Q: least-square fit for performance model

Hi,


Is there any literature discussing how suitable it is to use
least-square fitting to get the performance model of a set of
benchmarks? What I read so far is that: given a set of benchmarks,
decide what the performance factors there are, and then use the
least-square fitting to get the coefficient of each performance
factor. This model is then used to predict other
programs/benchmarks. Is there any other way to get the performance
model, besides the one which estimates the cost(time) of each
primitive operation, and count the number of that operation in the
program to predict the performance.


Chung-Hsing Hsu


chunghsu@paul.rutgers.edu
--

Return to the comp.compilers page.
Search the comp.compilers archives again.