1.caffe 上OpenBLAS的安装使用
BLAS是一个数学函数接口标准，有很多个实现。按照Caffe官方ubuntu的安装文档默认安装的是ATLAS。这个版本的BLAS不能利用多核CPU，我们将其换为OpenBLAS，可以利用多核CPU并行计算，加快Caffe的分类速度。
源代码安装配置OpenBLAS
-git代码到本地并安装
git clone https://github.com/xianyi/OpenBLAS.git
cd OpenBLAS
make -j4
make install
-修改Caffe的Makefile.config
BLAS_INCLUDE := /opt/OpenBLAS/include
BLAS_LIB := /opt/OpenBLAS/lib
-caffe重新make
make clean
make pycaffe
make all -j6
make test && runtest
在caffe安装好后，编译时要指定Makefile.config
# BLAS choice:
# atlas for ATLAS (default)
# mkl for MKL
# open for OpenBlas
BLAS := open
# Custom (MKL/ATLAS/OpenBLAS) include and lib directories.
# Leave commented to accept the defaults for your choice of BLAS
# (which should work)!
BLAS_INCLUDE := /usr/local/OpenBlas/include
BLAS_LIB := /usr/local/OpenBlas/lib
在runtest时会报一个错：：error while loading shared libraries: libopenblas.so.0: cannot open shared object file: No such file or directory.
解决方法：在/usr/lib/下建立一个 软链接将 libopenblas.so.0指向/openbls安装目录/lib/ libopenblas.so.0
sudo ln -s /opt/OpenBLAS/lib/libopenblas.so.0 /usr/lib/libopenblas.so.0
网友评论