emacs-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

--with-wide-int


From: Stefan Monnier
Subject: --with-wide-int
Date: Sun, 27 Nov 2011 23:19:43 -0500
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/24.0.91 (gnu/linux)

> Ah, sorry, I should have mentioned that I was thinking in the
> context of the patch for bug#9874, which improves performance on
> 32-bit platforms configured --with-wide int (so EMACS_INT is 64
> bits, ptrdiff_t 32) by shrinking struct Lisp_String as follows:
 
BTW, I'm less and less convinced that --with-wide-int makes sense.
So beside performance (both CPU and memory use) measurements showing
that this mode doesn't significantly suffer from the extra data
manipulation, I'd like to hear from concrete experience using it where
the wide-int was useful.

The reason why I'm skeptical is the following:
while with-wide-int pushes the theoretical buffer limit from 512MB to 4GB,
this only regards the limit imposed by Emacs's Lisp_Object reach,
whereas there are other limits:
- the OS may bring the limit down to 3GB or even 2GB.
- fragmentation is very likely to stop you as soon as you reach about
  half of the maximum, and is quite likely to do it even earlier.
So I expect the practical limit to be closer to 1GB.
While there are probably several mbox files in the "512MB-1GB" range,
I wonder how many users have such files while at the same time not
needing to push the limit even further (e.g. because they also have
other mbox files that are even larger).

And of course, we could also reach the 1GB limit without --with-wide-int
by reworking the tagging scheme (but I suspect the disadvantages would
similarly outweigh the potential benefits).


        Stefan



reply via email to

[Prev in Thread] Current Thread [Next in Thread]