DOC PREVIEW
Pitt CS 3150 - Energy Corollaries

This preview shows page 1-2-14-15-30-31 out of 31 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 31 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Sangyeun Cho, Rami Melhem, Michael MoengUniversity of PittsburghPresented By: Michael Moeng Amdahl’s Law Power Applications Problem Formulation Derivations Conclusions Maximum speedup for a program given◦ Sequential fraction s, parallel fraction p=1-s◦ N processors Speedup =  Assumes parallel portion can be perfectly parallelized Keep execution time constant and improve dynamic energy Assumes p ~ fα Find optimal serial and parallel frequency by taking derivative Keep execution time constant and improve energy Now we set speedup to x  Set fs, fp What are optimal serial and parallel frequencies? Now we set speedup to x  Set fs, fp What are optimal serial and parallel frequencies? Trivial – minimize frequencies Must consider static power! Serial work scompletes in time t Parallel work p=1-s completes in time 1/x-t◦ Recall that x is speedup, so 1/x is total time Energy is composed of ◦ Serial dynamic energy = ◦ Parallel dynamic energy = ◦ Static energy = First consider special case, x=1 (same execution time as serial case) Take derivative with respect to t First consider special case, x=1 (same execution time as serial case) Take derivative with respect tot  First consider special case, x=1 (same execution time as serial case) Take derivative with respect to t Relation between fs*and fp*is dependent on N, but not s! Note that static energy has no effect in this casefp=Fmaxfs=Fmax Relax execution time constraint◦ Set derivatives with respect to both xand tto 0◦ Solve equations Relax execution time constraint◦ Set derivatives with respect to both xand tto 0◦ Solve equations Relax execution time constraint◦ Set derivatives with respect to both xand tto 0◦ Solve equations Relax execution time constraint◦ Set derivatives with respect to both xand tto 0◦ Solve equations Relax execution time constraint◦ Set derivatives with respect to both xand tto 0◦ Solve equations Relax execution time constraint◦ Set derivatives with respect to both xand tto 0◦ Solve equations Previous solution necessitates λN ≤ α-1◦fs= Fmax Energy Delay Energy Delay Energy Delay Same relation between fs*and fp* Implications? Synchronization cost◦ Parallel architectures have some overhead for communication◦ As N increases, more total work is required, which can be expressed as a function◦ Total work is now s + p(1 + σ(N)) Synchronization cost◦ Parallel architectures have some overhead for communication◦ As N increases, more total work is required, which can be expressed as a function◦ Total work is now s + (p + σ(N)) Optimal serial and parallel frequencies have the same relationship◦ Still differ by N1/αMark D. Hill, University of Wisconsin-Madison andMichael R. Marty, GooglePresented By: Michael Moeng Chip has a total resource budget, defined in the number of ‘baseline’ cores that can be supported Assume performance for an r-BCE core grows by


View Full Document

Pitt CS 3150 - Energy Corollaries

Documents in this Course
JouleSort

JouleSort

12 pages

Load more
Download Energy Corollaries
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Energy Corollaries and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Energy Corollaries 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?