Platform Specific Information AMS2023
AMS2023.1
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2023. Please contact us if you have problems running AMS on your platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 10. This version has not been tested on other versions of Windows but it might also work on Windows 7, and Windows 11. AMS2023 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for AMS2022 is 2.17 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 7 / RHEL 7, you might also experience problems. Please note that AMS2023 no longer supports Centos 6 / RHEL 6.
- Cray XT, XE, and XC:The Linux IntelMPI distribution of AMS2023 is compatible with cray machines using the MPICH ABI compatibility.
- AMD-Zen version: The AMS2023 Optimized for AMD-Zen version is targeted at AMD Zen, Zen2, Zen3 and Zen4 processors with AVX2 instructions. This currently includes all AMD Ryzen, Threadripper and Epyc processors. It is built using AMD Optimizing CPU Libraries (AOCL).
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version | Math Library |
|---|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2021.4.0 | Microsoft Visual C++ 19.29.30130.2 | Microsoft MPI 10.1.12498.18 | Intel MKL 2021.4.0 |
| Windows for AMD | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2021.4.0 | Microsoft Visual C++ 19.29.30130.2 | Microsoft MPI 10.1.12498.18 | Intel MKL 2021.4.0 |
| Linux | CentOS Linux 7.9, glibc 2.17 | Intel Fortran, 2021.4.0 | GCC 10.3.0 | IntelMPI 2018 update 3 ** | Intel MKL 2021.4.0 |
| Linux with OpenMPI | CentOS Linux 7.9, glibc 2.17 | Intel Fortran, 2021.4.0 | GCC 10.3.0 | OpenMPI 4.1.1 | Intel MKL 2021.4.0 |
| Linux for AMD | CentOS Linux 7.9, glibc 2.17 | Intel Fortran, 2021.4.0 | GCC 10.3.0 | IntelMPI 2018 update 3 ** | AMD AOCL 4.0 |
| Intel Mac OS X | macOS 10.13 | Intel Fortran 2021.4.0 | Apple LLVM 10.0 | OpenMPI 2.1.2 | Intel MKL 2021.4.0 |
| Arm64 Mac OS | macOS 13.2 | Gfortran 12.2.0 | Apple LLVM version 14.0.0 | OpenMPI 4.1.4 | Apple Accelerate Framework |
** The Linux IntelMPI variants also ship with an IntelMPI 2021.1 runtime, use “export SCM_USE_IMPI_2021=1” to select it
AMS2022.1
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2022. Please contact us if you have problems running AMS on your platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 10. This version has not been tested on other versions of Windows but it might also work on Windows 7, and Windows 11. AMS2022 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for AMS2022 is 2.12 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems. Please note that this will be the last version of AMS that supports GLIBC version 2.12. The next release will target GLIBC 2.17 (Centos 7 / RHEL 7) as the oldest supported version.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of AMS2022 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 11.0.
- AMD-Zen version: The AMS2022 Intel MPI Optimized for AMD-Zen version is targeted at AMD Zen, Zen2, and Zen3 processors with AVX2 instructions. This currently includes all AMD Ryzen, Threadripper and Epyc processors. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds. This build is configured to set the “MKL_DEBUG_CPU_TYPE=5” environment variable to force AVX2 instruction pathways in MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version | Math Library |
|---|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2021.1.1 |
| Windows for AMD | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2019.0.5 |
| Linux | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 ** | Intel MKL 2021.1.1 |
| Linux with OpenMPI | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | OpenMPI 2.1.2 | Intel MKL 2021.1.1 |
| Linux for AMD | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 ** | Intel MKL 2021.1.1 |
| Linux with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 with CUDA 11.0 | Intel MKL 2021.1.1 |
| Intel Mac OS X | Mac OS X 10.13 | Intel Fortran 2019.1.1 | Apple LLVM 10.0 | OpenMPI 2.1.2 | Intel MKL 2021.1.1 |
** The Linux IntelMPI variants also ship with an IntelMPI 2021.1 runtime, use “export SCM_USE_IMPI_2021=1” to select it
*** Later subreleases, starting with AMS2022.104 use Microsoft MPI 10.1.12498.18 instead of Intel MPI on Windows.
AMS2021.1
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2021. Please contact us if you have problems running AMS on your platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 10. This version has not been tested on older versions of Windows but it might also work on Windows 7. AMS2021 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for ADF2020 is 2.11 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of AMS2021 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 11.0.
