multithreading - Building BLAS, ATLAS and LAPACK without OpenMP support -

i have been using prebuilt linear algebra libraries use openmp parallelize execution.

right working in project openmp parallelize tasks. set environment variable omp_num_threads.

some threads make use of blas , lapack linear algebra routines. these libraries parallel , take @ environment variable omp_num_threads see number of threads have use.

for requirements of project. need avoid nested parallelization making use of non parallel blas , lapack implementation.

what best library , how can built?

thank in advance.

you can use openblas. library includes 1 version of lapack routines , quite flexible respect threading.

the number of threads can defined either @ compile time:

$shell> make use_thread=0 

or using explicit environment variable:

export openblas_num_threads=4 

or using library api @ runtime:

void openblas_set_num_threads(int num_threads);