Re: [eigen] again a unit test precision question |

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]

*To*: eigen@xxxxxxxxxxxxxxxxxxx*Subject*: Re: [eigen] again a unit test precision question*From*: joel falcou <joel.falcou@xxxxxx>*Date*: Sat, 03 Jul 2010 14:02:20 +0200

Jitse Niesen wrote:

On Sat, 3 Jul 2010, Hauke Heibel wrote:Anyways, the redux_8 test is failing every now and then in redux.cpp line 105 (e.g. seed 1278148098). and it is failing with values 6.1214e-005 and 6.10948e-005The difference between these two values is 2e-7. The test adds 33floats between -1 and 1. The error analysis for addition uses absoluteerror, not relative error, so the correct test is whetherabs(sum computed by one method - sum computed by another method) <tolerance.

Could be better to measure error in ULP

**References**:**[eigen] again a unit test precision question***From:*Hauke Heibel

**Re: [eigen] again a unit test precision question***From:*Jitse Niesen

**Messages sorted by:**[ date | thread ]- Prev by Date:
**Re: [eigen] again a unit test precision question** - Next by Date:
**Re: [eigen] any remaining issues blocking beta1, speak up!** - Previous by thread:
**Re: [eigen] again a unit test precision question** - Next by thread:
**Re: [eigen] again a unit test precision question**

Mail converted by MHonArc 2.6.19+ | http://listengine.tuxfamily.org/ |