R J Cano on Thu, 20 Apr 2017 04:38:22 +0200


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: 4 Gigabytes of RAM are not enough!


Hi Bill!,

Merci,

about malining lists. Sure, no problem.

Cheers,

Remy

2017-04-19 14:13 GMT-05:00, Bill Allombert <Bill.Allombert@math.u-bordeaux.fr>:
> On Wed, Apr 19, 2017 at 05:01:28AM -0400, R J Cano wrote:
>>   Greetings,
>>
>>   This time featuring a problem that seems to be a RAM-memory's pacman
>> monster...
>>
>>   Briefly: If you are looking for arbitrary symmetric/antisymmetric
>> patterns inside the first N!-1 terms of oeis.org A217626, ...for N>6
>> the problem becomes (yet inclusive if having a very simple and
>> efficient script!) intractable due there is no enough RAM to complete
>> the corresponding report...
>>
>>  Attached to this mail comes a PARI/GP script which can be used to
>> appreciate that issue.
>>
>> The usage is simple, load it and set:
>>
>> p=vecDiff(A217626_vec_1(7));
>>
>> then run the following search&report:
>>
>> q=Patterns(p,1,5,1);
>>
>> Which tells Patterns( ) to search for any pattern of size 5, and
>> exclusively (e=1) size 5,
>>
>> For example the pattern [1,9,2,9,1] or the first 3!-1 terms of A217626
>> should be present (7-3+1)! = 120 times; However: The execution crashes
>> before allowing us to verify that!!, giving up due lacking of stack.
>
> The problem is that the output of Patterns is too large to fit
> in the memory, this is not a problem with the algorithm:
>
> ? p=vecDiff(A217626_vec_1(6));
> ? q=Patterns(p,1,5,1);
> ? sizebyte(q)
> %15 = 1998476512
>
> so q is almost 2GB. You need to find a more compact representation for
> the output of Patterns if you want to go farther
>
> Cheers,
> Bill.
>
> PS: please avoid crossposting between pari-dev and pari-users, a lot of
> people are subscribed to both lists and so receive it twice.
>