3 [![Join the chat at https://gitter.im/xianyi/OpenBLAS](https://badges.gitter.im/Join%20Chat.svg)](https://gitter.im/xianyi/OpenBLAS?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)
5 Travis CI: [![Build Status](https://travis-ci.org/xianyi/OpenBLAS.svg?branch=develop)](https://travis-ci.org/xianyi/OpenBLAS)
7 AppVeyor: [![Build status](https://ci.appveyor.com/api/projects/status/09sohd35n8nkkx64/branch/develop?svg=true)](https://ci.appveyor.com/project/xianyi/openblas/branch/develop)
9 [![Build Status](https://dev.azure.com/xianyi/OpenBLAS/_apis/build/status/xianyi.OpenBLAS?branchName=develop)](https://dev.azure.com/xianyi/OpenBLAS/_build/latest?definitionId=1&branchName=develop)
13 OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
15 Please read the documentation on the OpenBLAS wiki pages: <https://github.com/xianyi/OpenBLAS/wiki>.
19 We provide official binary packages for the following platform:
23 You can download them from [file hosting on sourceforge.net](https://sourceforge.net/projects/openblas/files/).
25 ## Installation from Source
27 Download from project homepage, https://xianyi.github.com/OpenBLAS/, or check out the code
28 using Git from https://github.com/xianyi/OpenBLAS.git.
29 Buildtime parameters can be chosen in Makefile.rule, see there for a short description of each option.
30 Most can also be given directly on the make or cmake command line.
34 Building OpenBLAS requires the following to be installed:
37 * A C compiler, e.g. GCC or Clang
38 * A Fortran compiler (optional, for LAPACK)
39 * IBM MASS (optional, see below)
43 Simply invoking `make` (or `gmake` on BSD) will detect the CPU automatically.
44 To set a specific target CPU, use `make TARGET=xxx`, e.g. `make TARGET=NEHALEM`.
45 The full target list is in the file `TargetList.txt`.
49 Set `CC` and `FC` to point to the cross toolchains, and set `HOSTCC` to your host C compiler.
50 The target must be specified explicitly when cross compiling.
54 * On an x86 box, compile this library for a loongson3a CPU:
56 make BINARY=64 CC=mips64el-unknown-linux-gnu-gcc FC=mips64el-unknown-linux-gnu-gfortran HOSTCC=gcc TARGET=LOONGSON3A
59 * On an x86 box, compile this library for a loongson3a CPU with loongcc (based on Open64) compiler:
61 make CC=loongcc FC=loongf95 HOSTCC=gcc TARGET=LOONGSON3A CROSS=1 CROSS_SUFFIX=mips64el-st-linux-gnu- NO_LAPACKE=1 NO_SHARED=1 BINARY=32
66 A debug version can be built using `make DEBUG=1`.
68 ### Compile with MASS support on Power CPU (optional)
70 The [IBM MASS](https://www.ibm.com/support/home/product/W511326D80541V01/other_software/mathematical_acceleration_subsystem) library consists of a set of mathematical functions for C, C++, and Fortran applications that are tuned for optimum performance on POWER architectures.
71 OpenBLAS with MASS requires a 64-bit, little-endian OS on POWER.
72 The library can be installed as shown:
76 wget -q http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/ubuntu/public.gpg -O- | sudo apt-key add -
77 echo "deb http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/ubuntu/ trusty main" | sudo tee /etc/apt/sources.list.d/ibm-xl-compiler-eval.list
79 sudo apt-get install libxlmass-devel.8.1.5
84 wget http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/rhel7/repodata/repomd.xml.key
85 sudo rpm --import repomd.xml.key
86 wget http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/rhel7/ibm-xl-compiler-eval.repo
87 sudo cp ibm-xl-compiler-eval.repo /etc/yum.repos.d/
88 sudo yum install libxlmass-devel.8.1.5
91 After installing the MASS library, compile OpenBLAS with `USE_MASS=1`.
92 For example, to compile on Power8 with MASS support: `make USE_MASS=1 TARGET=POWER8`.
94 ### Install to a specific directory (optional)
96 Use `PREFIX=` when invoking `make`, for example
99 make install PREFIX=your_installation_directory
102 The default installation directory is `/opt/OpenBLAS`.
104 ## Supported CPUs and Operating Systems
106 Please read `GotoBLAS_01Readme.txt` for older CPU models already supported by the 2010 GotoBLAS.
108 ### Additional supported CPUs
112 - **Intel Xeon 56xx (Westmere)**: Used GotoBLAS2 Nehalem codes.
113 - **Intel Sandy Bridge**: Optimized Level-3 and Level-2 BLAS with AVX on x86-64.
114 - **Intel Haswell**: Optimized Level-3 and Level-2 BLAS with AVX2 and FMA on x86-64.
115 - **Intel Skylake-X**: Optimized Level-3 and Level-2 BLAS with AVX512 and FMA on x86-64.
116 - **AMD Bobcat**: Used GotoBLAS2 Barcelona codes.
117 - **AMD Bulldozer**: x86-64 ?GEMM FMA4 kernels. (Thanks to Werner Saar)
118 - **AMD PILEDRIVER**: Uses Bulldozer codes with some optimizations.
119 - **AMD STEAMROLLER**: Uses Bulldozer codes with some optimizations.
120 - **AMD ZEN**: Uses Haswell codes with some optimizations.
124 - **ICT Loongson 3A**: Optimized Level-3 BLAS and the part of Level-1,2.
125 - **ICT Loongson 3B**: Experimental
129 - **ARMv6**: Optimized BLAS for vfpv2 and vfpv3-d16 (e.g. BCM2835, Cortex M0+)
130 - **ARMv7**: Optimized BLAS for vfpv3-d32 (e.g. Cortex A8, A9 and A15)
134 - **ARMv8**: Basic ARMV8 with small caches, optimized Level-3 and Level-2 BLAS
135 - **Cortex-A53**: same as ARMV8 (different cpu specifications)
136 - **Cortex A57**: Optimized Level-3 and Level-2 functions
137 - **Cortex A72**: same as A57 ( different cpu specifications)
138 - **Cortex A73**: same as A57 (different cpu specifications)
139 - **Falkor**: same as A57 (different cpu specifications)
140 - **ThunderX**: Optimized some Level-1 functions
141 - **ThunderX2T99**: Optimized Level-3 BLAS and parts of Levels 1 and 2
142 - **TSV110**: Optimized some Level-3 helper functions
146 - **POWER8**: Optimized BLAS, only for PPC64LE (Little Endian), only with `USE_OPENMP=1`
147 - **POWER9**: Optimized Level-3 BLAS (real) and some Level-1,2. PPC64LE with OpenMP only.
149 #### IBM zEnterprise System
151 - **Z13**: Optimized Level-3 BLAS and Level-1,2
152 - **Z14**: Optimized Level-3 BLAS and (single precision) Level-1,2
154 ### Support for multiple targets in a single library
156 OpenBLAS can be built for multiple targets with runtime detection of the target cpu by specifiying DYNAMIC_ARCH=1 in Makefile.rule, on the gmake command line or as -DDYNAMIC_ARCH=TRUE in cmake.
157 For **x86_64**, the list of targets this activates contains Prescott, Core2, Nehalem, Barcelona, Sandybridge, Bulldozer, Piledriver, Steamroller, Excavator, Haswell, Zen, SkylakeX. For cpu generations not included in this list, the corresponding older model is used. If you also specify DYNAMIC_OLDER=1, specific support for Penryn, Dunnington, Opteron, Opteron/SSE3, Bobcat, Atom and Nano is added. Finally there is an option DYNAMIC_LIST that allows to specify an individual list of targets to include instead of the default.
158 DYNAMIC_ARCH is also supported on **x86**, where it translates to Katmai, Coppermine, Northwood, Prescott, Banias,
159 Core2, Penryn, Dunnington, Nehalem, Athlon, Opteron, Opteron_SSE3, Barcelona, Bobcat, Atom and Nano.
160 On **ARMV8**, it enables support for CortexA53, CortexA57, CortexA72, CortexA73, Falkor, ThunderX, ThunderX2T99, TSV110 as well as generic ARMV8 cpus.
161 For **POWER**, the list encompasses POWER6, POWER8 and POWER9, on **ZARCH** it comprises Z13 and Z14.
162 The TARGET option can be used in conjunction with DYNAMIC_ARCH=1 to specify which cpu model should be assumed for all the
163 common code in the library, usually you will want to set this to the oldest model you expect to encounter.
