[prev in list] [next in list] [prev in thread] [next in thread] 

List:       lyx-devel
Subject:    Re: Computers at the meeting
From:       Kuba Ober <kuba () mareimbrium ! org>
Date:       2003-04-30 21:59:31
[Download RAW message or body]

On wtorek 29 kwiecień 2003 09:07 am, Lars Gullik Bjønnes wrote:
> Kuba Ober <kuba@mareimbrium.org> writes:
> | > > How else might we reduce memory requirements of a compile?
> | >
> | > It's the linking and part of it are long symbol names. I don't think
> | > there can be done much about it apart from the obvious "slash fat from
> | > interfaces"...
> |
> | Partial linking. Partial linking. Partial linking. Partial linking.
> | Par...
> |
> | It works like a charm, although I didn't specifically test it on LyX.
> | Although partial linking doesn't slash the amount of symbols in the final
> | executable's link, but it does slash the number of symbols handled on the
> | way to it. If done intelligently (as far as selection of .o files that go
> | into each .a), it can reduce link time *and* memory usage by about 50%.
> | That's what I'm seeing, even though my code is reasonably small. It's 25k
> | C++ lines, don't laugh at me, but the link times are a killer if it's all
> | done in a single linking, and the linker happily eats about 60mb while
> | linking it all at once (it uses Qt and a few other not-so-small
> | libraries).
>
> I'll try out partial linking again too then, even if we didn't see
> much improvment last time.

Maybe it's somehow specific to what libraries does one use.

As I said, I haven't tested it on anything besides my own project. At least 
theoretically it should help. But as you know, in theory there's no practice 
;-)

Cheers, Kuba Ober
[prev in list] [next in list] [prev in thread] [next in thread]