GammaLib

I believe the above has implications for the following CTools: ctbin, ctbkgcube, ctedispcube, ctexpcube, ctmodel, ctpsfcube, ctskymap

I’ve implemented some code to do a check on each observation inside a supplied XML observation list overlaps the sky map to be generated. The code that does this is located in 'GSkyMap::overlaps(const GSkyRegionCircle&)'. The check is done by first testing whether the central pointing position of the observation falls within the bounds of the sky map. If this is false, then pixel positions that border the sky map are tested for whether or not they fall into the observation region. The positions tested can be visualized as follows (for a 4x3 bin skymap), where '*' marks the positions tested.

     *   *   *   *   *   *
       +---+---+---+---+
     * |0,2|1,2|2,2|3,2| *
       +---+---+---+---+
     * |0,1|1,1|2,1|3,1| *
       +---+---+---+---+
     * |0,0|1,0|2,0|3,0| *
       +---+---+---+---+
     *   *   *   *   *   *

Quickly testing the performance for ctbin and ctbkgcube using the latest 1dc data give the following:

ctbin debug=yes inobs=../1dc.south/obs/obs_gps_baseline.xml nxpix=400 nypix=400 binsz=0.02 xref=0.0 yref=0.0 coordsys=GAL proj=CAR emin=0.1 emax=100 enumbins=30 ebinalg=LOG outcube=ctbin_test.fits

devel:
* 1011.93 seconds of CPU time
* Number of observed events .: 5530542
1980-skymap_overlap:
* 199.404 seconds of CPU time
* Number of observed events .: 5530542

ctbkgcube

devel:
* 7475.55 seconds of CPU time
1980-skymap_overlap:
* 2320.61 seconds of CPU time

The relevant branches are (note this affects both gammalib and ctools):
joshcardenzana/gammalib: 1980-skymap_overlap
joshcardenzana/ctools: 1980-skymap_overlap

P.S.
I also found a bug where GCTAEventList::dispose() was not actually freeing the memory when it cleared its m_events vector. I’ve also fixed this which dramatically reduces the amount of memory used by some of the ctools. The run of ctbin using the devel branch above used more than 30 GB of memory. After the changes, the same uses less than 1 GB.

I’m about to merge in the code.

I did some small modifications to the code since you introduced multiple return statements at various places. Each method should have a single exit point (see the coding conventions), so I modified the code in a way that the code always reached the return statement at the end.

Merged into devel.