[Date Index] [Thread Index] [Author Index]

Re: Prime[x] crashes kernel for large x

• To: mathgroup at smc.vnet.net
• Subject: [mg7870] Re: [mg7762] Prime[x] crashes kernel for large x
• From: sergio at scisun.sci.ccny.cuny.edu (Sergio Rojas)
• Date: Thu, 17 Jul 1997 15:35:35 -0400
• Organization: City College Of New York - Science
• Sender: owner-wri-mathgroup at wolfram.com

Does any one know if this kind of problem  has relation with the
" flaw in the precision of the floating point divide operation of the 
Pentium processor"  described in the Intel FAQ found in: 

http://support.intel.com/oem_developer/microprocessors/pentium/PPFPU.HTM
http://support.intel.com/enduser_reseller/pentium_processors/10191.HTM

Rojas

E-mail: sergio at scisun.sci.ccny.cuny.edu

Richard Finley (trfin at fiona.umsmed.edu) wrote:
: I get the same problem with Mma 3.0 (on a Pentium Pro 200 Win95 system) .. I
: have to restart the kernel before proceeding.  One other curiosity, I was
: trying to find out approximately what size number would cause it to crash
: and I noted that 2^26 was OK but 2^27 consistently crashes.  I started
: walking up from 2^26 + 40000000 and found that if I go from that to
: 2^26+50000000 it crashes, but if I perform consecutive trials from 40000000
: in steps of 1000000 I can continue to at least 150000000 (which is greater
: than 2^27) without crashing (that's where I ran out of patience...I don't
: know how far it would have continued.) Why computing Prime[n+k] should
: depend on whether you computed Prime[n] in the previous calculation is
: beyond my understanding of Mma but it is certainly true.  For example, if I
: compute Prime[2^26 + 40000000]//Timing  , it takes 8.47 sec the first time
: you do it, but if you compute it again it gives 0. sec.  It appears it
: "remembers" previous calculations and I presume it somehow uses the results
: in computations of larger primes...so Prime[2^26 + 50000000] crashes only if
: you didn't perform Prime[2^26 + 49000000] etc..first. Do others get the same
: result on other platforms? Any ideas from anyone ? -- Wolfram ?
: RF

: At 02:36 PM 7/8/97 -0400, you wrote:
: >
: >When I try to find a large prime, for example:
: >
: >Prime[2^32 - 1]
: >
: >The math kernel crashes with a divide error, and I can't restart it 
: >without quitting mathematica (3.0) and restarting.  Am I doing something 
: >wrong, or is this a bug? 
: >
: >Thanks,
: >
: >Michael
: >michael at qni.com
: >
: >
: >
: >
: >
: >