Download presentation
Presentation is loading. Please wait.
Published byHannah Brooks Modified over 9 years ago
1
Ten Pounds in a Five-Pound Sack Zac Lippard
2
Program Space as Cost How big is it? Space occupied is a principle cost! Is it worth paying the price for space, or is it better to use the money elsewhere? Eliminate unnecessary size expenditures by setting size targets, control sizes, and devise size-reduction techniques.
3
Size Control Size control is partly technical, partly managerial. Separate the system into subsystem components and set control sizes for each. The wise program manager prepares a kitty for later allocation.
4
OS/360 Setting size targets for core is not enough. Relied on cheap disk access and therefore slowed down performance. Even though core targets were met by the programmers, execution was incredible slow.
5
Whats the Lesson? Set total size budgets as well as resident- size budgets so as to not run over the amount of space you have available. Define specifically what a module must do when you set the size for it. Most important lesson: Foster a total- system, user-oriented attitude! Don’t go over the top!
6
Small Programs No space budgeting or controlling can do this. It takes invention and craftsmanship. Trade functions for size. It’s cheaper to by the package then everything individually. Must decide how fine-grained the program is to be.
7
Space-Time Trade-Offs More space, the faster the software. Managers can do two things to ensure this happens: Make sure the programmers are trained well. Use programming technology to functional preference Two sets of functions: the quick and the squeezed.
8
Invention Strategic breakthroughs in programming. Find code that creates a faster program. Cooley-Tukey Fast Fourier Transform Reduced computation time by using n 2 instead of n log n Representation of data and tables can help in further breakthroughs.
9
The End
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.