Cache Replacement Scheme based on Back Propagation Neural

advertisement
Cache Replacement Scheme based on Back Propagation Neural Networks
In a high performance computer systems mainly depends on memory systems bandwidth.
Cache is the simplest approach to achieve high speed memory. Cache uses spatial and temporal
locality to provide illusion of faster memory.
The line replacement algorithm that is utilized by the cache system majorly decides the
performance of the cache. In order to utilize cache to the maximum we have to use the optimal
algorithm for page replacement. But the shortcomings of this is we need to know which pages
will be accessed by the process in future but this isn’t possible as we cannot predict the future.
Algorithm nearest to optimal algorithm is LRU.
In this project we are trying to implement a better cache replacement algorithm using Back
propagation neural network for the set associative cache. We feed the address of the cache
into neural network and the output of network will be the set and page to be replaced in order
to achieve a better performance. The output will be compared with the page to be replaced
provided by various other algorithms like FIFO, LRU and MRU.
Reference:
Performance evaluation of a new cache replacement by prof. Khalid, H; Dept. of Electr.
Eng., City Univ. of New York, NY, USA
Download