memory allocation bug in do-bio subroutines
Bert Rubash
Four array declarations over the range 0 to nlev should be changed
to declarations over the range 1 to numc.
In the file /gotm-git/src/extras/bio/iow/bio_iow.F90
a ratio variable is declared as
REALTYPE :: rat(0:nlev,0:nlev)
but is initialized with indices in the 1 to numc range
rat(de,fl)=h(1)
rat(fl,am)=1./h(1)
rat(fl,de)=1./h(1)
and used over the 1 to numc range:
do i=1,numc
do j=1,numc
if (i.ne.j) pp(i,j,ci)=rat(j,i)*dd(j,i,ci)
end do
end do
...
similarly in gotm-git/src/extras/bio/mangan/bio_mangan.F90
REALTYPE :: rat(0:nlev,0:nlev)
but
rat(mn2,mn2_sed)=h(1)/1000.
rat(mno,mno_sed)=h(1)/1000.
rat(spm,spm_sed)=h(1)
rat(mn2_sed,mn2)=1000./h(1)
rat(mno_sed,mno)=1000./h(1)
rat(spm_sed,spm)=1.0/h(1)
and
do i=1,numc
do j=1,numc
pp(i,j,ci)=rat(j,i)*dd(j,i,ci)
end do
end do
end do
Also in gotm-git/src/extras/bio/mab/bio_mab.F90
REALTYPE :: rat(0:nlev,0:nlev)
but
rat(de,fl)=h(1)
rat(fl,am)=1./h(1)
and
do i=1,numc
do j=1,numc
if (i.ne.j) pp(i,j,ci)=rat(j,i)*dd(j,i,ci)
end do
end do
'rat' is declared in gotm-git/src/extras/bio/rolm/bio_rolm.F90
as REALTYPE :: rat(0:nlev,0:nlev)
and initialized but not used.
-Bert Rubash
Jorn Bruggeman
Cheers,
Jorn
Karsten Bolding
Thanks for the error report. I guess nlev has always been bigger than
numc as no wrong results has been reported.
As you might be aware - the support for the old GOTM/BIO will be phased
out in favor of FABM. No new features will be added to GOTM/BIO and the
code will also disappear. Prensently new models are added to FABM and
comparisons between old GOTM/BIO and new FABM implementations are done.
Karsten
Adolf STIPS
On Feb 14, 2012, at 7:56 AM, Karsten Bolding wrote:
> Hi Bert
>
> Thanks for the error report. I guess nlev has always been bigger than
> numc as no wrong results has been reported.
>
Unfortunately this is not correct, remember the respective error report from Gennadi, about 1.5 years ago.
He had exactly this case of using 10 layers in GOTM and bio_iow with fluff (11 variables)
and this failed, using 15 layers made it run.
Anyhow nice that even this case is solved now, ciao Adolf
Karsten Bolding
>
<snip>
>
> Unfortunately this is not correct, remember the respective error report from Gennadi, about 1.5 years ago.
searched in the mail archive - but can't find it
>
<snip>