Memory-related failures in QA

Clicked A Few Times

3:40:19 PM PST - Thu, Nov 12th 2015
Hi, I'm noticing an excessive number of QA failures arising from, e.g., segfaults, or things like ptsalloc: increase memory in input line 0 ------------------------------------------------------------------------ This build (version 6.6., Intel MPI 5, compilers 15.0.3) used 64-bit integer. Is it possible that the tests with manually spec'd memory segment sizes haven't taken this into account and need to be increased? Otherwise I'm unclear how the PNNL builds are passing testing, unless I'm doing something weird in the build (trying not to...). Thanks, Chris Failures 'NWChem execution failed' so far Running tests/dplot/dplot : segfault Running tests/h2o_diag_opt/h2o_diag_opt : ptsalloc Running tests/h2o_diag_opt_ub3lyp/h2o_diag_opt_ub3lyp : ptsalloc Running tests/h2o_cg_opt_rob3lyp/h2o_cg_opt_rob3lyp : ptsalloc Running tests/pspw/pspw : segfault Running tests/pspw_SiC/pspw_SiC : segfault Running tests/pspw_md/pspw_md : D1dB_init: mapping error 0 Running tests/paw/paw : segfault Running tests/pspw_polarizability/pspw_polarizability : segfault Running tests/pspw_stress : segfault

Forum Vet

5:29:33 PM PST - Thu, Nov 12th 2015
Chris Could you provide more details about the installation settings, for example: modules included, specs of the HW used, etc ... Could you please post all the env. variables you have set for the NWChem compilation (for example, have you set USE_NOIO)? One more question: how may processes were you using in the QA tests Thanks