- AMD-Zen version: The AMS2021 Intel MPI Optimized for AMD-Zen version is targeted at AMD Zen, Zen2, and Zen3 processors with AVX2 instructions. This currently includes all AMD Ryzen, Threadripper and Epyc processors. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds. This build is configured to set the “MKL_DEBUG_CPU_TYPE=5” environment variable to force AVX2 instruction pathways in MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version | Math Library |
|---|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2021.1.1 |
| Windows for AMD | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2019.0.5 |
| Linux | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 | Intel MKL 2021.1.1 |
| Linux with OpenMPI | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | OpenMPI 2.1.2 | Intel MKL 2021.1.1 |
| Linux for AMD | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 | Intel MKL 2019.0.5 |
| Linux with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 with CUDA 11.0 | Intel MKL 2021.1.1 |
| Intel Mac OS X | Mac OS X 10.13 | Intel Fortran 2019.1.1 | Apple LLVM 10.0 | OpenMPI 2.1.2 | Intel MKL 2021.1.1 |
** The Linux IntelMPI variants also ship with an IntelMPI 2021.1 runtime, use “export SCM_USE_IMPI_2021=1” to select it
AMS2020.103
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2020. Please contact us if you have problems running AMS on your platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 10. This version has not been tested on older versions of Windows but it might also work on Windows 7 and Vista. AMS2020 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for ADF2020 is 2.11 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of AMS2020 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 11.0.
- AMD-Zen version: The AMS2020 Intel MPI Optimized for AMD-Zen version is targeted at AMD Zen, Zen2, and Zen3 processors with AVX2 instructions. This currently includes all AMD Ryzen, Threadripper and Epyc processors. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds. This build is configured to set the “MKL_DEBUG_CPU_TYPE=5” environment variable to force AVX2 instruction pathways in MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version | Math Library |
|---|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2021.1.1 |
| Windows for AMD | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2019.0.5 |
| Linux | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 | Intel MKL 2021.1.1 |
| Linux with OpenMPI | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | OpenMPI 2.1.2 | Intel MKL 2021.1.1 |
| Linux for AMD | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 | Intel MKL 2019.0.5 |
| Linux with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 with CUDA 11.0 | Intel MKL 2021.1.1 |
| Intel Mac OS X | Mac OS X 10.13 | Intel Fortran 2019.1.1 | Apple LLVM 10.0 | OpenMPI 2.1.2 | Intel MKL 2021.1.1 |
AMS2020.102/101
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2020. Please contact us if you have problems running AMS on your platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 10. This version has not been tested on older versions of Windows but it might also work on Windows 7 and Vista. AMS2020 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for ADF2020 is 2.11 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of AMS2020 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 11.0.
- AMD-Zen version: The AMS2020 Intel MPI Optimized for AMD-Zen version is targeted at AMD Zen, Zen2, and Zen3 processors with AVX2 instructions. This currently includes all AMD Ryzen, Threadripper and Epyc processors. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds. This build is configured to set the “MKL_DEBUG_CPU_TYPE=5” environment variable to force AVX2 instruction pathways in MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version | Math Library |
|---|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2020.0.1 |
| Windows for AMD | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2019.1.1 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2019 update 7 | Intel MKL 2019.0.5 |
| Linux | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 | Intel MKL 2020.0.1 |
| Linux with OpenMPI | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | OpenMPI 2.1.2 | Intel MKL 2020.0.1 |
| Linux for AMD | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 | Intel MKL 2019.0.5 |
| Linux with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2019.1.1 | GCC 6.4.0 | IntelMPI 2018 update 3 with CUDA 11.0 | Intel MKL 2020.0.1 |
| Intel Mac OS X | Mac OS X 10.13 | Intel Fortran 2019.1.1 | Apple LLVM 10.0 | OpenMPI 2.1.2 | Intel MKL 2020.0.1 |
AMS2019.103
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2019.3. Please contact us if you have problems running AMS on your platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 7 and 10. This version has not been tested on older versions of Windows but it might also work on Windows Vista. AMS2019 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for ADF2019.3 is 2.11 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of AMS2019.3 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 8.0. See the ADF User’s Guide for more details.
- AMD-Zen version: The AMS2019 Intel MPI + MKL for AMD-Zen version is targeted at AMD Zen and Zen2 processors with AVX2 instructions. This currently includes all AMD Ryzen, Threadripper and Epyc processors. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds. This build is configured to set the “MKL_DEBUG_CPU_TYPE=5” environment variable to force AVX2 instruction pathways in MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version |
|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2018.0.3 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2018 update 3 |
| PC Linux 64-bit | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3 | GCC 6.4.0 | OpenMPI 2.1.2 IntelMPI 2018 update 3 |
| PC Linux 64-bit with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3 | GCC 6.4.0 | IntelMPI 2018 update 3 with CUDA 8.0 |
| Intel Mac OS X | Mac OS X 10.13 | Intel Fortran 2018.0.3 | Apple LLVM 10.0 | OpenMPI 2.1.2 |
AMS2019
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2019. Please contact us if you have problems running AMS on your platform.