164 Please not that it is not possible to combine support for different architectures, so no combined 32 and 64 bit or x86_64 and arm64 in the same library.
169 - **MinGW or Visual Studio (CMake)/Windows**: Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-use-OpenBLAS-in-Microsoft-Visual-Studio>.
170 - **Darwin/macOS/OSX/iOS**: Experimental. Although GotoBLAS2 already supports Darwin, we are not OSX/iOS experts.
171 - **FreeBSD**: Supported by the community. We don't actively test the library on this OS.
172 - **OpenBSD**: Supported by the community. We don't actively test the library on this OS.
173 - **NetBSD**: Supported by the community. We don't actively test the library on this OS.
174 - **DragonFly BSD**: Supported by the community. We don't actively test the library on this OS.
175 - **Android**: Supported by the community. Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-build-OpenBLAS-for-Android>.
176 - **AIX**: Supported on PPC up to POWER8
177 - **Haiku**: Supported by the community. We don't actively test the library on this OS.
178 - **SunOS**: Supported by the community. We don't actively test the library on this OS:
182 Statically link with `libopenblas.a` or dynamically link with `-lopenblas` if OpenBLAS was
183 compiled as a shared library.
185 ### Setting the number of threads using environment variables
187 Environment variables are used to specify a maximum number of threads.
191 export OPENBLAS_NUM_THREADS=4
192 export GOTO_NUM_THREADS=4
193 export OMP_NUM_THREADS=4
196 The priorities are `OPENBLAS_NUM_THREADS` > `GOTO_NUM_THREADS` > `OMP_NUM_THREADS`.
198 If you compile this library with `USE_OPENMP=1`, you should set the `OMP_NUM_THREADS`
199 environment variable; OpenBLAS ignores `OPENBLAS_NUM_THREADS` and `GOTO_NUM_THREADS` when
200 compiled with `USE_OPENMP=1`.
202 ### Setting the number of threads at runtime
204 We provide the following functions to control the number of threads at runtime:
207 void goto_set_num_threads(int num_threads);
208 void openblas_set_num_threads(int num_threads);
211 If you compile this library with `USE_OPENMP=1`, you should use the above functions too.
215 Please submit an issue in https://github.com/xianyi/OpenBLAS/issues.
219 * OpenBLAS users mailing list: https://groups.google.com/forum/#!forum/openblas-users
220 * OpenBLAS developers mailing list: https://groups.google.com/forum/#!forum/openblas-dev
224 Please see Changelog.txt to view the differences between OpenBLAS and GotoBLAS2 1.13 BSD version.
228 * Please read the [FAQ](https://github.com/xianyi/OpenBLAS/wiki/Faq) first.
229 * Please use GCC version 4.6 and above to compile Sandy Bridge AVX kernels on Linux/MinGW/BSD.
230 * Please use Clang version 3.1 and above to compile the library on Sandy Bridge microarchitecture.
231 Clang 3.0 will generate the wrong AVX binary code.
232 * Please use GCC version 6 or LLVM version 6 and above to compile Skylake AVX512 kernels.
233 * The number of CPUs/cores should less than or equal to 256. On Linux `x86_64` (`amd64`),
234 there is experimental support for up to 1024 CPUs/cores and 128 numa nodes if you build
235 the library with `BIGNUMA=1`.
236 * OpenBLAS does not set processor affinity by default.
237 On Linux, you can enable processor affinity by commenting out the line `NO_AFFINITY=1` in
238 Makefile.rule. However, note that this may cause
239 [a conflict with R parallel](https://stat.ethz.ch/pipermail/r-sig-hpc/2012-April/001348.html).
240 * On Loongson 3A, `make test` may fail with a `pthread_create` error (`EAGAIN`).
241 However, it will be okay when you run the same test case on the shell.
245 1. [Check for open issues](https://github.com/xianyi/OpenBLAS/issues) or open a fresh issue
246 to start a discussion around a feature idea or a bug.
247 2. Fork the [OpenBLAS](https://github.com/xianyi/OpenBLAS) repository to start making your changes.
248 3. Write a test which shows that the bug was fixed or that the feature works as expected.
249 4. Send a pull request. Make sure to add yourself to `CONTRIBUTORS.md`.
253 Please read [this wiki page](https://github.com/xianyi/OpenBLAS/wiki/Donation).