Ticket #1339 (assigned defect)

Opened 1 year ago

Last modified 5 months ago

Solaris: number of partitions broken

Reported by: mabshoff Assigned to: mabshoff (accepted)
Priority: major Milestone: sage-3.4
Component: combinatorics Keywords:
Cc:

Description

Hello,

number of partitions is broken on Solaris/Sparc: 

sage -t  devel/sage-main/sage/combinat/combinat.py          **********************************************************************
File "combinat.py", line 1869:
    sage: number_of_partitions(100000)
Expected:
    274935105697756965126775163209863526881734293159800547582031259843021473281149641730550507416607366215901578447742962489
4049306307020046179276449303351011607934245719015571894350972531246610845200636955893446424871682878983218234500926285383140
4597021307130674510624419227311238999702284408609370935531629697851569569892196108480158600569421098519
Got:
    274935105697756965126775163209863526881734293159800547582031259843021473281149641730550507416607366215901578447742962489
4049306307020046179276449303351011607934245719015571894350972531246610845200636955893446424871682878983218234500926285383140
4597021307130674510624419227311238999702284408609370935531629697851569569892196108480158600582558780007
**********************************************************************
File "combinat.py", line 1896:

but also more worringly the "small" case: 

**********************************************************************
File "combinat.py", line 1924:
    sage: len([n for n in [1..500] if number_of_partitions(n) != number_of_partitions(n,algorithm='pari')])
Expected:
    0
Got:
    245
**********************************************************************

Cheers,

Michael

Change History

12/25/2007 07:15:40 PM changed by mabshoff

I suspect floating point precision issues. Didn't we do something on MaxOSX/PPC about this by lowering certain precision bounds?

Cheers,

Michael

07/29/2008 10:42:19 AM changed by mabshoff

  • owner changed from mhansen to mabshoff.
  • status changed from new to assigned.

The solution here is to reduce the various precisions used in the partition code. For qd we have to reduce the precision to *170* bits which indicates either a severe bug on our end or something is seriously wrong in qd.

Cheers,

Michael