ralna / spral

Sparse Parallel Robust Algorithms Library
https://ralna.github.io/spral/
Other
104 stars 27 forks source link

SSMFE segfaults under ifort with OpenBLAS #148

Closed jfowkes closed 11 months ago

jfowkes commented 11 months ago

When compiled with the ifort compiler, the SSMFE tests segfault on both Linux and Mac:

FAIL: ssmfe_test
./test-driver: line 112:  8183 Segmentation fault      (core dumped) "$@" >> "$log_file" 2>&1
FAIL: ssmfe_ciface_test
FAIL: ssmfe_test
./test-driver: line 112:  9147 Segmentation fault: 11  "$@" >> "$log_file" 2>&1
FAIL: ssmfe_ciface_test

we should investigate and fix the segfault.

jfowkes commented 11 months ago

I have been able to reproduce this segfault locally, gdb reports errors such as the following:

Thread 22 "ssmfe_test" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffc15d4700 (LWP 693940)]
0x00007ffff6fd8a00 in ?? () from /usr/lib/x86_64-linux-gnu/libopenblas.so.0
(gdb) bt
#0  0x00007ffff6fd8a00 in ?? () from /usr/lib/x86_64-linux-gnu/libopenblas.so.0
#1  0x00007ffff6fd8b00 in ?? () from /usr/lib/x86_64-linux-gnu/libopenblas.so.0
#2  0x00007ffff61528c4 in ?? () from /usr/lib/x86_64-linux-gnu/libopenblas.so.0
#3  0x00007ffff6152aa7 in ?? () from /usr/lib/x86_64-linux-gnu/libopenblas.so.0
#4  0x00007ffff5c27609 in start_thread (arg=<optimised out>) at pthread_create.c:477
#5  0x00007ffff3acd133 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

so this looks to be an issue with using OpenBLAS with Intel pthreads under SSMFE.

jfowkes commented 11 months ago

@amontoison TL;DR we need to use Intel MKL not OpenBLAS with the Intel compilers :see_no_evil: