Front page | perl.perl5.porters |
Postings from November 2012
On 1 November 2012 09:50, bulk 88 <bulk88@hotmail.com> wrote: > > > > ---------------------------------------- >> Date: Thu, 1 Nov 2012 09:33:41 +0100 >> Subject: Re: Eliminating the "rehash" mechanism for 5.18 >> From: demerphq@gmail.com >> To: bulk88@hotmail.com >> >> Did you mean this to be offlist? > > No. That was an accident. > >> >> On 1 November 2012 09:09, bulk 88 <bulk88@hotmail.com> wrote: >> > >> > >> > >> >> There are multiple objectives. I had hoped to see a performance boost. > > I just hope the replacement implementation will not be a performance decrease. Obviously me too. >> As far as I can tell that would still work. In fact it seems to me >> like things would be better to some degree, as there would be no more >> "two hashes" and the difference between PERL_HASH() and >> PERL_HASH_INTERNAL_() would be eliminated (and actually >> PERL_HASH_INTERNAL_() just goes away). > > The difference between the 2 macros is just where the hash key is > obtained from, > "U32 hash_PeRlHaSh = (internal ? PL_rehash_seed : PL_hash_seed);" which > is constant folded away for PERL_HASH to PL_hash_seed. And in the new implementation it becomes U32 hash_PeRlHaSh = PERL_HASH ^ len; (the len there is provisional) >> The thing is the layout would be different process to process, but not >> internally to a process. > > That is fine for me. My unpublished code studies and remembers the > positions in the array and linked list of the hash elements at runtime. > >> >> On the other hand if we ever introduced per-hash hashing, (as a >> security measure) then it would break. Or we would have to give you a >> way to control it, or something. I'm hoping we don't need per-hash >> hashing if we have random hash seeding. > > I could imagine restricted hashes having a per restricted hash hash > that guarentees no keys collide and that all those keys are accessed > through array alone, no linked list. Or maybe a future linked-list-less > hash api where reads and replace are cheaper than current hash api > but inserts are very costly. Interesting ideas. Yves -- perl -Mre=debug -e "/just|another|perl|hacker/"Thread Previous