You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I’m getting a crash and the error “SIGFPE floating-point exception - erroneous arithmetic operation” when running with Medlyn stomatal conductance and DEBUG=TRUE, on Perlmutter CPUs, gnu compiler (global 4x5 run). With DEBUG=FALSE it runs but I get a lot of warnings “ FATESWARN: 1 m: Stomatal conductance error check - weak convergence: 0.168128E+06 0.159592E+06 “
The error with DEBUG=TRUE is being triggered here:
I guess it depends... If it is a tiny negative number, then I think it makes sense to assume this is just truncation error and set q=0. If not, then I think the model should crash.
Per the software meeting today, @JessicaNeedham noted that the sqrt number is on the order of -1e-7, so this isn't numerical precision.
We should add a check inside the function to make sure that we check b*b-4*a*c and output diagnostic warnings (i.e. values of a, b, and c), but from discussion we need to diagnose why we're getting a combo of these values in upstream code.
I’m getting a crash and the error “SIGFPE floating-point exception - erroneous arithmetic operation” when running with Medlyn stomatal conductance and DEBUG=TRUE, on Perlmutter CPUs, gnu compiler (global 4x5 run). With DEBUG=FALSE it runs but I get a lot of warnings “ FATESWARN: 1 m: Stomatal conductance error check - weak convergence: 0.168128E+06 0.159592E+06 “
The error with DEBUG=TRUE is being triggered here:
fates/biogeophys/FatesPlantRespPhotosynthMod.F90
Line 1812 in 44ba97a
because
( b*b - 4._r8*a*c)
is negative and we’re asking for a square root.My parameter values are:
fates_leaf_stomatal_intercept = 10000, 10000, 10000, 10000, 10000, 10000,
10000, 10000, 10000, 10000, 10000, 40000 ;
fates_leaf_stomatal_slope_medlyn = 4.12, 2.35, 2.35, 4.12, 4.45, 4.45,
4.7, 4.7, 4.7, 2.22, 5.25, 1.62 ;
Can we wrap it in a max(0, ) term or should the model crash if this term becomes negative?
The text was updated successfully, but these errors were encountered: