| To: | scsh@martigny.ai.mit.edu |
|---|---|
| Subject: | Re: cons-streams and SCSH |
| From: | shivers@ai.mit.edu (Olin Shivers) |
| Date: | 22 Aug 1996 12:23:00 -0400 |
| Organization: | Artificial Intelligence Lab, MIT |
| Reply-to: | shivers@ai.mit.edu |
It's a limitation of S48's GC. I just received the following report from
Kelsey.
-Olin
From: kelsey@research.nj.nec.com (Richard Kelsey)
Yes. Because the byte-code compiler doesn't do any free-variable
analysis, every closure contains its complete lexical environment.
For the first definition of SEARCH-STREAM the LP closure contains
a pointer to the original lazy-list S, causing the heap to blow out.
For Kali I wrote an additional compiler pass that does the necessary
analysis and produces minimal closures. The downside is that loading
code takes over 50% longer.
|
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: cons-streams and SCSH, Sean Doran |
|---|---|
| Next by Date: | Re: cons-streams and SCSH, Sean Doran |
| Previous by Thread: | Re: cons-streams and SCSH, Sean Doran |
| Next by Thread: | Re: cons-streams and SCSH, Sean Doran |
| Indexes: | [Date] [Thread] [Top] [All Lists] |