4 |Gromacs| programs may be influenced by the use of
5 environment variables. First of all, the variables set in
6 the ``GMXRC`` file are essential for running and
7 compiling |Gromacs|. Some other useful environment variables are
8 listed in the following sections. Most environment variables function
9 by being set in your shell to any non-NULL value. Specific
10 requirements are described below if other values need to be set. You
11 should consult the documentation for your shell for instructions on
12 how to set environment variables in the current shell, or in configuration
13 files for future shells. Note that requirements for exporting
14 environment variables to jobs run under batch control systems vary and
15 you should consult your local documentation for details.
20 Print constraint virial and force virial energy terms.
23 |Gromacs| automatically backs up old
24 copies of files when trying to write a new file of the same
25 name, and this variable controls the maximum number of
26 backups that will be made, default 99. If set to 0 it fails to
27 run if any output file already exists. And if set to -1 it
28 overwrites any output file without making a backup.
31 if this is explicitly set, no cool quotes
32 will be printed at the end of a program.
35 prevent dumping of step files during
36 (for example) blowing up during failure of constraint
40 dump all configurations to a :ref:`pdb`
41 file that have an interaction energy less than the value set
42 in this environment variable.
45 ``GMX_VIEW_XVG``, ``GMX_VIEW_EPS`` and ``GMX_VIEW_PDB``, commands used to
46 automatically view :ref:`xvg`, :ref:`xpm`, :ref:`eps`
47 and :ref:`pdb` file types, respectively; they default to ``xv``, ``xmgrace``,
48 ``ghostview`` and ``rasmol``. Set to empty to disable
49 automatic viewing of a particular file type. The command will
50 be forked off and run in the background at the same priority
51 as the |Gromacs| tool (which might not be what you want).
52 Be careful not to use a command which blocks the terminal
53 (e.g. ``vi``), since multiple instances might be run.
55 ``GMX_VIRIAL_TEMPERATURE``
56 print virial temperature energy term
59 the size of the buffer for file I/O. When set
60 to 0, all file I/O will be unbuffered and therefore very slow.
61 This can be handy for debugging purposes, because it ensures
62 that all files are always totally up-to-date.
65 set display color for logo in :ref:`gmx view`.
67 ``GMX_PRINT_LONGFORMAT``
68 use long float format when printing
72 Applies for computational electrophysiology setups
73 only (see reference manual). The initial structure gets dumped to
74 :ref:`pdb` file, which allows to check whether multimeric channels have
75 the correct PBC representation.
79 ``GMX_PRINT_DEBUG_LINES``
80 when set, print debugging info on line numbers.
83 number of steps that elapse between dumping
84 the current DD to a PDB file (default 0). This only takes effect
85 during domain decomposition, so it should typically be
86 0 (never), 1 (every DD phase) or a multiple of :mdp:`nstlist`.
88 ``GMX_DD_NST_DUMP_GRID``
89 number of steps that elapse between dumping
90 the current DD grid to a PDB file (default 0). This only takes effect
91 during domain decomposition, so it should typically be
92 0 (never), 1 (every DD phase) or a multiple of :mdp:`nstlist`.
95 general debugging trigger for every domain
96 decomposition (default 0, meaning off). Currently only checks
97 global-local atom index mapping for consistency.
100 over-ride the number of DD pulses used
101 (default 0, meaning no over-ride). Normally 1 or 2.
103 There are a number of extra environment variables like these
104 that are used in debugging - check the code!
106 Performance and Run Control
107 ---------------------------
108 ``GMX_DO_GALACTIC_DYNAMICS``
109 planetary simulations are made possible (just for fun) by setting
110 this environment variable, which allows setting :mdp:`epsilon-r` to -1 in the :ref:`mdp`
111 file. Normally, :mdp:`epsilon-r` must be greater than zero to prevent a fatal error.
