Idea: Use of gethash/puthash for GNU Emacs 21?

Idea: Use of gethash/puthash for GNU Emacs 21?

[Christoph Conrad]

Hello,

Gnus uses gnus-gethash/gnus-sethash, which in fact uses obarrays
(hashes). I don't know if there is a significant speedup of using
native gethash/puthash in the case of big hash tables and GNU Emacs
Version greater/equal 21. At least there should be a significant
memory saving, as far as i remember early discussions about them.

When i get time in the next days i will do some performance
comparisons.

Best regards,
cu, -cc-
-- 
=> GNU Emacs Webring @ <http://www.gnusoftware.com/WebRing/> <=
Look Ma, this man can twist his fingers as if they were made of rubber,
isn't that amazing? -- Not really, he's been using emacs for years...!

Re: Idea: Use of gethash/puthash for GNU Emacs 21?

[Christoph Conrad]

    > When i get time in the next days i will do some performance
    > comparisons.

Ok, done so. Used Lars Magne Ingebrigtsen's benchmark.el (from
quimby.gnus.org).

(benchmark 10000 (gnus-sethash (int-to-string (random 1000))
                               (random 1000)
                               hashtable))

(benchmark 10000 (gnus-gethash (int-to-string (random 1000))
                               hashtable))


the same for puthash/gethash (with makehash called without args). I
made garbage collections before each measurement. gc-cons-threshold's
value is 3000000 and the repeat count to benchmark only 10000 to
prevent any garbage collections.

The result is astonishing: performance is nearly identical for this
_very_ limited testset.

Best regards,
cu, -cc-
-- 
=> GNU Emacs Webring @ <http://www.gnusoftware.com/WebRing/> <=
Look Ma, this man can twist his fingers as if they were made of rubber,
isn't that amazing? -- Not really, he's been using emacs for years...!

Re: Idea: Use of gethash/puthash for GNU Emacs 21?

[Lars Magne Ingebrigtsen]

Christoph Conrad <christoph.conrad@gmx.de> writes:

[puthash vs. gnus-sethash]

> The result is astonishing: performance is nearly identical for this
> _very_ limited testset.

I would have thought the only difference would be in the memory usage
-- except for garbage collecting (and swapping :-), gnus-sethash
should be pretty optimal.  I mean, it just uses `intern', and since
one of the most central parts of Emacs is symbol lookup, it's bound to
be pretty optimized...

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Idea: Use of gethash/puthash for GNU Emacs 21?

[Dave Love]

>>>>> "LMI" == Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

 LMI> I mean, it just uses `intern', and since one of the most central
 LMI> parts of Emacs is symbol lookup, it's bound to be pretty
 LMI> optimized...

The performance of interning probably has room for improvement, though
as far as I remember, the ...hash functions use the same algorithm and
parameters.

The salient feature of hash tables in Gnus is that various code
depends on them being obarrays, since it uses `read' to intern into
them.  That said, the Gnus functions should still be used
consistently, rather than explicitly creating obarrays in various
places.