All geek questions in one place

Numismas with ATLAS or OpenBLAS?

Having spent a huge amount of time creating ATLAS from the source code, I found that libopenblas and libatals are in the OpenSUSE 13.1 repository. My questions

  • Simple installation (without configuration on your own computer) of "libatlas" in the repository really improves computing performance?

  • Is OpenBLAS better than ATLAS, or just better than just installing libatlas in a Linux flavor store? See For faster use of R, use OpenBLAS instead: better than ATLAS, trivially switching to Ubuntu .

  • I followed this post Compiling Numpy with OpenBLAS , but cannot find the module "numpy.core._dotblas". Moreover, I could not create Numpy with ATLAS and OpenBLAS at the same time.

  • Can someone post a .py file or bash code to compare between ATLAS and OpenBLAS? For instance.

  • I built Numpy-1.9 using my own ATLAS , compiled OpenBLAS from the source code, and installed "libopenblaso" (OpenMP version) and "libopenblasp" (pthreads version) in the OpenSUSE 13.1 repository. How to set up links and libraries so that Numpy-1.9 can use OpenBLAS instead of ATLAS without reinstalling the Numpy-1.9 package.

Note. If you install "libatlas" in the repository, ATLAS is not configured for your computer and cannot significantly calculate the performance. So I first built and configured ATLAS, and then built Numpy with my own ATLAS. After that, I tried to associate OpenBLAS with Numpy, but could not.

Thank you very much in advance!

Thanks @Dmitry for the quick reply! But the issues have not been resolved ...

Installation

$ sudo zypper in libopenblasp0 The following NEW package is going to be installed: libopenblasp0 1 new package to install. Overall download size: 3.0 MiB. After the operation, additional 30.3 MiB will be used. Continue? [y/n/? shows all options] (y): Retrieving package libopenblasp0-0.2.11-11.1.x86_64 (1/1), 3.0 MiB ( 30.3 MiB unpacked) Retrieving: libopenblasp0-0.2.11-11.1.x86_64.rpm ...........................[done (2.1 MiB/s)] (1/1) Installing: libopenblasp0-0.2.11-11.1 ............................................[done] Additional rpm output: /sbin/ldconfig: Can't link /usr/lib64//usr/local/atlas/lib/libtatlas.so to libopenblas.so.0

Q: Why is there a funny double slash? .. 64 // usr .. "?

Link the library

 $ /usr/sbin/update-alternatives --config libblas.so.3 Selection Path Priority Status ------------------------------------------------------------ 0 /usr/local/atlas/lib/libtatlas.so 70 auto mode 1 /usr/lib64/blas/libblas.so.3 50 manual mode 2 /usr/lib64/libopenblasp.so.0 20 manual mode 3 /usr/local/atlas/lib/libcblas.a 50 manual mode 4 /usr/local/atlas/lib/libptcblas.a 60 manual mode 5 /usr/local/atlas/lib/libsatlas.so 65 manual mode * 6 /usr/local/atlas/lib/libtatlas.so 70 manual mode

Q: Is this configuration great because some of the static .a libraries are linked?

Note: "libopenblasp.so.0" is automatically linked after "zypper in", while all "satin" libraries are manually created with the command:

 $ /usr/sbin/update-alternatives --install /usr/lib64/blas/libblas.so.3 libblas.so.3 /usr/local/atlas/lib/libxxxx.x <Integer>
+6
source share
1 answer

You can switch between BLAS and LAPACK implementations using update alternatives. For instance:

 /usr/sbin/update-alternatives --config libblas.so.3

http://en.opensuse.org/openSUSE:Science_Linear_algebra_libraries

I do not know about atlas packages, but we provide openblas with multiarch support. The Math kernel includes all optimizations optimized for the processor and dynamically sets the correct option. Note that there are 3 versions of the openblas library: serial, pthreads, and openmp. You need to use pthreads or openmp.

+5
source

Source: https://habr.com/ru/post/975247/

More articles:


All Articles