Coordinates cannot be read under AGRIF

[fschwarzkopf]

AGRIF crashes when reading 1_coordinates.nc, in particular in domhgr.F90:

CALL iom_get( inum, jpdom_data, 'glamt', glamt )

Test configuration is an ORCA05 with 1/10 of a degree tropical Atlantic nest parallelized on 4 PE (4x1) on NEC-SX9

Following ERROR message appears:

90 Fatal exception PROG=lib_mpp.sub_loop_mpprecv ELN=3415(400041c1c)
SIGXABT: Distributed parallel program aborted

Called from lib_mpp.mpprecv ELN=3379(40004189c)
Called from lib_mpp.sub_loop_mpp_lnk_2d ELN=1624(4000294dc)
Called from lib_mpp.mpp_lnk_2d ELN=1436(400025748)
Called from iom.sub_loop_iom_get_123d ELN=994(40035f330)
Called from iom.iom_get_123d ELN=675(40035b3a4)
Called from iom.iom_g2d ELN=613(40035a5bc)
Called from domhgr.sub_loop_hgr_read ELN=739(4004e7f8c)
Called from domhgr.hgr_read ELN=659(4004e718c)
Called from domhgr.sub_loop_dom_hgr ELN=271(4004df88c)
Called from domhgr.dom_hgr ELN=63(4004de024)
Called from domain.sub_loop_dom_init ELN=176(40036ffa8)
Called from domain.dom_init ELN=96(40036f520)
Called from opa.sub_loop_opa_init ELN=374(4002ee7e4)
Called from opa.opa_init ELN=234(4002ed908)
Called from sub_loop_agrif_initvalues ELN=197(4002f3bf0)
Called from agrif_initvalues ELN=119(4002f33c0)
Called from agrif_clustering.agrif_init_hier ELN=1329(40078e1c8)
Called from agrif_util.agrif_regrid ELN=778(400690fc0)
Called from agrif_util.agrif_step ELN=635(400690300)
Called from opa.sub_loop_opa_model ELN=194(4002eccd8)
Called from opa.opa_model ELN=127(4002ec764)
Called from model ELN=19(4000009c0)

We assume that the bug appears in iom.F90:

!--- overlap areas and extra hallows (mpp)

IF( PRESENT(pv_r2d) .AND. idom /= jpdom_unknown ) THEN

CALL lbc_lnk( pv_r2d,'Z',-999.,'no0' )

Regards,
Franziska Schwarzkopf

[fschwarzkopf]

Update:

• We tested the original sample ORAC2_LIM AGRIF zoom on 4X1 and 1X1 PE the former failed at the same point as the original problem; the latter passed this point.
• We replaced the 1_coordinates.nc and AGRIF_FixedGrids.in files by a set of proofed ones > The run failed at the same point.
• We deceased our optimisation level > The run failed at the same point.

Are there any suggestions what to test next or other ideas?

Thanks, Franziska