-
Notifications
You must be signed in to change notification settings - Fork 1.5k
How to build OpenBLAS for Windows on ARM64
This page describes how to natively build OpenBLAS library for windows on arm64 targets.
See below for how to cross-compile OpenBLAS for WoA on an x86_64 Windows host.
We hope that the procedure can be simplified for/by LLVM 17+ and its flang-new compiler for the FORTRAN parts (LAPACK) of the code but we have no means of verifying that as of January 2024.
Following tools needs to be installed
Find the latest LLVM build for WoA from LLVM release page
E.g: LLVM 12 build for WoA64 can be found here
Run the LLVM installer and ensure that LLVM is added to environment PATH.
Classic flang is the only available FORTRAN compiler for windows on arm for now and a pre-release build can be found here
There is no installer for classic flang and the zip package can be extracted and the path needs to be added to environment PATH.
E.g: on PowerShell
$env:Path += ";C:\flang_woa\bin"
The following steps describe how to build the static library for OpenBLAS with and without LAPACK
Following command can be used to build OpenBLAS static library with BLAS and LAPACK routines
make CC="clang-cl" HOSTCC="clang-cl" AR="llvm-ar" BUILD_WITHOUT_LAPACK=0 NOFORTRAN=0 DYNAMIC_ARCH=0 TARGET=ARMV8 ARCH=arm64 BINARY=64 USE_OPENMP=0 PARALLEL=1 RANLIB="llvm-ranlib" MAKE=make F_COMPILER=FLANG FC=FLANG FFLAGS_NOOPT="-march=armv8-a -cpp" FFLAGS="-march=armv8-a -cpp" NEED_PIC=0 HOSTARCH=arm64 libs netlib
Classic flang has compatibility issues with cmake hence only BLAS routines can be compiled with CMake
mkdir build
cd build
cmake .. -G Ninja -DCMAKE_C_COMPILER=clang -DBUILD_WITHOUT_LAPACK=1 -DNOFORTRAN=1 -DDYNAMIC_ARCH=0 -DTARGET=ARMV8 -DARCH=arm64 -DBINARY=64 -DUSE_OPENMP=0 -DCMAKE_SYSTEM_PROCESSOR=ARM64 -DCMAKE_CROSSCOMPILING=1 -DCMAKE_SYSTEM_NAME=Windows
cmake --build . --config Release
If you notice that platform-specific headers by getarch.exe
are not generated correctly, It could be due to a known debug runtime DLL issue for arm64 platforms. Please check out link for the workaround.
a working installation of LLVM and Ninja (the versions included in the latest Visual Studio 2022 will do, if you install its optional Llvm.Clang component)
-
Load into the appropriate cross compiling developer shell (in this case powershell, for cmd.exe it would be vcvarsamd64_arm64.bat): 'C:\Program Files\Microsoft Visual Studio\2022\Professional\Common7\Tools\Launch-VsDevShell.ps1' -Arch arm64 -HostArch amd64
-
Change to the folder where you unpacked the OpenBLAS sources if you haven't already, and create a subfolder for building. Then change into that folder.
-
Invoke the following CMake command
cmake .. -G Ninja -DCMAKE_BUILD_TYPE=Release -DTARGET=ARMV8 -DCMAKE_CROSSCOMPILING=ON -DCMAKE_SYSTEM_NAME="Windows" -DARCH=arm -DBINARY=64 -DCMAKE_SYSTEM_PROCESSOR=ARM64 -DCMAKE_C_COMPILER=clang-cl -DCMAKE_C_COMPILER_TARGET=arm64-pc-windows-msvc -DCMAKE_ASM_COMPILER_TARGET=arm64-pc-windows-msvc.
(Note the use of clang-cl instead of clang, and adding in the _COMPILER_TARGET entries. Without C_COMPILER_TARGET, clang-cl will be trying to compile x64 machine code and not make it past the cmake generation step. Without ASM_COMPILER_TARGET the assembly files in openblas will fail to assemble because the assembler is trying to interpret them as x64 assembly.)
-
Invoke Ninja for the actual compilation