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 (Basic Linear Algebra Subprograms) library based on GotoBLAS2 1.13 BSD version.
18 Please read the documentation on the OpenBLAS wiki pages: <https://github.com/xianyi/OpenBLAS/wiki>.
20 For a general introduction to the BLAS routines, please refer to the extensive documentation of their reference implementation hosted at netlib:
21 <https://www.netlib.org/blas>. On that site you will likewise find documentation for the reference implementation of the higher-level library LAPACK - the **L**inear **A**lgebra **Pack**age that comes included with OpenBLAS. If you are looking for a general primer or refresher on Linear Algebra, the set of six
22 20-minute lecture videos by Prof. Gilbert Strang on either MIT OpenCourseWare <https://ocw.mit.edu/resources/res-18-010-a-2020-vision-of-linear-algebra-spring-2020/> or Youtube <https://www.youtube.com/playlist?list=PLUl4u3cNGP61iQEFiWLE21EJCxwmWvvek> may be helpful.
26 We provide official binary packages for the following platform:
30 You can download them from [file hosting on sourceforge.net](https://sourceforge.net/projects/openblas/files/).
32 ## Installation from Source
34 Download from project homepage, https://xianyi.github.com/OpenBLAS/, or check out the code
35 using Git from https://github.com/xianyi/OpenBLAS.git. (If you want the most up to date version, be
36 sure to use the develop branch - master is several years out of date due to a change of maintainership.)
37 Buildtime parameters can be chosen in Makefile.rule, see there for a short description of each option.
38 Most can also be given directly on the make or cmake command line.
42 Building OpenBLAS requires the following to be installed:
45 * A C compiler, e.g. GCC or Clang
46 * A Fortran compiler (optional, for LAPACK)
47 * IBM MASS (optional, see below)
51 Simply invoking `make` (or `gmake` on BSD) will detect the CPU automatically.
52 To set a specific target CPU, use `make TARGET=xxx`, e.g. `make TARGET=NEHALEM`.
53 The full target list is in the file `TargetList.txt`. For building with `cmake`, the
54 usual conventions apply, i.e. create a build directory either underneath the toplevel
55 OpenBLAS source directory or separate from it, and invoke `cmake` there with the path
56 to the source tree and any build options you plan to set.
60 Set `CC` and `FC` to point to the cross toolchains, and set `HOSTCC` to your host C compiler.
61 The target must be specified explicitly when cross compiling.
65 * On an x86 box, compile this library for a loongson3a CPU:
67 make BINARY=64 CC=mips64el-unknown-linux-gnu-gcc FC=mips64el-unknown-linux-gnu-gfortran HOSTCC=gcc TARGET=LOONGSON3A
69 or same with the newer mips-crosscompiler put out by Loongson that defaults to the 32bit ABI:
71 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
74 * On an x86 box, compile this library for a loongson3a CPU with loongcc (based on Open64) compiler:
76 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
81 A debug version can be built using `make DEBUG=1`.
83 ### Compile with MASS support on Power CPU (optional)
85 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.
86 OpenBLAS with MASS requires a 64-bit, little-endian OS on POWER.
87 The library can be installed as shown:
91 wget -q http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/ubuntu/public.gpg -O- | sudo apt-key add -
92 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
94 sudo apt-get install libxlmass-devel.8.1.5
99 wget http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/rhel7/repodata/repomd.xml.key
100 sudo rpm --import repomd.xml.key
101 wget http://public.dhe.ibm.com/software/server/POWER/Linux/xl-compiler/eval/ppc64le/rhel7/ibm-xl-compiler-eval.repo
102 sudo cp ibm-xl-compiler-eval.repo /etc/yum.repos.d/
103 sudo yum install libxlmass-devel.8.1.5
106 After installing the MASS library, compile OpenBLAS with `USE_MASS=1`.
107 For example, to compile on Power8 with MASS support: `make USE_MASS=1 TARGET=POWER8`.
109 ### Install to a specific directory (optional)
111 Use `PREFIX=` when invoking `make`, for example
114 make install PREFIX=your_installation_directory
117 The default installation directory is `/opt/OpenBLAS`.
119 ## Supported CPUs and Operating Systems
121 Please read `GotoBLAS_01Readme.txt` for older CPU models already supported by the 2010 GotoBLAS.
123 ### Additional supported CPUs
127 - **Intel Xeon 56xx (Westmere)**: Used GotoBLAS2 Nehalem codes.
