For adjoint simulation, NSOURCES hardcoded constraint <=1000 in initialize_simulations.f90

[amir-allam]

For adjoint simulations (SIMULATION_TYPE=3), a CMTSOLUTION file with more than 1000 sources will cause the following error:
for adjoint simulations, NSOURCES <= 1000
Error detected, aborting MPI... proc 66

This comes from line 399 in initialize_simulation.f90:
if (SIMULATION_TYPE /= 1 .and. NSOURCES > 1000) call exit_mpi(myrank, 'for adjoint simulations, NSOURCES <= 1000')

I don't see why this constraint exists. Shouldn't it be either a parameter, or just equal to the maximum of either sources or receivers? There are many reasons a user might want to create a kernel with more than 1000 sources.

[danielpeter]

closed by cb7bfc5

for an explanation, see the email below on the seismo-dev list:

hi Carl,

the limit is somewhat arbitrary and i think goes back to the implementation from Qinya. for pure adjoint simulations (SIMULATION_TYPE == 2), the code outputs displacement (S00001.NT.BXX.semd,..) and strains (S00001.NT.SEE.semd,..) as well as source derivative kernels (src_frechet.00001,..) all for each point source. the naming convention for those files used something like (.., i5.5,..), which would have limit the number of sources to 99999.

for now, i submitted a commit which uses the same convention as in the global code. in the global code, we use a hard limit of 999,999, because file names use (..,i6.6,..). if that is still too low, please let me know and i increase it further.

best wishes,
daniel

[carltape]

Thank you, Daniel! Great to have such an explicit error message and to have
consistency with the global code.

Carl

On Tue, Nov 4, 2014 at 5:14 AM, daniel peter notifications@github.com
wrote:

closed by cb7bfc5
cb7bfc5

for an explanation, see the email below on the seismo-dev list:

hi Carl,

the limit is somewhat arbitrary and i think goes back to the
implementation from Qinya. for pure adjoint simulations (SIMULATION_TYPE ==
2), the code outputs displacement (S00001.NT.BXX.semd,..) and strains
(S00001.NT.SEE.semd,..) as well as source derivative kernels
(src_frechet.00001,..) all for each point source. the naming convention for
those files used something like (.., i5.5,..), which would have limit the
number of sources to 99999.

for now, i submitted a commit which uses the same convention as in the
global code. in the global code, we use a hard limit of 999,999, because
file names use (..,i6.6,..). if that is still too low, please let me know
and i increase it further.

best wishes,
daniel

—
Reply to this email directly or view it on GitHub
#279 (comment)
.

[komatits]

email below added as a comment in both source codes (since it is a
useful explanation).

Dimitri.

On 11/04/2014 03:14 PM, daniel peter wrote:

closed by cb7bfc5
cb7bfc5

for an explanation, see the email below on the seismo-dev list:

hi Carl,

the limit is somewhat arbitrary and i think goes back to the
implementation from Qinya. for pure adjoint simulations (SIMULATION_TYPE
== 2), the code outputs displacement (S00001.NT.BXX.semd,..) and strains
(S00001.NT.SEE.semd,..) as well as source derivative kernels
(src_frechet.00001,..) all for each point source. the naming convention
for those files used something like (.., i5.5,..), which would have
limit the number of sources to 99999.

for now, i submitted a commit which uses the same convention as in the
global code. in the global code, we use a hard limit of 999,999, because
file names use (..,i6.6,..). if that is still too low, please let me
know and i increase it further.

best wishes,
daniel

—
Reply to this email directly or view it on GitHub
#279 (comment)
Bug from
https://github.com/notifications/beacon/AFjDKT6fyleDNlEoxBXxkkzFgP7WJRbWks5nKNa7gaJpZM4C0HiC.gif

{"@context":"http://schema.org","@type":"EmailMessage","description":"View
this Issue on
GitHub","action":{"@type":"ViewAction","url":"https://github.com/geodynamics/specfem3d/issues/279#issuecomment-61644661","name":"View
Issue"}}

Dimitri Komatitsch
CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
UPR 7051, Marseille, France http://komatitsch.free.fr