Related articles |
---|
Compiler-generated file access hints tuahanj@cs.curtin.edu.au (Nor Jaidi) (1997-04-02) |
Re: Compiler-generated file access hints monnier+comp.compilers@tequila.cs.yale.edu (Stefan Monnier) (1997-04-03) |
Re: Compiler-generated file access hints jaidi@cs.curtin.edu.au (Nor Jaidi) (1997-04-06) |
From: | Nor Jaidi <jaidi@cs.curtin.edu.au> |
Newsgroups: | comp.compilers |
Date: | 6 Apr 1997 22:26:42 -0400 |
Organization: | Curtin University of Technology |
References: | 97-04-015 97-04-032 |
Keywords: | optimize |
Stefan Monnier wrote:
> No special string or file knowledge wouldn't be necessary any more in
> the compiler. Of course, then you'll also have to make sure this
> skeleton program is run "a little in advance".
I like your idea and will put it in my list of things to squeeze my
brain with. But it is still a forest to me. Running "a little in
advance" means (as I understand it) interlacing the skeleton code with
the "normal" code and making sure that the prefetch version of file
access are placed as far back as possible from the normal version to
allow the prefetch to complete. We have to scan the source backward,
scan for dependencies backward (possibly into function calls that
might change vars that store filename, seek position, etc) and
backpatch the object tree. Hei, someone please tell me: do you see
any neat way to do this? Multi-pass? That's not neat.
--
Nor Jaidi Tuah
Research student
Curtin University
Western Australia
tuahanj@cs.curtin.edu.au
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.