Proebsting's Law

James Davis jamesd at
Wed Dec 14 13:46:36 GMT 2005

Simon Wistow wrote:

> Put it this way - since Moore's law has doubled computing power
> 9roughly) every 18 months (or roughly a million times in 30 years) that
> means the speed up between a a naive 35 year old compiler and gcc with
> no optimisations would need to be 250,000 fold (if I'm doing my maths
> right, I'm probably not) to make a significant difference.
> How much optimisation can you do to the basic compiler cycle. Have there 
> been amazing new innovations in Tokenising?

No, am not arguing that the improvement in speed is significant compared
to the increase in speed through hardware

This doesn't mean people should stop researching compiler design because
hardware gets faster. Would you really be happy using the current
version of your compiler in 2018 let alone 2043? With changes in
architecture it's foolish to presume that no unforeseen and significant,
optimizations are yet to be found or even that a better compiler model
becomes standard. The whole argument presumes that our current state of
technology is 'good enough' and will never be improved upon.

Perhaps the inspiration for a better model comes to a researcher in the
middle of working on the most insignificant optimization? Who knows. The
value of research isn't always purely the end product.

Similar arguments come up frequently, why do people use supercomputers
when they could just wait ten years for a fast enough workstation to
solve the problem. It's because problems need solving now.

Why do researchers work on better parallel algorithms especially since
there'll be faster hardware out by the time they complete their
research. It's because this interests them and not hardware design.

>>I know compilers go back even further than 35 years ago, that was my
>>point. Why pick 35 years?  
> Because, according to the 
> A0 compiler was written in 1952 and this law was first posited in 
> 1997-1998.

Still confused as to the choice of 35 years, that's 45 :-)


-- - Free hosting for charities

More information about the mailing list