Skip to content

Latest commit

 

History

History
59 lines (44 loc) · 1.09 KB

results.rst

File metadata and controls

59 lines (44 loc) · 1.09 KB

The mathlibs-tester scaling performance

depending on the OpenMP MKL-threading

  • python setup.py --fc=ifort --int64 --mkl=parallel build_ifort_mklpar_i8
  • python setup.py --fc=ifort --int64 --mkl=sequential build_ifort_mklseq_i8

The DGESV (LAPACK); DGEMM (BLAS) and DNRM2(BLAS) library routines are called.

Run as

/bin/example N i

where N is the size of the matrix, i is the print level (recommended 1).

The lxir071.gsi.de interactive server

Intel(R) Xeon(R) CPU L5506 @ 2.13GHz; 8 CPU

  • export MKL_NUM_THREADS=#
  • export MKL_DYNAMIC=false

/bin/example 3000 1

# time
seq 16.497
1 16.141
2 8.485
4 4.650
8 3.397

The hpcc.umb.sk SIVVP cluster

12 Intel CPU per node

  • export MKL_NUM_THREADS=#
  • export MKL_DYNAMIC=false

/bin/example 7000 1

# time
seq 127.839
2 66.642
4 35.038
8 18.428
12 13.861