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 Drone CI: [![Build Status](https://cloud.drone.io/api/badges/xianyi/OpenBLAS/status.svg?branch=develop)](https://cloud.drone.io/xianyi/OpenBLAS/)
11 [![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)
16 OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
18 Please read the documentation on the OpenBLAS wiki pages: <https://github.com/xianyi/OpenBLAS/wiki>.
22 We provide official binary packages for the following platform:
26 You can download them from [file hosting on sourceforge.net](https://sourceforge.net/projects/openblas/files/).
28 ## Installation from Source
30 Download from project homepage, https://xianyi.github.com/OpenBLAS/, or check out the code
31 using Git from https://github.com/xianyi/OpenBLAS.git. (If you want the most up to date version, be
32 sure to use the develop branch - master is several years out of date due to a change of maintainership.)
33 Buildtime parameters can be chosen in Makefile.rule, see there for a short description of each option.
34 Most can also be given directly on the make or cmake command line.
38 Building OpenBLAS requires the following to be installed:
41 * A C compiler, e.g. GCC or Clang
42 * A Fortran compiler (optional, for LAPACK)
43 * IBM MASS (optional, see below)
47 Simply invoking `make` (or `gmake` on BSD) will detect the CPU automatically.
48 To set a specific target CPU, use `make TARGET=xxx`, e.g. `make TARGET=NEHALEM`.
49 The full target list is in the file `TargetList.txt`.
53 Set `CC` and `FC` to point to the cross toolchains, and set `HOSTCC` to your host C compiler.
54 The target must be specified explicitly when cross compiling.
58 * On an x86 box, compile this library for a loongson3a CPU:
60 make BINARY=64 CC=mips64el-unknown-linux-gnu-gcc FC=mips64el-unknown-linux-gnu-gfortran HOSTCC=gcc TARGET=LOONGSON3A
62 or same with the newer mips-crosscompiler put out by Loongson that defaults to the 32bit ABI:
64 make HOSTCC=gcc CC='/opt/mips-loongson-gcc7.3-linux-gnu/2019.06-29/bin/mips-linux-gnu-gcc -mabi=64' FC='/opt/mips-loongson-gcc7.3-linux-gnu/2019.06-29/bin/mips-linux-gnu-gfortran -mabi=64' TARGET=LOONGSON3A
67 * On an x86 box, compile this library for a loongson3a CPU with loongcc (based on Open64) compiler:
69 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
74 A debug version can be built using `make DEBUG=1`.
76 ### Compile with MASS support on Power CPU (optional)
78 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.
79 OpenBLAS with MASS requires a 64-bit, little-endian OS on POWER.
80 The library can be installed as shown:
84 wget -q http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/ubuntu/public.gpg -O- | sudo apt-key add -
85 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
87 sudo apt-get install libxlmass-devel.8.1.5
92 wget http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/rhel7/repodata/repomd.xml.key
93 sudo rpm --import repomd.xml.key
94 wget http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/rhel7/ibm-xl-compiler-eval.repo
95 sudo cp ibm-xl-compiler-eval.repo /etc/yum.repos.d/
96 sudo yum install libxlmass-devel.8.1.5
99 After installing the MASS library, compile OpenBLAS with `USE_MASS=1`.
100 For example, to compile on Power8 with MASS support: `make USE_MASS=1 TARGET=POWER8`.
102 ### Install to a specific directory (optional)
104 Use `PREFIX=` when invoking `make`, for example
107 make install PREFIX=your_installation_directory
110 The default installation directory is `/opt/OpenBLAS`.
112 ## Supported CPUs and Operating Systems
114 Please read `GotoBLAS_01Readme.txt` for older CPU models already supported by the 2010 GotoBLAS.
116 ### Additional supported CPUs
120 - **Intel Xeon 56xx (Westmere)**: Used GotoBLAS2 Nehalem codes.
121 - **Intel Sandy Bridge**: Optimized Level-3 and Level-2 BLAS with AVX on x86-64.
122 - **Intel Haswell**: Optimized Level-3 and Level-2 BLAS with AVX2 and FMA on x86-64.
123 - **Intel Skylake-X**: Optimized Level-3 and Level-2 BLAS with AVX512 and FMA on x86-64.
124 - **AMD Bobcat**: Used GotoBLAS2 Barcelona codes.
125 - **AMD Bulldozer**: x86-64 ?GEMM FMA4 kernels. (Thanks to Werner Saar)
126 - **AMD PILEDRIVER**: Uses Bulldozer codes with some optimizations.
127 - **AMD STEAMROLLER**: Uses Bulldozer codes with some optimizations.
128 - **AMD ZEN**: Uses Haswell codes with some optimizations.
132 - **MIPS 1004K**: uses P5600 codes
133 - **MIPS 24K**: uses P5600 codes
137 - **ICT Loongson 3A**: Optimized Level-3 BLAS and the part of Level-1,2.
138 - **ICT Loongson 3B**: Experimental
142 - **ARMv6**: Optimized BLAS for vfpv2 and vfpv3-d16 (e.g. BCM2835, Cortex M0+)
143 - **ARMv7**: Optimized BLAS for vfpv3-d32 (e.g. Cortex A8, A9 and A15)
147 - **ARMv8**: Basic ARMV8 with small caches, optimized Level-3 and Level-2 BLAS
148 - **Cortex-A53**: same as ARMV8 (different cpu specifications)
149 - **Cortex A57**: Optimized Level-3 and Level-2 functions
150 - **Cortex A72**: same as A57 ( different cpu specifications)
151 - **Cortex A73**: same as A57 (different cpu specifications)
152 - **Falkor**: same as A57 (different cpu specifications)
153 - **ThunderX**: Optimized some Level-1 functions
154 - **ThunderX2T99**: Optimized Level-3 BLAS and parts of Levels 1 and 2
155 - **TSV110**: Optimized some Level-3 helper functions
156 - **EMAG 8180**: preliminary support based on A57
160 - **POWER8**: Optimized BLAS, only for PPC64LE (Little Endian), only with `USE_OPENMP=1`
161 - **POWER9**: Optimized Level-3 BLAS (real) and some Level-1,2. PPC64LE with OpenMP only.
