MrBayes
MrBayes copied to clipboard
NBISweden
Calculation hangs on model with partition
Test example contains aligned 56 chloroplast sequences downloaded from NCBI. There are 60 partitions (genes and concatenated inter gene regions). Running that calculations hangs around iteration 6000. Hanging means logging stops and CPU is on 100%.
Nexus file specifies seeds, so it is reproducible.
When calculation is done without partitions everything works.
I created this data with my script, and I suppose that I created it somehow wrong, but I don't know what to check more. So I am more asking for advise what to check in my data.
Things I checked:
-
created few variants of data and some hangs sooner, some later.
-
run calculation on my laptop (Intel i7) and on local server (Xeon). Both runs Ubuntu, first 20.04, second 18.04. Both runs MrBayes version 3.2.7a.
-
run developer version (https://github.com/NBISweden/MrBayes/archive/develop.zip). It hangs even sooner, on iteration 3500.
-
run current GitHub version (967f7832 - Oct 20). Same as developer, hangs on iteration 3500.
-
compiled with configuration: CFLAGS="-DDEBUG_FBDPR -DDEBUG_CONSTRAINTS -DDEBUG_NOSHORTCUTS -DDEBUG_LIKELIHOOD -DDEBUG_LNLIKELIHOOD -DDEBUG_SETSTDQMATRIX -DDEBUG_TIPROBS_STD -DDEBUG_MB_BEAGLE_FLOW" ./configure and commenting lines "#undef DEBUG_*" in mcmc.c. No check with an error was displayed.
-
same things happens with MPI version. That is how I found it since I started lot of runs and detected that some of them hung.
Thank you orig_2_2chains.zip
I debugged problem and it seems that execution hangs in function Move_RateMult_Slider() (proposal.c), in while loop in line 14616 (git version): https://github.com/NBISweden/MrBayes/blob/develop/src/proposal.c#L14616
Variables min and max contain the same value of 0.5. With that it is not possible to find y between them. I didn't notice suspicious value(s) in arrays value and subValue.