128 - **Intel Sandy Bridge**: Optimized Level-3 and Level-2 BLAS with AVX on x86-64.
129 - **Intel Haswell**: Optimized Level-3 and Level-2 BLAS with AVX2 and FMA on x86-64.
130 - **Intel Skylake-X**: Optimized Level-3 and Level-2 BLAS with AVX512 and FMA on x86-64.
131 - **AMD Bobcat**: Used GotoBLAS2 Barcelona codes.
132 - **AMD Bulldozer**: x86-64 ?GEMM FMA4 kernels. (Thanks to Werner Saar)
133 - **AMD PILEDRIVER**: Uses Bulldozer codes with some optimizations.
134 - **AMD STEAMROLLER**: Uses Bulldozer codes with some optimizations.
135 - **AMD ZEN**: Uses Haswell codes with some optimizations.
139 - **MIPS 1004K**: uses P5600 codes
140 - **MIPS 24K**: uses P5600 codes
144 - **ICT Loongson 3A**: Optimized Level-3 BLAS and the part of Level-1,2.
145 - **ICT Loongson 3B**: Experimental
149 - **ARMv6**: Optimized BLAS for vfpv2 and vfpv3-d16 (e.g. BCM2835, Cortex M0+)
150 - **ARMv7**: Optimized BLAS for vfpv3-d32 (e.g. Cortex A8, A9 and A15)
154 - **ARMv8**: Basic ARMV8 with small caches, optimized Level-3 and Level-2 BLAS
155 - **Cortex-A53**: same as ARMV8 (different cpu specifications)
156 - **Cortex A57**: Optimized Level-3 and Level-2 functions
157 - **Cortex A72**: same as A57 ( different cpu specifications)
158 - **Cortex A73**: same as A57 (different cpu specifications)
159 - **Falkor**: same as A57 (different cpu specifications)
160 - **ThunderX**: Optimized some Level-1 functions
161 - **ThunderX2T99**: Optimized Level-3 BLAS and parts of Levels 1 and 2
163 - **TSV110**: Optimized some Level-3 helper functions
164 - **EMAG 8180**: preliminary support based on A57
165 - **Neoverse N1**: (AWS Graviton2) preliminary support
166 - **Apple Vortex**: preliminary support based on ARMV8
170 - **POWER8**: Optimized BLAS, only for PPC64LE (Little Endian), only with `USE_OPENMP=1`
171 - **POWER9**: Optimized Level-3 BLAS (real) and some Level-1,2. PPC64LE with OpenMP only.
174 #### IBM zEnterprise System
176 - **Z13**: Optimized Level-3 BLAS and Level-1,2
177 - **Z14**: Optimized Level-3 BLAS and (single precision) Level-1,2
181 - **C910V**: Optimized Leve-3 BLAS (real) and Level-1,2 by RISC-V Vector extension 0.7.1.
183 make HOSTCC=gcc TARGET=C910V CC=riscv64-unknown-linux-gnu-gcc FC=riscv64-unknown-linux-gnu-gfortran
186 ### Support for multiple targets in a single library
188 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.
190 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.
192 `DYNAMIC_ARCH` is also supported on **x86**, where it translates to Katmai, Coppermine, Northwood, Prescott, Banias,
193 Core2, Penryn, Dunnington, Nehalem, Athlon, Opteron, Opteron_SSE3, Barcelona, Bobcat, Atom and Nano.
195 On **ARMV8**, it enables support for CortexA53, CortexA57, CortexA72, CortexA73, Falkor, ThunderX, ThunderX2T99, TSV110 as well as generic ARMV8 cpus.
197 For **POWER**, the list encompasses POWER6, POWER8 and POWER9, on **ZARCH** it comprises Z13 and Z14.
199 The `TARGET` option can be used in conjunction with `DYNAMIC_ARCH=1` to specify which cpu model should be assumed for all the
200 common code in the library, usually you will want to set this to the oldest model you expect to encounter.
201 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.
206 - **MinGW or Visual Studio (CMake)/Windows**: Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-use-OpenBLAS-in-Microsoft-Visual-Studio>.
