, deadlock competition and other problems. 5. What are the three ways to process vectors? What are the characteristics of each? (1) Horizontal processing method. Vector calculation is performed horizontally from left to right in a row manner. (2) Vertical processing method. Vector calculation is performed vertically from top to bottom in a column manner. (3) Vertical and horizontal processing. A method that combines horizontal processing and vertical processing. 6. What are the conditions for vector processors to achieve linking? (1) There is no conflict between vector registers and computing units. (2) Only the cycle in which the first result is sent to the vector register can be linked. (3) The time for the two preceding instructions to produce the results must be equal. (4) The vector lengths of the two vector instructions must be equal. 7. List the four routing methods in the interconnection network? And point out their respective advantages and disadvantages? (1) Line switching needs to frequently establish a physical path from the source node to the destination node before transmitting a message, which will cost a lot. (2) The buffer of the store-and-forward addressing packet is large, which is not conducive to VLSI implementation; the delay is large and is proportional to the distance between the nodes. (3) Virtual direct access does not need to wait until the entire message is buffered before making a route selection. As long as the message header used for addressing is received, the communication delay is independent of the number of nodes; it is also not conducive to VLSI implementation. (4) The buffer of each node of wormhole addressing is small, which is easy to implement in VLSI; lower network delay. 8. What are the main features of SIMD computers using parallel processors? (1) Fast speed and great potential (2) Good modularity, easy production and maintenance (3) High reliability, easy to achieve fault tolerance and reconstruction (4) Low efficiency. Compared with pipeline processors, vector processors, etc., it relies on resource duplication rather than time overlap, so its efficiency is lower. (5) Great potential. It mainly relies on increasing the number of PEs. Compared with pipeline processors that mainly rely on shortening the clock, its potential for speed increase is much greater. (6) It relies on the interconnection network and parallel algorithms. The interconnection network determines the connection mode between PEs and also determines the algorithms that the parallel processor can adapt to. (7) A high-performance scalar processor is required. If a machine has an extremely high vector processing speed, but a scalar processing speed of only one million times per second, then for a problem with scalar operations accounting for 10%, the total effective speed is no more than 10 million times per second. 9. What are the five basic instructions for designing a computer? (1) Data transfer instructions (2) Operation instructions (3) Program control instructions (4) Input and output instructions (5) Processing control and debugging instructions 10. When a pipeline encounters a bottleneck, how should the pipeline be handled to improve the efficiency of the pipeline? First, subdivide the "bottleneck" part of the pipeline; second, repeat the setting of bottleneck functional segments to allow multiple bottleneck functional segments to work in parallel. 11. When designing a computer system, what are the main principles for determining data representation? (Red) There are three main principles for determining data representation: one is to shorten the running time of the program, the second is to reduce the amount of communication between the CPU and the main memory, and the third is the versatility and utilization of this data representation. 12. What is a cluster system? A cluster system is a collection of multiple independent computers that are connected to each other. These computers can be stand-alone or multi-processor systems (PC workstations or SMPs). Each node has its own memory, I/O devices and operating system. The cluster is a single system for users and applications. It can provide a low-cost, high-performance environment and fast and reliable services. 13. What are the key technologies of cluster systems? (1) Efficient communication system; (2) Parallel programming environment; (3) Parallel programming language; (4) Load balancing technology (5) Parallel program adjustment technology (6) Reliability technology 14. What is a parallel processor? It is suitable for high-speed numerical calculations, has a relatively fixed structure, relies on resource duplication as a parallel measure, and the interconnection network stipulates the connection mode of the processing unit. It is basically a computer dedicated to vector processing. 15. What are the three main parts of Illiac IV? (1) Array of processing units specifically for array operations (2) Array controller (3) Standard B6700 computer 16. What is the main difference between a multiprocessor system and a parallel processor? (1) The parallelism of parallel processors lies within the instructions, while the parallelism of multiprocessors lies outside the instructions. (2) Parallel processors bring together the same operating system, and each PE is started directly by instructions to work simultaneously. Multiprocessors use special instructions to represent concurrent relationships. When a task starts to execute, it can derive other tasks that are executed at the same time. If there are more tasks than processors, they enter the task queue and wait. (3) Parallel processors have only one CU, which is naturally synchronized. The execution time of multiple processors may be different. 17. In what technical aspects has China's "Tianhe-2" supercomputer achieved major breakthroughs? (1) Multi-array configurable cooperative parallel architecture; (2) High-speed expansion interconnection communication technology (3) Efficient heterogeneous cooperative technology; (4) Isolation-based security control technology; (5) Virtualized network simulation computing technology (6) Large-scale system fault tolerance technology; (8) System energy consumption comprehensive technology. 18. In a multiprocessor system, what causes the cache consistency problem, and what are the two main methods to solve it? Cache consistency problems are mainly caused by shared writable data, process migration, I/O transmission, etc. At present, the monitoring protocol and directory-based protocol are mainly used, which are suitable for different structures respectively. 19. The similarities and differences between superscalar processors and superpipeline processors. The working mode of superpipeline processors is different from that of superscalar processors. Supervariable processors increase the performance of processors at the cost of increasing hardware resources, while superpipeline processors increase a small amount of hardware and improve the performance of processors by fully overlapping the work of each component hardware. From the time-space diagram of the pipeline, the supervariable processor uses spatial parallelism, while the superpipeline processor uses temporal parallelism. 20. What are CISC and RISC? Some commonly used functions that were originally implemented by software are implemented by hardware instruction systems. This computer system is called a complex instruction system, namely CISC; RISC is a reduced instruction system. It is a design concept of computer architecture. It is not a product, but a milestone in the history of the development of modern computer structure.

This post is from Wireless Connectivity
 

Guess Your Favourite
Just looking around
Find a datasheet?

EEWorld Datasheet Technical Support

Related articles more>>

EEWorld
subscription
account

EEWorld
service
account

Automotive
development
circle

Copyright © 2005-2024 EEWORLD.com.cn, Inc. All rights reserved 京B2-20211791 京ICP备10001474号-1 电信业务审批[2006]字第258号函 京公网安备 11010802033920号
快速回复 返回顶部 Return list