Staged Tuning: A Hybrid (Compile/Install-time) Technique for Improving Utilization of Performance-asymmetric Multicores

Emerging trends towards performance-asymmetric multicore pro-
cessors (AMPs) are posing new challenges, because for effective
utilization of AMPs, code sections of a program must be assigned
to cores such that the resource needs of the code sections closely
match  the  resources  available  at  the  assigned  core.  Computing
this assignment can be difficult especially in the presence of un-
known or many target AMPs. We observe that finding a mapping
between the code segment characteristics and the core character-
istics is inexpensive enough, compared to finding a mapping be-
tween the code segments and the cores, that it can be deferred un-
til installation-time for more precise decision. We present staged
tuning which combines extensive compile time  analysis  with in-
telligent binary customization at install-time. Staged tuning is like
staged compilation, just for core assignment. Our evaluation shows
that staged tuning is effective in improving the utilization of AMPs.
We see a 23% speedup over untuned workloads.

 
.....
Neural Network Training
We  use  the  FANN  library  [32]  for constructing and training our neural networks. In our experiments,we  compute  a  grouping  (i.e.,  core  assignment)  for  each  individual benchmark.

Staged Tuning: A Hybrid (Compile/Install-time) Technique for Improving Utilization of Performance-asymmetric Multicores

tyler Sondag ( Intel Labs ) Hridesh Rajan ( Iowa State University )