1 .. NOTE: Below is a useful bash one-liner to verify whether there are variables in this file
2 .. no longer present in the code.
3 .. ( export INPUT_FILE='docs/user-guide/environment-variables.rst' GIT_PAGER="cat "; for s in $(grep '^`' $INPUT_FILE | sed 's/`//g' | sed 's/,/ /g'); do count=$(git grep $s | grep -v $INPUT_FILE | wc -l); [ $count -eq 0 ] && printf "%-30s%s\n" $s $count; done ; )
4 .. Another useful one-liner to find undocumentedvariables:
5 .. ( export INPUT_FILE=docs/user-guide/environment-variables.rst; GIT_PAGER="cat "; for ss in `for s in $(git grep getenv | sed 's/.*getenv("\(.*\)".*/\1/' | sort -u | grep '^[A-Z]'); do [ $(grep $s $INPUT_FILE -c) -eq 0 ] && echo $s; done `; do git grep $ss ; done )
10 |Gromacs| programs may be influenced by the use of
11 environment variables. First of all, the variables set in
12 the ``GMXRC`` file are essential for running and
13 compiling |Gromacs|. Some other useful environment variables are
14 listed in the following sections. Most environment variables function
15 by being set in your shell to any non-NULL value. Specific
16 requirements are described below if other values need to be set. You
17 should consult the documentation for your shell for instructions on
18 how to set environment variables in the current shell, or in configuration
19 files for future shells. Note that requirements for exporting
20 environment variables to jobs run under batch control systems vary and
21 you should consult your local documentation for details.
26 Print constraint virial and force virial energy terms.
29 Neighbour list dump level; default 0.
32 |Gromacs| automatically backs up old
33 copies of files when trying to write a new file of the same
34 name, and this variable controls the maximum number of
35 backups that will be made, default 99. If set to 0 it fails to
36 run if any output file already exists. And if set to -1 it
37 overwrites any output file without making a backup.
40 if this is explicitly set, no cool quotes
41 will be printed at the end of a program.
44 prevent dumping of step files during
45 (for example) blowing up during failure of constraint
49 dump all configurations to a :ref:`pdb`
50 file that have an interaction energy less than the value set
51 in this environment variable.
54 ``GMX_VIEW_XVG``, ``GMX_VIEW_EPS`` and ``GMX_VIEW_PDB``, commands used to
55 automatically view :ref:`xvg`, :ref:`xpm`, :ref:`eps`
56 and :ref:`pdb` file types, respectively; they default to ``xv``, ``xmgrace``,
57 ``ghostview`` and ``rasmol``. Set to empty to disable
58 automatic viewing of a particular file type. The command will
59 be forked off and run in the background at the same priority
60 as the |Gromacs| tool (which might not be what you want).
61 Be careful not to use a command which blocks the terminal
62 (e.g. ``vi``), since multiple instances might be run.
65 the size of the buffer for file I/O. When set
66 to 0, all file I/O will be unbuffered and therefore very slow.
67 This can be handy for debugging purposes, because it ensures
68 that all files are always totally up-to-date.
71 set display color for logo in :ref:`gmx view`.
73 ``GMX_PRINT_LONGFORMAT``
74 use long float format when printing
78 Applies for computational electrophysiology setups
79 only (see reference manual). The initial structure gets dumped to
80 :ref:`pdb` file, which allows to check whether multimeric channels have
81 the correct PBC representation.
83 ``GMX_TRAJECTORY_IO_VERBOSITY``
84 Defaults to 1, which prints frame count e.g. when reading trajectory
85 files. Set to 0 for quiet operation.
87 ``GMX_ENABLE_GPU_TIMING``
88 Enables GPU timings in the log file for CUDA. Note that CUDA timings
89 are incorrect with multiple streams, as happens with domain
90 decomposition or with both non-bondeds and PME on the GPU (this is
91 also the main reason why they are not turned on by default).
93 ``GMX_DISABLE_GPU_TIMING``
94 Disables GPU timings in the log file for OpenCL.
98 ``GMX_PRINT_DEBUG_LINES``
99 when set, print debugging info on line numbers.
102 number of steps that elapse between dumping
103 the current DD to a PDB file (default 0). This only takes effect
104 during domain decomposition, so it should typically be
105 0 (never), 1 (every DD phase) or a multiple of :mdp:`nstlist`.
107 ``GMX_DD_NST_DUMP_GRID``
108 number of steps that elapse between dumping
109 the current DD grid to a PDB file (default 0). This only takes effect
110 during domain decomposition, so it should typically be
111 0 (never), 1 (every DD phase) or a multiple of :mdp:`nstlist`.
114 general debugging trigger for every domain
115 decomposition (default 0, meaning off). Currently only checks
116 global-local atom index mapping for consistency.
119 over-ride the number of DD pulses used
120 (default 0, meaning no over-ride). Normally 1 or 2.
122 ``GMX_DISABLE_ALTERNATING_GPU_WAIT``
123 disables the specialized polling wait path used to wait for the PME and nonbonded
124 GPU tasks completion to overlap to do the reduction of the resulting forces that
125 arrive first. Setting this variable switches to the generic path with fixed waiting
128 ``GMX_TEST_REQUIRED_NUMBER_OF_DEVICES``
129 sets the number of GPUs required by the test suite. By default, the test suite would
130 fall-back to using CPU if GPUs could not be detected. Set it to a positive integer value
131 to ensure that at least this at least this number of usable GPUs are detected. Default:
132 0 (not testing GPU availability).
