Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Precise automatable analytical modeling of the cache behavior of codes with indirections
Andrade D., Fraguela B., Doallo R. ACM Transactions on Architecture and Code Optimization4 (3):16-es,2007.Type:Article
Date Reviewed: Nov 28 2007

The modeling of cache performance and the prediction of cache miss probability are difficult problems. Modeling the cache performance by trace-driven simulations has been widely studied in the past. However, the validity of these studies is hardware dependent, and these studies do not provide a good insight into the cache miss observations. Analytical models are expected to elucidate cache miss behaviors, and are strongly needed for implementing code optimization by production compilers.

The probabilistic miss equation (PME) model was an earlier analytical model, but it relies on the assumption that codes have regular access patterns. Like other analytical models, the model discussed in this paper is based on the assumption that the irregular access to the referenced array is uniformly distributed in the array. This assumption eases the modeling of the irregular access patterns posed for modeling the operations of the cache. The authors admit that most irregular accesses in real programs do not follow a uniform distribution. The authors also claim that experiments show that their model accurately predicts the behavior of several codes with nonuniformly distributed irregular accesses.

Even though the uniform distributed irregular patterns can closely approximate many nonuniformly distributed patterns, the assumption itself detaches the possible interactions between a program code and the operation of the cache. Namely, an originally closed-loop interaction is approximated by an open-loop model. The assumption about the uniform distributed access patterns limits the analytical model in this paper to non-real-time applications. In real-time applications, a much stronger interaction may exist between the application code and the cache operations.

Reviewer:  Jun Liu Review #: CR134984
Bookmark and Share
 
Performance Analysis And Design Aids (B.8.2 )
 
 
Performance of Systems (C.4 )
 
Would you recommend this review?
yes
no
Other reviews under "Performance Analysis And Design Aids": Date
Circuit design with VHDL
Pedroni V., MIT Press, Cambridge, MA, 2004. Type: Book (9780262162241)
Dec 20 2004
An Algorithm for Optimally Exploiting Spatial and Temporal Locality in Upper Memory Levels
Temam O. IEEE Transactions on Computers 48(2): 150-158, 1999. Type: Article
Aug 1 1999
Understanding Why Correlation Profiling Improves the Predictability of Data Cache Misses in Nonnumeric Applications
Mowry T., Luk C. IEEE Transactions on Computers 49(4): 369-384, 2000. Type: Article
Jun 1 2001
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy