Re: Proposal: immediate strings

[Dmitry Antipov]

On 05/29/2012 11:38 AM, Paul Eggert wrote:

> On 05/28/2012 11:55 PM, Dmitry Antipov wrote:
> > To avoid using 'packed' attribute and speedup an inefficient access to 
> > u.dat.size
> > field, it's possible to use 'ptrdiff_t X : BITS_PER_PTRDIFF_T - 2'
> > bitfields - it's still wide enough to hold the values up to STRING_BYTES_BOUND.
>
> That's not true on 32-bit hosts configured --with-wide-int.
> On such hosts, a string can contain up to PTRDIFF_MAX - 1 bytes.
> This proposal would lower the limit to PTRDIFF_MAX / 4 bytes.
> In contrast, the proposal in
> <http://lists.gnu.org/archive/html/emacs-devel/2012-05/msg00460.html>
> would merely lower it to PTRDIFF_MAX - 65536 bytes.

OK, this allows

min (MOST_POSITIVE_FIXNUM, (ptrdiff_t) min (SIZE_MAX, PTRDIFF_MAX - 1) - 1)

in all configurations (so PTRDIFF_MAX - 2 for --with-wide-int), eventually
uses _all_ bits by eliminating signed values.

Dmitry

immstr-dev.patch
Description: Text document