strumpack test fails on perlmutter

[wspear]

@pghysels

The strumpack test defined here: https://github.com/E4S-Project/testsuite/tree/master/validation_tests/strumpack fails for the perlmutter install under e4s 22.11 with the following variants:

-- linux-sles15-zen3 / [email protected] -------------------------------
jslue4d [email protected]+butterflypack+c_interface~count_flops~cuda~ipo+mpi+openmp+parmetis~rocm~scotch+shared~slate~task_timers+zfp build_system=cmake build_type=RelWithDebInfo

With the following console output. (It seems that it runs to near completion and segfaults at the end)

strumpack~rocm %gcc: jslue4d
+ export OMP_NUM_THREADS=1
+ OMP_NUM_THREADS=1
+ ./testPoisson2d 100 --sp_disable_gpu
solving 2D 100x100 Poisson problem with 1 right hand sides
# Initializing STRUMPACK
# using 1 OpenMP thread(s)
# matrix equilibration, r_cond = 1 , c_cond = 1 , type = N
# initial matrix:
#   - number of unknowns = 10,000
#   - number of nonzeros = 49,600
# nested dissection reordering:
#   - Geometric reordering
#   - strategy parameter = 8
#   - number of separators = 1,967
#   - number of levels = 12
#   - nd time = 0.000863288
#   - symmetrization time = 2.67616e-06
# symbolic factorization:
#   - nr of dense Frontal matrices = 1,967
#   - symb-factor time = 0.000889237
# multifrontal factorization:
#   - estimated memory usage (exact solver) = 4.50818 MB
#   - minimum pivot, sqrt(eps)*|A|_1 = 8.42937e-08
#   - replacing of small pivots is not enabled
#   - factor time = 0.0113813
#   - factor nonzeros = 563,522
#   - factor memory = 4.50818 MB
REFINEMENT it. 0	res =      442.368	rel.res =            1	bw.error =            1
REFINEMENT it. 1	res =  7.89021e-14	rel.res =  1.78363e-16	bw.error =   5.8175e-16
# DIRECT/GMRES solve:
#   - abs_tol = 1e-10, rel_tol = 1e-06, restart = 30, maxit = 5000
#   - number of Krylov iterations = 1
#   - solve time = 0.00290866
# COMPONENTWISE SCALED RESIDUAL = 5.68529e-16
# relative error = ||x-x_exact||_F/||x_exact||_F = 3.96446e-15
./run.sh: line 7: 97192 Segmentation fault      ./testPoisson2d 100 --sp_disable_gpu

[pghysels]

@lisaclaus from NERSC recommends replacing cray-libsci/21.08.1.2 with cray-libsci/22.06.1.3 on Perlmutter.