Illegal instruction (core dumped) when execute HPL

[jliu41]

OS: RHEL 9.3 5.14.0-362.8.1.el9_3.x86_64

gcc: 11.4.1 20230605 (Red Hat 11.4.1-2)

oneAPI Base Toolkit: BaseKit_p_2024.1.0.596

HPC Toolkit: HPCKit_p_2024.1.0.560

CPU: Sierra Forrest-SP

HPL version: l_onemklbench_p_2024.1.0_517

Always get Illegal instruction (core dumped) when execute HPL.

Tried to add export MKL_ENABLE_INSTRUCTIONS=AVX2 and -march=native into build.sh script

But got Illegal instruction with compiled binary

core dump msg: kernel: traps: xhpl_intel64_dy[680773] trap invalid opcode ip:7f1eeccc5a77 sp:7ffcb9522ca8 error:0 in libmpi.so.12.0.0[7f1eec800000+b9c000]

[Gennady_F_Intel · Accepted solution]

This is MPI issue. Could you please try to 

export I_MPI_COLL_INTRANODE=pt2pt

export I_MPI_SHM = clx_avx2

 

[jliu41]

Yes, the error disappear with these two MPI environment variable.

Will MPI fix this mismatch with CPU flags?

[Gennady_F_Intel]

The next version of MPI ( 2021.13) will fix all issues.