Re: Code for converting between Elisp and Calc floats

[Stefan Monnier <monnier@iro.umontreal.ca>]

> There are few areas of importance.  One would be buffer_size.  Putting

bignums for buffer size is a bad idea.  It would require an enormous
amount of code changes and a very significant slow down (buffer
positions are used extensively everywhere in the C code, including
performance critical code such as the rendering engine, and changing
them from scalar values to something else would slow things down
tremendously).

> The buffer size problem would likely be less radically addressed by
> adding another bit of tagging, and then use half of the possible
> values for integers, giving a range of -2^30 .. 2^30-1.

Currently we have -2^28..2^28-1 and we can easily extend it to
2^29..2^29-1 without any noticeable downside (I have been using such
a change for a while now).

> That's what XEmacs does, and it goes a far way of extending the point
> of nuisance.

Not far enough, AFAICT: if you regularl edit files somewhere between
512M and 1G, then it's quite likely that you'll also want to edit files
larger than 1G.  Currently in 64bit mode, you supposedly can edit files
larger but in practice when I tried it, I bumped into other bugs when
reaching the 2G limit (or maybe even earlier) IIRC.


        Stefan