112 See webpage_ for example input files for a planetary simulation.
114 ``GMX_ALLOW_CPT_MISMATCH``
115 when set, runs will not exit if the
116 ensemble set in the :ref:`tpr` file does not match that of the
119 ``GMX_CUDA_NB_EWALD_TWINCUT``
120 force the use of twin-range cutoff kernel even if :mdp:`rvdw` equals
121 :mdp:`rcoulomb` after PP-PME load balancing. The switch to twin-range kernels is automated,
122 so this variable should be used only for benchmarking.
124 ``GMX_CUDA_NB_ANA_EWALD``
125 force the use of analytical Ewald kernels. Should be used only for benchmarking.
127 ``GMX_CUDA_NB_TAB_EWALD``
128 force the use of tabulated Ewald kernels. Should be used only for benchmarking.
130 ``GMX_CUDA_STREAMSYNC``
131 force the use of cudaStreamSynchronize on ECC-enabled GPUs, which leads
132 to performance loss due to a known CUDA driver bug present in API v5.0 NVIDIA drivers (pre-30x.xx).
133 Cannot be set simultaneously with ``GMX_NO_CUDA_STREAMSYNC``.
135 ``GMX_DISABLE_CUDALAUNCH``
136 disable the use of the lower-latency cudaLaunchKernel API even when supported (CUDA >=v7.0).
137 Should only be used for benchmarking purposes.
140 times all code during runs. Incompatible with threads.
142 ``GMX_CYCLE_BARRIER``
143 calls MPI_Barrier before each cycle start/stop call.
146 build domain decomposition cells in the order
147 (z, y, x) rather than the default (x, y, z).
149 ``GMX_DD_USE_SENDRECV2``
150 during constraint and vsite communication, use a pair
151 of ``MPI_Sendrecv`` calls instead of two simultaneous non-blocking calls
152 (default 0, meaning off). Might be faster on some MPI implementations.
154 ``GMX_DLB_BASED_ON_FLOPS``
155 do domain-decomposition dynamic load balancing based on flop count rather than
156 measured time elapsed (default 0, meaning off).
157 This makes the load balancing reproducible, which can be useful for debugging purposes.
158 A value of 1 uses the flops; a value > 1 adds (value - 1)*5% of noise to the flops to increase the imbalance and the scaling.
160 ``GMX_DLB_MAX_BOX_SCALING``
161 maximum percentage box scaling permitted per domain-decomposition
162 load-balancing step (default 10)
164 ``GMX_DD_RECORD_LOAD``
165 record DD load statistics for reporting at end of the run (default 1, meaning on)
167 ``GMX_DETAILED_PERF_STATS``
168 when set, print slightly more detailed performance information
169 to the :ref:`log` file. The resulting output is the way performance summary is reported in versions
170 4.5.x and thus may be useful for anyone using scripts to parse :ref:`log` files or standard output.
172 ``GMX_DISABLE_SIMD_KERNELS``
173 disables architecture-specific SIMD-optimized (SSE2, SSE4.1, AVX, etc.)
174 non-bonded kernels thus forcing the use of plain C kernels.
176 ``GMX_DISABLE_CUDA_TIMING``
177 timing of asynchronously executed GPU operations can have a
178 non-negligible overhead with short step times. Disabling timing can improve performance in these cases.
180 ``GMX_DISABLE_GPU_DETECTION``
181 when set, disables GPU detection even if :ref:`gmx mdrun` was compiled
184 ``GMX_GPU_APPLICATION_CLOCKS``
185 setting this variable to a value of "0", "ON", or "DISABLE" (case insensitive)
186 allows disabling the CUDA GPU allication clock support.
188 ``GMX_DISRE_ENSEMBLE_SIZE``
189 the number of systems for distance restraint ensemble
190 averaging. Takes an integer value.
193 emulate GPU runs by using algorithmically equivalent CPU reference code instead of
194 GPU-accelerated functions. As the CPU code is slow, it is intended to be used only for debugging purposes.
195 The behavior is automatically triggered if non-bonded calculations are turned off using ``GMX_NO_NONBONDED``
196 case in which the non-bonded calculations will not be called, but the CPU-GPU transfer will also be skipped.