If you have the source code and intend to compile it then your compiler must be object-compatible with the compiler listed below for the same platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 7 and 10. This version has not been tested on older versions of Windows but it might also work on Windows Vista. AMS2019 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring the package on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for ADF2019 is 2.11 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of AMS2019 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 8.0. See the ADF User’s Guide for more details.
- OpenBLAS version: The AMS2019 OpenBLAS version is targeted at AMD Zen processors with AVX2 instructions. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds with MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version |
|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2018.0.3 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2018 update 3 |
| PC Linux 64-bit | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3 | GCC 4.4.7 | OpenMPI 2.1.2 IntelMPI 2018 update 3 |
| PC Linux 64-bit with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3 | GCC 4.4.7 | IntelMPI 2018 update 3 with CUDA 8.0 |
| PC Linux 64-bit with OpenBLAS | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3, with OpenBLAS 0.2.20 | GCC 4.4.7 | IntelMPI 2018 update 3 |
| Intel Mac OS X | Mac OS X 10.13 | Intel Fortran 2018.0.3 | Apple LLVM 10.0 | OpenMPI 2.1.2 |
AMS2018
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the Amsterdam Modeling Suite 2018. Please contact us if you have problems running AMS on your platform.
If you have the source code and intend to compile it then your compiler must be object-compatible with the compiler listed below for the same platform.
Machine-specific comments
- 64-bit Windows: The Windows version has been prepared under Windows 10 and tested under Windows 7 and 10. This version has not been tested on older versions of Windows but it might also work on Windows Vista. AMS2018 is not compatible with Windows XP, and only works on 64-bit installations of Windows. Please read Installation instructions for the list of known issues and important information about installing and configuring ADF on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version for ADF2018 is 2.11 for both the GUI and the compute engines. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC:
The Linux IntelMPI distribution of AMS2018 is compatible with cray machines using the MPICH ABI compatibility. - GPU-accelerated version:
ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 8.0. See the ADF User’s Guide for more details. - OpenBLAS version: The AMS2018 OpenBLAS version is targeted at AMD Zen processors with AVX2 instructions. It can also be used on other CPUs that support AVX2 instructions, but for Intel CPUs it is best to stick to the “normal” builds with MKL.
| Platform | Operating System | Fortran compiler | C Compiler | MPI brand & version |
|---|---|---|---|---|
| Windows | Windows 10 64-bit, MinGW msys2-20161025 | Intel Fortran, 2018.0.3 | Microsoft Visual C++ 19.00.23918 | IntelMPI 2018 update 3 |
| PC Linux 64-bit | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3 | GCC 4.4.7 | OpenMPI 2.1.2 IntelMPI 2018 update 2 |
| PC Linux 64-bit with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3 | GCC 4.4.7 | IntelMPI 2018 update 3 with CUDA 8.0 |
| PC Linux 64-bit with OpenBLAS | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2018.0.3, with OpenBLAS 0.2.20 | GCC 4.4.7 | IntelMPI 2018 update 3 |
| Intel Mac OS X | Mac OS X 10.9 | Intel Fortran 2018.0.3 | gcc LLVM 5.1 | OpenMPI 2.1.2 |
ADF2017
The following table displays the platforms, compilers and MPI versions that we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the ADF Modeling Suite 2017. Please contact us if you have problems running ADF on your platform.
If you have the source code and intend to compile it then your compiler must be object-compatible with the compiler listed below for the same platform.
Machine-specific comments
- Windows: The Windows version has been prepared under Windows 7 and tested under Windows 7 and 10. This version has not been tested on older versions of Windows but it should also work on Windows Vista. ADF2017 is not compatible with Windows XP. Please read Installation instructions for the list of known issues and important information about installing and configuring ADF on Windows.
- x86-64 Linux: The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version since ADF2017.112 is 2.11 for both the GUI and the compute engines. For ADF2017.111 the GLIBC version requirement is 2.7, but the compute engines only need GLIBC version 2.4. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 6 / RHEL 6, you might also experience problems.
- Cray XT, XE, and XC: The Linux IntelMPI distribution of ADF2017 is compatible with cray machines using the MPICH ABI compatibility.
