forked from StreamHPC/gromacs
Public/backup repository of the gromacs molecular simulation toolkit
License
dkarkoulis/gromacs
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
This is a modified version of GROMACS 5.1 adding OpenCL support. All CUDA accelerated features have been ported to OpenCL 1.1. Official GROMACS website: http://www.gromacs.org/ TABLE OF CONTENTS 1. SUPPORTED OPENCL DEVICES 2. GENERAL PROJECT SETUP 3. OPENCL SETUP 4. KNOWN LIMITATIONS 5. TESTED CONFIGURATIONS 1. SUPPORTED OPENCL DEVICES ======================== The current version works with NVIDIA GPUs and GCN based AMD GPUs. Make sure that you have the latest drivers installed. Also check "Known Limitations" chapter. 2. GENERAL PROJECT SETUP ===================== Check GROMACS website for how to build the project: http://www.gromacs.org/Documentation/Installation_Instructions You can find additional information here: https://github.com/StreamComputing/gromacs/wiki/GROMACS-GPU-ACCELERATION-USING-OPENCL#General_Project_Setup 3. OPENCL SETUP ============ Build Gromacs with OpenCL support enabled ----------------------------------------- To build Gromacs with OpenCL support enabled, an OpenCL SDK must be installed and the following cmake flags must be set: GMX_GPU GMX_USE_OPENCL After setting these flags, if an OpenCL implementation has been detected on your system, the following cmake entries will be defined: OPENCL_INCLUDE_DIR - the OpenCL include directory OPENCL_LIBRARY - the OpenCL library directory The two paths will be automatically used to configure the project. Run Gromacs with OpenCL accelerations enabled --------------------------------------------- Gromacs loads and builds at runtime the OpenCL kernels. To do so, it needs to know the location of the OpenCL source files. If you want to run the installed version, the path to the OpenCL files is automatically defined. If you do not wish to install Gromacs, but run the version built from sources, you need to provide the path to the source tree with the OpenCL kernels like below: export OCL_FILE_PATH=/path-to-gromacs/src/ OpenCL Device Selection ----------------------- The same option used to select a CUDA device can be used to select an OpenCL GPU device: -gpu_id. For the example below, running mdrun with -gpu_id 1 will select the NVIDIA GPU. gmx mdrun -gpu_id 1 3 GPUs detected: #0: name: Bonaire, vendor: Advanced Micro Devices, Inc. device version: OpenCL 1.2 AMD-APP (1411.4), comp. units: 12, stat: compatible #1: name: GeForce GTX 750 Ti, vendor: NVIDIA Corporation device version: OpenCL 1.1 CUDA, comp. units: 5, stat: compatible #2: name: Intel(R) HD Graphics 4600, vendor: Intel(R) Corporation device version: OpenCL 1.2 , comp. units: 20, stat: compatible 1 GPU user-selected for this run. Mapping of GPU ID to the 1 PP rank in this node: GeForce GTX 750 Ti Environment Variables For OpenCL -------------------------------- Currently, several environment variables exist that help customize some aspects of the OpenCL version of Gromacs. They are mostly related to the runtime compilation of OpenCL kernels, but they are also used on the device selection. OCL_FORCE_CPU - Force the selection of a CPU device instead of a GPU. This exists only for debugging purposes. Do not expect Gromacs to function properly with this option on, it is solely for the simplicity of stepping in a kernel and see what is happening. OCL_FILE_PATH - Is the full path to Gromacs src folder. Useful when gmx is called from a folder other than the installation/bin folder. OCL_NOFASTGEN - Generates and compiles all algorithm flavors, otherwise only the flavor required for the simulation is generated and compiled. OCL_DEBUG - Use in conjunction with OCL_FORCE_CPU or with an AMD device. It adds the debug flag to the compiler options (-g). OCL_FASTMATH - Adds the option cl-fast-relaxed-math to the compiler options (in the CUDA version this is enabled by default, it is likely that the same will happen with the OpenCL version soon) OCL_NOOPT - Disable optimisations. Adds the option cl-opt-disable to the compiler options. 4. KNOWN LIMITATIONS ================= - Intel CPUs are not supported - Intel GPUs are not supported - The current implementation is not compatible with OpenCL devices that are not using warp/wavefronts or for which the warp/wavefront size is not a multiple of 32 - The following kernels are known to produce incorrect results: nbnxn_kernel_ElecEwQSTab_VdwLJ_VF_prune_opencl nbnxn_kernel_ElecEwQSTab_VdwLJ_F_prune_opencl 5. TESTED CONFIGURATIONS ===================== Tested devices: NVIDIA GPUs: GeForce GTX 660M, GeForce GTX 750Ti AMD GPUs: FirePro W5100, HD 7950, FirePro W9100, Radeon R7 M260 Tested kernels: Kernel |Benchmark test |Remarks -------------------------------------------------------------------------------------------------------- nbnxn_kernel_ElecCut_VdwLJ_VF_prune_opencl |d.poly-ch2 | nbnxn_kernel_ElecCut_VdwLJ_F_opencl |d.poly-ch2 | nbnxn_kernel_ElecCut_VdwLJ_F_prune_opencl |d.poly-ch2 | nbnxn_kernel_ElecCut_VdwLJ_VF_opencl |d.poly-ch2 | nbnxn_kernel_ElecRF_VdwLJ_VF_prune_opencl |adh_cubic with rf_verlet.mdp | nbnxn_kernel_ElecRF_VdwLJ_F_opencl |adh_cubic with rf_verlet.mdp | nbnxn_kernel_ElecRF_VdwLJ_F_prune_opencl |adh_cubic with rf_verlet.mdp | nbnxn_kernel_ElecEwQSTab_VdwLJ_VF_prune_opencl |adh_cubic_vsites with pme_verlet_vsites.mdp |Failed nbnxn_kernel_ElecEwQSTab_VdwLJ_F_prune_opencl |adh_cubic_vsites with pme_verlet_vsites.mdp |Failed Input data used for testing - Benchmark data sets available here: ftp://ftp.gromacs.org/pub/benchmarks * * * * * The GROMACS OpenCL team StreamComputing - www.streamcomputing.eu Vincent Hindriksen Teemu Virolainen Anca Hamuraru Contributors Dimitris Karkoulis * * * * * GROMACS is free software, distributed under the GNU Lesser General Public License, version 2.1 However, scientific software is a little special compared to most other programs. Both you, we, and all other GROMACS users depend on the quality of the code, and when we find bugs (every piece of software has them) it is crucial that we can correct it and say that it was fixed in version X of the file or package release. For the same reason, it is important that you can reproduce other people's result from a certain GROMACS version. The easiest way to avoid this kind of problems is to get your modifications included in the main distribution. We'll be happy to consider any decent code. If it's a separate program it can probably be included in the contrib directory straight away (not supported by us), but for major changes in the main code we appreciate if you first test that it works with (and without) MPI, threads, double precision, etc. If you still want to distribute a modified version or use part of GROMACS in your own program, remember that the entire project must be licensed according to the requirements of the LGPL v2.1 license under which you received this copy of GROMACS. We request that it must clearly be labeled as derived work. It should not use the name "official GROMACS", and make sure support questions are directed to you instead of the GROMACS developers. Sorry for the hard wording, but it is meant to protect YOUR reseach results! * * * * *
About
Public/backup repository of the gromacs molecular simulation toolkit
Resources
License
Stars
Watchers
Forks
Packages 0
No packages published
Languages
- C 93.0%
- C++ 5.1%
- TeX 1.1%
- Python 0.5%
- Cuda 0.2%
- Fortran 0.1%