199 disable exiting upon encountering a corrupted frame in an :ref:`edr`
200 file, allowing the use of all frames up until the corruption.
203 update forces when invoking ``mdrun -rerun``.
206 set in the same way as ``mdrun -gpu_id``, ``GMX_GPU_ID``
207 allows the user to specify different GPU id-s, which can be useful for selecting different
208 devices on different compute nodes in a cluster. Cannot be used in conjunction with ``mdrun -gpu_id``.
210 ``GMX_IGNORE_FSYNC_FAILURE_ENV``
211 allow :ref:`gmx mdrun` to continue even if
215 when set to a floating-point value, overrides the default tolerance of
216 1e-5 for force-field floating-point parameters.
218 ``GMX_MAXCONSTRWARN``
219 if set to -1, :ref:`gmx mdrun` will
220 not exit if it produces too many LINCS warnings.
223 use the generic C kernel. Should be set if using
224 the group-based cutoff scheme and also sets ``GMX_NO_SOLV_OPT`` to be true,
225 thus disabling solvent optimizations as well.
228 neighbor list balancing parameter used when running on GPU. Sets the
229 target minimum number pair-lists in order to improve multi-processor load-balance for better
230 performance with small simulation systems. Must be set to a non-negative integer,
231 the 0 value disables list splitting.
232 The default value is optimized for supported GPUs (NVIDIA Fermi to Maxwell),
233 therefore changing it is not necessary for normal usage, but it can be useful on future architectures.
236 use neighbor list and kernels based on charge groups.
239 when set, print detailed neighbor search cycle counting.
241 ``GMX_NBNXN_EWALD_ANALYTICAL``
242 force the use of analytical Ewald non-bonded kernels,
243 mutually exclusive of ``GMX_NBNXN_EWALD_TABLE``.
245 ``GMX_NBNXN_EWALD_TABLE``
246 force the use of tabulated Ewald non-bonded kernels,
247 mutually exclusive of ``GMX_NBNXN_EWALD_ANALYTICAL``.
249 ``GMX_NBNXN_SIMD_2XNN``
250 force the use of 2x(N+N) SIMD CPU non-bonded kernels,
251 mutually exclusive of ``GMX_NBNXN_SIMD_4XN``.
253 ``GMX_NBNXN_SIMD_4XN``
254 force the use of 4xN SIMD CPU non-bonded kernels,
255 mutually exclusive of ``GMX_NBNXN_SIMD_2XNN``.
258 disables optimized all-vs-all kernels.
260 ``GMX_NO_CART_REORDER``
261 used in initializing domain decomposition communicators. Rank reordering
262 is default, but can be switched off with this environment variable.
264 ``GMX_NO_LJ_COMB_RULE``
265 force the use of LJ paremeter lookup instead of using combination rules
266 in the non-bonded kernels.
268 ``GMX_NO_CUDA_STREAMSYNC``
269 the opposite of ``GMX_CUDA_STREAMSYNC``. Disables the use of the
270 standard cudaStreamSynchronize-based GPU waiting to improve performance when using CUDA driver API
271 ealier than v5.0 with ECC-enabled GPUs.
273 ``GMX_NO_INT``, ``GMX_NO_TERM``, ``GMX_NO_USR1``
274 disable signal handlers for SIGINT,
275 SIGTERM, and SIGUSR1, respectively.
278 do not use separate inter- and intra-node communicators.
281 skip non-bonded calculations; can be used to estimate the possible
282 performance gain from adding a GPU accelerator to the current hardware setup -- assuming that this is
283 fast enough to complete the non-bonded calculations while the CPU does bonded force and PME computation.
286 when set, do not add virial contribution to COM pull forces.
289 shell positions are not predicted.
292 turns off solvent optimizations; automatic if ``GMX_NB_GENERIC``
296 the ideal number of charge groups per neighbor searching grid cell is hard-coded
297 to a value of 10. Setting this environment variable to any other integer value overrides this hard-coded
301 set the number of OpenMP or PME threads (overrides the number guessed by
305 use P3M-optimized influence function instead of smooth PME B-spline interpolation.
