Return to the
comp.compilers page.
Search the
comp.compilers archives again.
| Related articles |
|---|
| General byte-codes reference mak@imakhno.freeserve.co.uk (Makhno) (2000-12-07) |
| Re: General byte-codes reference s337240@student.uq.edu.au (Trent Waddington) (2000-12-08) |
| Re: General byte-codes reference evilzr@yahoo.com (Daniel Dunbar) (2000-12-08) |
| Re: General byte-codes reference anton@mips.complang.tuwien.ac.at (2000-12-11) |
| Re: General byte-codes reference midkiff@watson.ibm.com (2000-12-11) |
| Re: General byte-codes reference Norman_member@newsguy.com (Norman Culver) (2000-12-18) |
| Re: General byte-codes reference brangdon@cix.compulink.co.uk (2000-12-18) |
| Re: General byte-codes reference patc@acm.org (Pat Caudill) (2000-12-18) |
| Re: General byte-codes reference sjmeyer@www.tdl.com (2000-12-20) |
| Re: General byte-codes reference midkiff@watson.ibm.com (2000-12-21) |
| Re: General byte-codes reference anton@mips.complang.tuwien.ac.at (2000-12-31) |
| From: | midkiff@watson.ibm.com (Sam Midkiff) |
| Newsgroups: | comp.compilers |
| Date: | 11 Dec 2000 02:12:46 -0500 |
| Organization: | AT&T Worldnet |
| References: | 00-12-030 00-12-032 |
| Keywords: | interpreter, performance, comment |
| Posted-Date: | 11 Dec 2000 02:12:46 EST |
Work by David Padua at Illinois and Keshav Pingali at Cornell indicate
that even with Matlab, big wins come from dynamically compiling the
code.
>[The overhead is typically more like 10:1. On systems like Matlab where
>each operator does a great deal of work, e.g., extract eigenvalues, the
>space savings could plausibly win due to less paging. -John]
Sam Midkiff
[Interesting stuff, see Pingali's paper at
http://www.cs.cornell.edu/Info/Projects/Bernoulli/papers/dsl99.ps
which points out that if you compile Matlab into Fortran, you can
host many checks from runtime to compile time. This looks to me
similar to the advantage you get from compiling APL. -John]
Return to the
comp.compilers page.
Search the
comp.compilers archives again.