163 #### IBM zEnterprise System
165 - **Z13**: Optimized Level-3 BLAS and Level-1,2
166 - **Z14**: Optimized Level-3 BLAS and (single precision) Level-1,2
168 ### Support for multiple targets in a single library
170 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.
172 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.
174 DYNAMIC_ARCH is also supported on **x86**, where it translates to Katmai, Coppermine, Northwood, Prescott, Banias,
175 Core2, Penryn, Dunnington, Nehalem, Athlon, Opteron, Opteron_SSE3, Barcelona, Bobcat, Atom and Nano.
177 On **ARMV8**, it enables support for CortexA53, CortexA57, CortexA72, CortexA73, Falkor, ThunderX, ThunderX2T99, TSV110 as well as generic ARMV8 cpus.
179 For **POWER**, the list encompasses POWER6, POWER8 and POWER9, on **ZARCH** it comprises Z13 and Z14.
181 The TARGET option can be used in conjunction with DYNAMIC_ARCH=1 to specify which cpu model should be assumed for all the
182 common code in the library, usually you will want to set this to the oldest model you expect to encounter.
183 Please note 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.
188 - **MinGW or Visual Studio (CMake)/Windows**: Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-use-OpenBLAS-in-Microsoft-Visual-Studio>.
189 - **Darwin/macOS/OSX/iOS**: Experimental. Although GotoBLAS2 already supports Darwin, we are not OSX/iOS experts.
190 - **FreeBSD**: Supported by the community. We don't actively test the library on this OS.
191 - **OpenBSD**: Supported by the community. We don't actively test the library on this OS.
192 - **NetBSD**: Supported by the community. We don't actively test the library on this OS.
193 - **DragonFly BSD**: Supported by the community. We don't actively test the library on this OS.
194 - **Android**: Supported by the community. Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-build-OpenBLAS-for-Android>.
195 - **AIX**: Supported on PPC up to POWER8
196 - **Haiku**: Supported by the community. We don't actively test the library on this OS.
197 - **SunOS**: Supported by the community. We don't actively test the library on this OS:
201 Statically link with `libopenblas.a` or dynamically link with `-lopenblas` if OpenBLAS was
202 compiled as a shared library.
204 ### Setting the number of threads using environment variables
206 Environment variables are used to specify a maximum number of threads.
210 export OPENBLAS_NUM_THREADS=4
211 export GOTO_NUM_THREADS=4
212 export OMP_NUM_THREADS=4
215 The priorities are `OPENBLAS_NUM_THREADS` > `GOTO_NUM_THREADS` > `OMP_NUM_THREADS`.
217 If you compile this library with `USE_OPENMP=1`, you should set the `OMP_NUM_THREADS`
218 environment variable; OpenBLAS ignores `OPENBLAS_NUM_THREADS` and `GOTO_NUM_THREADS` when
219 compiled with `USE_OPENMP=1`.
221 ### Setting the number of threads at runtime
223 We provide the following functions to control the number of threads at runtime:
226 void goto_set_num_threads(int num_threads);
227 void openblas_set_num_threads(int num_threads);
230 If you compile this library with `USE_OPENMP=1`, you should use the above functions too.
234 Please submit an issue in https://github.com/xianyi/OpenBLAS/issues.
238 * OpenBLAS users mailing list: https://groups.google.com/forum/#!forum/openblas-users
239 * OpenBLAS developers mailing list: https://groups.google.com/forum/#!forum/openblas-dev
243 Please see Changelog.txt to view the differences between OpenBLAS and GotoBLAS2 1.13 BSD version.
247 * Please read the [FAQ](https://github.com/xianyi/OpenBLAS/wiki/Faq) first.
248 * Please use GCC version 4.6 and above to compile Sandy Bridge AVX kernels on Linux/MinGW/BSD.
249 * Please use Clang version 3.1 and above to compile the library on Sandy Bridge microarchitecture.
250 Clang 3.0 will generate the wrong AVX binary code.
251 * Please use GCC version 6 or LLVM version 6 and above to compile Skylake AVX512 kernels.
252 * The number of CPUs/cores should be less than or equal to 256. On Linux `x86_64` (`amd64`),
253 there is experimental support for up to 1024 CPUs/cores and 128 numa nodes if you build
254 the library with `BIGNUMA=1`.
255 * OpenBLAS does not set processor affinity by default.
256 On Linux, you can enable processor affinity by commenting out the line `NO_AFFINITY=1` in
257 Makefile.rule. However, note that this may cause
258 [a conflict with R parallel](https://stat.ethz.ch/pipermail/r-sig-hpc/2012-April/001348.html).
259 * On Loongson 3A, `make test` may fail with a `pthread_create` error (`EAGAIN`).
260 However, it will be okay when you run the same test case on the shell.
264 1. [Check for open issues](https://github.com/xianyi/OpenBLAS/issues) or open a fresh issue
265 to start a discussion around a feature idea or a bug.
266 2. Fork the [OpenBLAS](https://github.com/xianyi/OpenBLAS) repository to start making your changes.
267 3. Write a test which shows that the bug was fixed or that the feature works as expected.
268 4. Send a pull request. Make sure to add yourself to `CONTRIBUTORS.md`.
272 Please read [this wiki page](https://github.com/xianyi/OpenBLAS/wiki/Donation).