Date: Mon, 7 Oct 1996 10:53:10 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: ache@nagual.ru (=?KOI8-R?Q?=E1=CE=C4=D2=C5=CA_=FE=C5=D2=CE=CF=D7?=) Cc: rkw@dataplex.net, terry@lambert.org, freebsd-hackers@FreeBSD.org Subject: Re: I plan to change random() for -current (was Re: rand() and random()) Message-ID: <199610071753.KAA14386@phaeton.artisoft.com> In-Reply-To: <199610071726.VAA01295@nagual.ru> from "=?KOI8-R?Q?=E1=CE=C4=D2=C5=CA_=FE=C5=D2=CE=CF=D7?=" at Oct 7, 96 09:26:07 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > If YOU want a different function, by all means write one, test it, and use it. > > > > However, DO NOT change the existing functions. PERIOD. > > The function must conforms standard - it is main rule. > What you say is "golden code" syndrome. > > We already change some math functions and nobody complains. A typical run is one million relativistically invariant collisions which pass the physics constraints that have been imposed. If I run experiment #2 to compare results with experiment #1, I will not know that you have perterbed the code. It will not be obvious to me from my results in experiment #2. Thank God you have not yet damaged rand48; the physics calculations I run which depend on rand48 (specifically, drand48) remaining intact take nearly a full week to run, on average. Unless I took another week each time a change occurred (assumming anyone red-flagged the change, like they should have), any change you made would be undetectable without 130+ CPU *hours* of calculation, and a comparison of over *450MB* of output. For obvious reasons of storage, when I run 30 or so of these a year, I can keep only the more recent results online... I do not have 15G of storage available online for this work. Instead, I print out a graphical report, and get some nice hard lines at the known particles, another line at where the carrier of the weak force should be, and various other lines. It is the graphical report which gets kept. Clearly, if you were to damage the generator, you would potentially introduce periodicity which would throw off any attempt to verify the numbers when I wanted to check the actual events for a given report. Regards, Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199610071753.KAA14386>