McStas PSD_Detector (T v Wuure) gas-detector has "undefined behaviour" in McStas 3.6.1 #2267
Description
Looking at the results table from 3.6.1 I spotted numeric fluctuations / diverging output in HZB_NEAT testcase 1.
Experimenting a little with an Intel Ubuntu machine as well as my M2 MacBook Pro I found erratic output, depending on run/os conditions. All output earlier than the last PSD_Detector is identical.
macOS M2:
mcrun -c HZB_NEAT.instr -s1000 lambda=6 | grep Detector_I
Detector: Detector_I=2904.46 Detector_ERR=36.487 Detector_N=19860 "NEAT.psd"
macOS M2 mpi:
mcrun -c HZB_NEAT.instr -s1000 lambda=6 --mpi=2 | grep Detector_I
Detector: Detector_I=2871.53 Detector_ERR=36.3542 Detector_N=19698 "NEAT.psd"
Ubuntu 24.04.1 (Debian package):
mcrun -c HZB_NEAT.instr -s1000 lambda=6 | grep Detector_I
Detector: Detector_I=527.293 Detector_ERR=15.5595 Detector_N=3629 "NEAT.psd"
Ubuntu 24.04.1 (Debian package) mpi:
mcrun -c HZB_NEAT.instr -s1000 lambda=6 --mpi=2 | grep Detector_I
Detector: Detector_I=2871.53 Detector_ERR=36.3542 Detector_N=19698 "NEAT.psd"
Ubuntu 24.04.1 (conda package):
mcrun -c HZB_NEAT.instr -s1000 lambda=6 | grep Detector_I
Detector: Detector_I=527.293 Detector_ERR=15.5595 Detector_N=3629 "NEAT.psd"
Ubuntu 24.04.1 (conda package) mpi:
mcrun -c HZB_NEAT.instr -s1000 lambda=6 --mpi=2 | grep Detector_I
Detector: Detector_I=529.904 Detector_ERR=15.6395 Detector_N=3663 "NEAT.psd"
Investigations ongoing, observations:
- no warnings / hints from cppcheck...
- It seems that the same effect can be observed in ILL_IN4 and ILL_IN5 - but with fluctuations of lower magnitude...
- Interestingly, Test_PSD_Detector does not show anything... Plot thickens...
- Memory-issue quite probable... Exchanging Isotropic_Sqw for Incoherent, use of an MCPL_ouput to dump the beam all makes the issue (seemingly) go away...