Amdahl's Law in the Multicore Era

Amdahl's Law in the Multicore Era,
Mark D. Hill and Michael R. Marty,
IEEE Computer, July 2008.
Paper: pdf

2017 Retrospective: pdf
YouTube Video (52 minutes): Google TechTalk 02/2009
Related Talks: HPCA Keynote 02/2008 and CS Colloquium
Original Technical Report (UW CS-TR-2007-1593, April 2007): pdf
Gene Amdahl and Michael Marty at Wisconsin CS Reception in Bay Area, March 2008: jpg photo

Matlab files used for paper graphs:

scmp.m, acmp.m, dcmp.m

Interactive Grapher:

Number of BCEs (N):

Line 1: fraction parallel (f): perf(x) = Symmetric Asymmetric Dynamic

Line 2: fraction parallel (f): perf(x) = Symmetric Asymmetric Dynamic

Line 3: fraction parallel (f): perf(x) = Symmetric Asymmetric Dynamic

Line 4: fraction parallel (f): perf(x) = Symmetric Asymmetric Dynamic

Note: x is the number of BCEs harnessed for faster core(s) (was r in the paper)
Note: Pressing ReDraw when re-enabling disabled lines can be REALLY slow and temporarily tie up the browser



For defining perf(x), you can use the standard operators
+ - * / ( )
the constants of the JavaScript Math object
E LN2 LN10 LOG2E LOG10E PI SQRT1_2 SQRT2
and the functions of the Math object
abs acos asin atan ceil cos exp floor log max min pow random round sin sqrt tan
to define a function.