307 ``GMX_PME_THREAD_DIVISION``
308 PME thread division in the format "x y z" for all three dimensions. The
309 sum of the threads in each dimension must equal the total number of PME threads (set in
313 if the number of domain decomposition cells is set to 1 for both x and y,
314 decompose PME in one dimension.
316 ``GMX_REQUIRE_SHELL_INIT``
317 require that shell positions are initiated.
319 ``GMX_REQUIRE_TABLES``
320 require the use of tabulated Coulombic
321 and van der Waals interactions.
324 the minimum value for soft-core sigma. **Note** that this value is set
325 using the :mdp:`sc-sigma` keyword in the :ref:`mdp` file, but this environment variable can be used
326 to reproduce pre-4.5 behavior with respect to this parameter.
329 should contain multiple masses used for test particle insertion into a cavity.
330 The center of mass of the last atoms is used for insertion into the cavity.
333 use graph for bonded interactions.
335 ``GMX_VERLET_BUFFER_RES``
336 resolution of buffer size in Verlet cutoff scheme. The default value is
337 0.001, but can be overridden with this environment variable.
340 Not strictly a |Gromacs| environment variable, but on large machines
341 the hwloc detection can take a few seconds if you have lots of MPI processes.
342 If you run the hwloc command `lstopo out.xml` and set this environment
343 variable to point to the location of this file, the hwloc library will use
344 the cached information instead, which can be faster.
347 the ``mpirun`` command used by :ref:`gmx tune_pme`.
350 the :ref:`gmx mdrun` command used by :ref:`gmx tune_pme`.
353 sets the default value for :mdp:`nstlist`, preventing it from being tuned during
354 :ref:`gmx mdrun` startup when using the Verlet cutoff scheme.
356 ``GMX_USE_TREEREDUCE``
357 use tree reduction for nbnxn force reduction. Potentially faster for large number of
358 OpenMP threads (if memory locality is important).
360 .. _opencl-management:
364 Currently, several environment variables exist that help customize some aspects
365 of the OpenCL_ version of |Gromacs|. They are mostly related to the runtime
366 compilation of OpenCL kernels, but they are also used in device selection.
368 ``GMX_OCL_NOGENCACHE``
369 If set, disable caching for OpenCL kernel builds. Caching is
370 normally useful so that future runs can re-use the compiled
371 kernels from previous runs. Currently, caching is always
372 disabled, until we solve concurrency issues.
375 Enable OpenCL binary caching. Only intended to be used for
376 development and (expert) testing as neither concurrency
377 nor cache invalidation is implemented safely!
379 ``GMX_OCL_NOFASTGEN``
380 If set, generate and compile all algorithm flavors, otherwise
381 only the flavor required for the simulation is generated and
385 Adds the option ``cl-fast-relaxed-math`` to the compiler
386 options (in the CUDA version this is enabled by default, it is likely that
387 the same will happen with the OpenCL version soon)
390 If defined, the OpenCL build log is always written to the
391 mdrun log file. Otherwise, the build log is written to the
392 log file only when an error occurs.
395 If defined, it enables verbose mode for OpenCL kernel build.
396 Currently available only for NVIDIA GPUs. See ``GMX_OCL_DUMP_LOG``
397 for details about how to obtain the OpenCL build log.
399 ``GMX_OCL_DUMP_INTERM_FILES``
401 If defined, intermediate language code corresponding to the
402 OpenCL build process is saved to file. Caching has to be
403 turned off in order for this option to take effect (see
404 ``GMX_OCL_NOGENCACHE``).
