[self-interest] Re: Self ported to Linux

Ian Woollard ian.woollard at tesco.net
Thu Jan 31 00:09:23 UTC 2002

I'm not sure how it would pan out, I do know that Self 4.0 on
a sparc had the option of storing the compiled code cache to
disk together with the snapshot- I turned it off because the
file it generated was huge IRC.

Modern disks are faster of course, and MUCH bigger.

Still, it would be interesting to know if there are any clues 
that the compiler could write to disk that would help it next 
time around; that would also improve performance. Can the code 
cache be compressed in any way?

Jecel Assumpcao Jr wrote:

> On Tuesday 29 January 2002 20:30, Samantha Atkins wrote:
>>Jecel Assumpcao Jr wrote:
>>>[Self 2.0 compiler features]
>>The really nice option would be to have the uber-compiler
>>available for streamlining production code but not turned on
>>during normal development cycles.
> You would think so. After all, the runtime compilers probably make lots 
> of bad decisions due to the need to get the job done in a fraction of a 
> second. With minutes or hours to work on the sources (and let's be 
> generous and add some profiling information gathered from running the 
> code under typical conditions) it seems obvious that the results would 
> be significantly better.
> I am not sure that this would be the case in practice, however. The 
> uber-compiler might have to generate a very large executable. 
> Applications normally go through several "execution phases" with very 
> different characteristics and the generated code would have to deal 
> with all of them. Given that processor speed has been improving faster 
> than disk speed, it is possible that the time spent loading super 
> optimized code from the disk (probably due to swapping in the virtual 
> memory) could negate any gains relative to compiling and running lower 
> quality code with the NIC/SIC.
> This seems like an interesting research question. The answer probably 
> varies from application to application.
> -- Jecel
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ 

- Ian Woollard (ian.woollard at tesco.net)

"Is a planetary surface the right place for an expanding 
technological civilization?"
- Gerard O'Neill

More information about the Self-interest mailing list