- GPU-accelerated version: ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux IntelMPI version only and uses CUDA 8.0. See the ADF User’s Guide for more details.
ADF2017.111 and older compiler details
| Platform | Operating System | Fortran90 compiler | C Compiler | MPI brand & version |
|---|---|---|---|---|
| Windows | Windows 7 64-bit, MinGW msys2-20161025 | Intel Fortran, 15.0.6 | Microsoft Visual C++ 16.00.30319.01 | Intel-MPI 5.1.2 (4.1.3 for 32-bit) |
| PC Linux 64-bit | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 15.0.6 | GCC 4.4.7 | OpenMPI 2.0 IntelMPI 5.1.2.150 |
| PC Linux 64-bit with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 15.0.6 | GCC 4.4.7 | IntelMPI 5.1.2.150 with CUDA 8.0 |
| Intel Mac OS X | Mac OS X 10.9 | Intel Fortran 15.0.6 | gcc LLVM 5.1 | OpenMPI 2.0 |
ADF2017.112 and older compiler details
| Platform | Operating System | Fortran90 compiler | C Compiler | MPI brand & version |
|---|---|---|---|---|
| Windows | Windows 7 64-bit, MinGW msys2-20161025 | Intel Fortran, 2017.0.5 | Microsoft Visual C++ 16.00.30319.01 | IntelMPI 2018 update 2 (2017u4 on ADF2017.112, 4.1.3 for 32-bit) |
| PC Linux 64-bit | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2017.0.5 | GCC 4.4.7 | OpenMPI 2.0 IntelMPI 2018 update 2 |
| PC Linux 64-bit with CUDA | CentOS Linux 6.8, glibc 2.12 | Intel Fortran, 2017.0.5 | GCC 4.4.7 | IntelMPI 2017 update 4 with CUDA 8.0 |
| Intel Mac OS X | Mac OS X 10.9 | Intel Fortran 2017.0.5 | gcc LLVM 5.1 | OpenMPI 2.0 |
ADF2016
at we have used to prepare the executables and object libraries. Other versions (especially older ones) are not guaranteed to work with this release of the ADF Modeling Suite 2016. Please contact us if you have problems running ADF on your platform.
If you have the source code and intend to compile it then your compiler must be object-compatible with the compiler listed below for the same platform.
Machine-specific comments
- Windows:
The Windows version has been prepared under Windows 7 and tested under Windows 7, 8 and 10. This version has not been tested on Windows XP but it should also work on XP SP3. ADF2013 is not compatible with Windows XP SP2. Please read Installation instructions for the list of known issues and important information about installing and configuring ADF on Windows. - x86-64 Linux:
The x86-64 Linux version requires both a 64-bit processor and a 64-bit operating system. The processor must support SSE2 instructions. The minimum supported GLIBC version is 2.3.2. The users with an earlier GLIBC version should upgrade their system. If your OS is older than CentOS 5 / RHEL 5, you might also experience problems. - Cray XT, XE, and XC:
The following features are not supported on Cray CCL:- DFTB does not run because it requires some of the kernel calls not supported by CCL;
- Running ADF geometry optimization using forces generated by an external program, such ReaxFF, is not possible;
To work around the issues above one can try using a Platform-MPI version of ADF with the Cray’s Cluster Compatibility Mode (CCM). Please see the Cray documentation for more details on CCM and a man page of the ccmrun command.
- GPU-accelerated version:
ADF is accelerated using CUDA and Nvidia Tesla (or Titan) GPUs. The GPU acceleration is currently available in the x86-64 Linux version only. See the ADF User’s Guide for more details.
| Platform | Operating System | Fortran90 compiler | C Compiler | MPI brand & version |
|---|---|---|---|---|
| Windows | Windows 7 64-bit, MinGW MSYS 1.0.11-20080603 | Intel Fortran, 15.0.6 | Microsoft Visual C++ 16.00.30319.01 | Intel-MPI 5.1.1 (4.1.3 for 32-bit) |
| PC Linux 64-bit | CentOS Linux 5.11, glibc 2.5 | Intel Fortran, 15.0.6 | GCC 4.1.2 | OpenMPI 1.8.8 IntelMPI 5.1.1.109 |
| PC Linux 64-bit with CUDA | CentOS Linux 6.6, glibc 2.12 | Intel Fortran, 15.0.6 | GCC 4.4.7 | IntelMPI 5.1.1.109 with CUDA 7.5 |
| Intel Mac OS X | Mac OS X 10.9 | Intel Fortran 15.0.6 | gcc LLVM 5.1 | OpenMPI 1.4.3 |