search

glenco / SLsimLib

Library for Gravitational Lensing Simulations
MIT License
2 stars 1 forks source link

Field Sim of Millennium NFW + NSIE galaxie: Concave Hull takes too long #82

Closed leier closed 8 years ago

leier commented 8 years ago

@rbmetcalf

For the current master SHA: bba416d a program that tries to calculate the caustics runs too long (>15 min - i.e. maybe forever - for a small area around the largest halo). The revision before the rbm/nsie-normalization merger works fine, so I searched for the first commit that introduced this problem: it's SHA: d997be7 .

My guess is: the changes in lines 348-355 in lens_halos.cpp are responsible for this behaviour.

leier commented 8 years ago

All Rsize in https://github.com/glenco/SLsimLib/blob/master/MultiPlane/lens_halos.cpp#L352-355 must be Rmax.

Rmax is private variable of the LensHalo base class and is at this point filled with Rmax from https://github.com/glenco/CosmoLib/blob/master/nfw.cpp#L141 which was as I see right now renamed to Rsize ...

So apparently variable names in CosmoLib were changed too ...

this is very problematic for everyone who runs CosmoLib with an older version of SLsimLib or an older version of CosmoLib with are new master SLsimLib.

leier commented 8 years ago

Changing Rsize to Rmax in https://github.com/glenco/SLsimLib/blob/master/MultiPlane/lens_halos.cpp#L352-355 is necessary: in fact the Rmax is not equal to 1.2*Rsize. Rmax is used as we go through the empty constructor of LensHaloNFW and isotropic halos are assumed, so the distinction doesn't make sense here. In view of this, the renaming of the CosmoLib function wasn't really necessary. I created a new branch and will do a pull-request right away.