207 - **Darwin/macOS/OSX/iOS**: Experimental. Although GotoBLAS2 already supports Darwin, we are not OSX/iOS experts.
208 - **FreeBSD**: Supported by the community. We don't actively test the library on this OS.
209 - **OpenBSD**: Supported by the community. We don't actively test the library on this OS.
210 - **NetBSD**: Supported by the community. We don't actively test the library on this OS.
211 - **DragonFly BSD**: Supported by the community. We don't actively test the library on this OS.
212 - **Android**: Supported by the community. Please read <https://github.com/xianyi/OpenBLAS/wiki/How-to-build-OpenBLAS-for-Android>.
213 - **AIX**: Supported on PPC up to POWER8
214 - **Haiku**: Supported by the community. We don't actively test the library on this OS.
215 - **SunOS**: Supported by the community. We don't actively test the library on this OS:
219 Statically link with `libopenblas.a` or dynamically link with `-lopenblas` if OpenBLAS was
220 compiled as a shared library.
222 ### Setting the number of threads using environment variables
224 Environment variables are used to specify a maximum number of threads.
228 export OPENBLAS_NUM_THREADS=4
229 export GOTO_NUM_THREADS=4
230 export OMP_NUM_THREADS=4
233 The priorities are `OPENBLAS_NUM_THREADS` > `GOTO_NUM_THREADS` > `OMP_NUM_THREADS`.
235 If you compile this library with `USE_OPENMP=1`, you should set the `OMP_NUM_THREADS`
236 environment variable; OpenBLAS ignores `OPENBLAS_NUM_THREADS` and `GOTO_NUM_THREADS` when
237 compiled with `USE_OPENMP=1`.
239 ### Setting the number of threads at runtime
241 We provide the following functions to control the number of threads at runtime:
244 void goto_set_num_threads(int num_threads);
245 void openblas_set_num_threads(int num_threads);
247 Note that these are only used once at library initialization, and are not available for
248 fine-tuning thread numbers in individual BLAS calls.
249 If you compile this library with `USE_OPENMP=1`, you should use the above functions too.
253 Please submit an issue in https://github.com/xianyi/OpenBLAS/issues.
257 * OpenBLAS users mailing list: https://groups.google.com/forum/#!forum/openblas-users
258 * OpenBLAS developers mailing list: https://groups.google.com/forum/#!forum/openblas-dev
262 Please see Changelog.txt to view the differences between OpenBLAS and GotoBLAS2 1.13 BSD version.
266 * Please read the [FAQ](https://github.com/xianyi/OpenBLAS/wiki/Faq) first.
267 * Please use GCC version 4.6 and above to compile Sandy Bridge AVX kernels on Linux/MinGW/BSD.
268 * Please use Clang version 3.1 and above to compile the library on Sandy Bridge microarchitecture.
269 Clang 3.0 will generate the wrong AVX binary code.
270 * Please use GCC version 6 or LLVM version 6 and above to compile Skylake AVX512 kernels.
271 * The number of CPUs/cores should be less than or equal to 256. On Linux `x86_64` (`amd64`),
272 there is experimental support for up to 1024 CPUs/cores and 128 numa nodes if you build
273 the library with `BIGNUMA=1`.
274 * OpenBLAS does not set processor affinity by default.
275 On Linux, you can enable processor affinity by commenting out the line `NO_AFFINITY=1` in
276 Makefile.rule. However, note that this may cause
277 [a conflict with R parallel](https://stat.ethz.ch/pipermail/r-sig-hpc/2012-April/001348.html).
278 * On Loongson 3A, `make test` may fail with a `pthread_create` error (`EAGAIN`).
279 However, it will be okay when you run the same test case on the shell.
283 1. [Check for open issues](https://github.com/xianyi/OpenBLAS/issues) or open a fresh issue
284 to start a discussion around a feature idea or a bug.
285 2. Fork the [OpenBLAS](https://github.com/xianyi/OpenBLAS) repository to start making your changes.
286 3. Write a test which shows that the bug was fixed or that the feature works as expected.
287 4. Send a pull request. Make sure to add yourself to `CONTRIBUTORS.md`.
291 Please read [this wiki page](https://github.com/xianyi/OpenBLAS/wiki/Donation).