How to use the ncores parameter in secr.fit() function .

[Sachin Vardhan]

Hi, 

I am doing some likelihood analysis in R using the secr package. The boundary I am struct into is that when implementing the logic it is only using one core(R is single-threaded), but there is one parameter in your secr.fit() "ncores=1", when I am not getting it. Is it being used to run the function/analysis on multiple cores or it mean something else. 

Regards 

Sachin Vardhan

[Murray Efford]

The help for secr.fit says

"If ncores > 1 the parallel package will be used to create processes on multiple cores (see Parallel for more). Specifying extra cores may improve the speed of multi-session analyses (it may also slow them down, as data must be copied back and forth). There is presently no benefit for single-session analyses."

See ?Parallel for more

[Sachin Vardhan]

I read that in the documentation. But I am not getting it how to use it. Specifying externally to ncores= 4 , it still uses only one core despite of that. And the point is how to create the multi-session analysis for that. ?

[Murray Efford]

Basically, don't do it. In the current version of secr you won't gain anything unless perhaps you are already fitting a multi-session model. This may change in future. If speed is really an issue then find other ways to speed up the analysis - combining occasions, reducing the number of mask points etc.