BLAS for Nim

This is a Nim wrapper for the BLAS routines.

You can import nimblas/cblas to use the standard BLAS interface, or just import nimblas for a version that is more Nim-friendly.

The Nim version removes the prefixes and uses dispatch based on types instead. This means that, for instance, both cblas_saxpy and cblas_daxpy become simply axpy, and the correct version is chosen by checking the size of parameters at the usage site.

Only a subset of BLAS is available under nimblas, with more operations added on necessity.

For a higher-level linear algebra library based on this, check out

• neo
• arraymancer.

Linking BLAS implementations

The library requires to link some BLAS implementation to perform the actual linear algebra operations. By default, it tries to link whatever is the default system-wide BLAS implementation.

You can link against a different BLAS implementation by a combination of:

• changing the path for linked libraries (use --clibdir for this).
• using the --define:blas flag. By default (i.e. if you don’t set this flag), the system tries to load a BLAS library by looking for the most common blas library file names according to the underling operating system (e.g. blas.dll, openblas.dll, libopenblas.dll, mkl_intel_lp64.dll in Windows, libblas.so, libcblas.so or libopenblas.so on Linux, etc). However, if you want to link to one specific library, skipping the automatic search, you can specify it with this flag. For instance, the Linux Arch distribution and its derivatives (such as Manjaro) expose the C API for BLAS inside a library called libcblas.so (unlike most other distributions that put it into libblas.so). To explicitly link with that library, you can set --define:blas=cblas. Note the missing lib prefix and .so suffix, which nimblas adds automatically (similarly on windows you should not include the .dll extension when setting this flag).

For more examples, see the tasks inside nimblas.nimble.

(Previously there was a more ad hoc mechanism using flags called -d:atlas, -d:openblas or -d:mkl, which is deprecated as of NimBLAS 0.2.)

Packages for various BLAS implementations are available from the package managers of many Linux distributions. On OSX one can add the brew formulas from Homebrew Science, such as brew install homebrew/science/openblas. On Windows you can download pre-built binaries from the OpenBLAS github repository and add the library folder to your PATH or copy it into your executable folder.

You may also need to add suitable paths for the includes and library dirs. On OSX, this should do the trick

switch("clibdir", "/usr/local/opt/openblas/lib")
switch("cincludes", "/usr/local/opt/openblas/include")

If you have problems with MKL, you may want to link it statically. Just pass the options

--dynlibOverride:mkl_intel_lp64
--passL:${PATH_TO_MKL}/libmkl_intel_lp64.a

to enable static linking.