406 - NVIDIA GPUs: PTX code is saved in the current directory
407 with the name ``device_name.ptx``
408 - AMD GPUs: ``.IL/.ISA`` files will be created for each OpenCL
409 kernel built. For details about where these files are
410 created check AMD documentation for ``-save-temps`` compiler
414 Use in conjunction with ``OCL_FORCE_CPU`` or with an AMD device.
415 It adds the debug flag to the compiler options (-g).
418 Disable optimisations. Adds the option ``cl-opt-disable`` to the
421 ``GMX_OCL_FORCE_CPU``
422 Force the selection of a CPU device instead of a GPU. This
423 exists only for debugging purposes. Do not expect |Gromacs| to
424 function properly with this option on, it is solely for the
425 simplicity of stepping in a kernel and see what is happening.
427 ``GMX_OCL_DISABLE_I_PREFETCH``
428 Disables i-atom data (type or LJ parameter) prefetch allowig
431 ``GMX_OCL_ENABLE_I_PREFETCH``
432 Enables i-atom data (type or LJ parameter) prefetch allowig
433 testing on platforms where this behavior is not default.
435 ``GMX_OCL_NB_ANA_EWALD``
436 Forces the use of analytical Ewald kernels. Equivalent of
437 CUDA environment variable ``GMX_CUDA_NB_ANA_EWALD``
439 ``GMX_OCL_NB_TAB_EWALD``
440 Forces the use of tabulated Ewald kernel. Equivalent
441 of CUDA environment variable ``GMX_OCL_NB_TAB_EWALD``
443 ``GMX_OCL_NB_EWALD_TWINCUT``
444 Forces the use of twin-range cutoff kernel. Equivalent of
445 CUDA environment variable ``GMX_CUDA_NB_EWALD_TWINCUT``
447 ``GMX_DISABLE_OCL_TIMING``
448 Disables timing for OpenCL operations
450 ``GMX_OCL_FILE_PATH``
451 Use this parameter to force |Gromacs| to load the OpenCL
452 kernels from a custom location. Use it only if you want to
453 override |Gromacs| default behavior, or if you want to test
456 Analysis and Core Functions
457 ---------------------------
459 accuracy in Gaussian L510 (MC-SCF) component program.
461 ``GMX_QM_ORCA_BASENAME``
462 prefix of :ref:`tpr` files, used in Orca calculations
463 for input and output file names.
466 when set to a nonzero value, Gaussian QM calculations will
467 iteratively solve the CP-MCSCF equations.
469 ``GMX_QM_MODIFIED_LINKS_DIR``
470 location of modified links in Gaussian.
473 used by :ref:`gmx do_dssp` to point to the ``dssp``
474 executable (not just its path).
477 directory where Gaussian is installed.
480 name of the Gaussian executable.
482 ``GMX_DIPOLE_SPACING``
483 spacing used by :ref:`gmx dipoles`.
486 sets the maximum number of residues to be renumbered by
487 :ref:`gmx grompp`. A value of -1 indicates all residues should be renumbered.
489 ``GMX_FFRTP_TER_RENAME``
490 Some force fields (like AMBER) use specific names for N- and C-
491 terminal residues (NXXX and CXXX) as :ref:`rtp` entries that are normally renamed. Setting
492 this environment variable disables this renaming.
495 ``gunzip`` executable, used by :ref:`gmx wham`.
498 name of X11 font used by :ref:`gmx view`.
501 the time unit used in output files, can be
502 anything in fs, ps, ns, us, ms, s, m or h.
504 ``GMX_QM_GAUSSIAN_MEMORY``
505 memory used for Gaussian QM calculation.
508 name of the ``multiprot`` executable, used by the
509 contributed program ``do_multiprot``.
512 number of CPUs to be used for Gaussian QM calculation
515 directory where Orca is installed.
518 simulated annealing step size for Gaussian QM calculation.
520 ``GMX_QM_GROUND_STATE``
521 defines state for Gaussian surface hopping calculation.
524 name of the ``total`` executable used by the contributed
525 ``do_shift`` program.
528 make :ref:`gmx energy` and :ref:`gmx eneconv`
532 where to find VMD plug-ins. Needed to be
533 able to read file formats recognized only by a VMD plug-in.
536 base path of VMD installation.
539 sets viewer to ``xmgr`` (deprecated) instead of ``xmgrace``.