Skip to content
Reference implementation of Batched BLAS routines
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
include
install
script
src
testing
tools
Doxyfile
LICENSE
Makefile.gen
README
README.md
history.log
mainpage.dox
make.inc
make.inc.bak
make.inc.mkl
make.inc.sys.example
setup.py

README.md

BBLAS-ref

Reference implementation of Batched BLAS routines

Please note that this is a draft specification which is currently open to suggestions from the community and may be subject to large changes in the future.

Compilation is achieved using the script setup.py. For instance to compile a basic version using only MKL we can run the following.
source /opt/intel/compilers_and_libraries/linux/bin/compilervars.sh intel64
./setup.py --blaslib="-Wl,--no-as-needed -L${MKLROOT}/lib/intel64 -lmkl_intel_lp64 -lmkl_core -lmkl_gnu_thread -lpthread -lm -ldl"

By running ./setup.py -h you can see the options required to compile with CUDA and MAGMA.

This compilation will also create all the documentation in the "docs" folder. After compilation is complete (don't worry that "make install" fails) tests can be run using the script testing/run_tests.py. By default running ./run_tests.py with no input arguments will run all available tests.

You can’t perform that action at this time.