Re: GiST penalty functions [PoC] - Mailing list pgsql-hackers

From Andrew Borodin
Subject Re: GiST penalty functions [PoC]
Date
Msg-id CAJEAwVFgUnE+hk8P0MCUmuzm05731jg2ZeqTA5stPPYUFQ8w-g@mail.gmail.com
Whole thread Raw
In response to Re: GiST penalty functions [PoC]  (Greg Stark <stark@mit.edu>)
List pgsql-hackers
>Personally I wouldn't be very happy about an IEEE754 assumption.
Ok, so let's avoid autoconf man. There is no real explanations of the
ground for this assumption, just a reference to paper of David
Goldberg (and there is no metion about IEEE754 is absoulte
everywhere). BTW, may be we can ask David Goldberg how to hack that
float properly? I do  not see any math sense in this:

pack_float(float actualValue, int realm)
{
... int realmAjustment = *((int*)&actualValue)/4; float something = *((float*)&realmAdjustment)
...
}
No wonder it's not in libs.
Nither I see a way to implement it with ldexp siblings.
>I did go to the trouble of testing Postgres on a VAX and we fixed the few
instances where it had such dependencies for a net gain.
Greg, could you please point out those places to see how exaclty it
should be #ifdef'd?

Regrads, Andrey Borodin.



pgsql-hackers by date:

Previous
From: Peter Geoghegan
Date:
Subject: Re: Parallel tuplesort (for parallel B-Tree index creation)
Next
From: Andres Freund
Date:
Subject: Re: pg_sequence catalog