134 There are a number of extra environment variables like these
135 that are used in debugging - check the code!
137 Performance and Run Control
138 ---------------------------
139 ``GMX_DO_GALACTIC_DYNAMICS``
140 planetary simulations are made possible (just for fun) by setting
141 this environment variable, which allows setting :mdp:`epsilon-r` to -1 in the :ref:`mdp`
142 file. Normally, :mdp:`epsilon-r` must be greater than zero to prevent a fatal error.
143 See webpage_ for example input files for a planetary simulation.
145 ``GMX_BONDED_NTHREAD_UNIFORM``
146 Value of the number of threads per rank from which to switch from uniform
147 to localized bonded interaction distribution; optimal value dependent on
148 system and hardware, default value is 4.
150 ``GMX_GPU_NB_EWALD_TWINCUT``
151 force the use of twin-range cutoff kernel even if :mdp:`rvdw` equals
152 :mdp:`rcoulomb` after PP-PME load balancing. The switch to twin-range kernels is automated,
153 so this variable should be used only for benchmarking.
155 ``GMX_GPU_NB_ANA_EWALD``
156 force the use of analytical Ewald kernels. Should be used only for benchmarking.
158 ``GMX_GPU_NB_TAB_EWALD``
159 force the use of tabulated Ewald kernels. Should be used only for benchmarking.
161 ``GMX_DISABLE_CUDA_TIMING``
162 Deprecated. Use ``GMX_DISABLE_GPU_TIMING`` instead.
165 perform domain decomposition halo exchange communication operations (on coordinate and force buffers)
166 directly on GPU memory spaces, without the staging of data through CPU memory, where possible.
168 ``GMX_GPU_PME_PP_COMMS``
169 when the simulation uses a separate PME rank, perform communication operations between PP and PME rank
170 (for coordinate and force buffers) directly on GPU memory spaces, without the staging of data through CPU
171 memory, where possible.
173 ``GMX_GPU_SYCL_NO_SYNCHRONIZE``
174 disable synchronizations between different GPU streams in SYCL build, instead relying on SYCL runtime to
175 do scheduling based on data dependencies. Experimental.
178 times all code during runs. Incompatible with threads.
180 ``GMX_CYCLE_BARRIER``
181 calls MPI_Barrier before each cycle start/stop call.
184 build domain decomposition cells in the order
185 (z, y, x) rather than the default (x, y, z).
187 ``GMX_DD_USE_SENDRECV2``
188 during constraint and vsite communication, use a pair
189 of ``MPI_Sendrecv`` calls instead of two simultaneous non-blocking calls
190 (default 0, meaning off). Might be faster on some MPI implementations.
192 ``GMX_DLB_BASED_ON_FLOPS``
193 do domain-decomposition dynamic load balancing based on flop count rather than
194 measured time elapsed (default 0, meaning off).
195 This makes the load balancing reproducible, which can be useful for debugging purposes.
196 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.
198 ``GMX_DLB_MAX_BOX_SCALING``
199 maximum percentage box scaling permitted per domain-decomposition
200 load-balancing step (default 10)
202 ``GMX_DD_RECORD_LOAD``
203 record DD load statistics for reporting at end of the run (default 1, meaning on)
205 ``GMX_DETAILED_PERF_STATS``
206 when set, print slightly more detailed performance information
207 to the :ref:`log` file. The resulting output is the way performance summary is reported in versions
208 4.5.x and thus may be useful for anyone using scripts to parse :ref:`log` files or standard output.
210 ``GMX_DISABLE_SIMD_KERNELS``
211 disables architecture-specific SIMD-optimized (SSE2, SSE4.1, AVX, etc.)
212 non-bonded kernels thus forcing the use of plain C kernels.
214 ``GMX_DISABLE_GPU_TIMING``
215 timing of asynchronously executed GPU operations can have a
216 non-negligible overhead with short step times. Disabling timing can improve performance in these cases.
218 ``GMX_DISABLE_GPU_DETECTION``
219 when set, disables GPU detection even if :ref:`gmx mdrun` was compiled
222 ``GMX_GPU_APPLICATION_CLOCKS``
223 setting this variable to a value of "0", "ON", or "DISABLE" (case insensitive)
224 allows disabling the CUDA GPU allication clock support.
226 ``GMX_DISRE_ENSEMBLE_SIZE``
227 the number of systems for distance restraint ensemble
228 averaging. Takes an integer value.
231 emulate GPU runs by using algorithmically equivalent CPU reference code instead of
232 GPU-accelerated functions. As the CPU code is slow, it is intended to be used only for debugging purposes.
235 disable exiting upon encountering a corrupted frame in an :ref:`edr`
236 file, allowing the use of all frames up until the corruption.
239 update forces when invoking ``mdrun -rerun``.
241 ``GMX_FORCE_UPDATE_DEFAULT_GPU``
242 Force update to run on the GPU by default, overriding the ``mdrun -update auto`` option. Works similar to setting
243 ``mdrun -update gpu``, but (1) falls back to the CPU code-path, if set with input that is not supported and
244 (2) can be used to run update on GPUs in multi-rank cases. The latter case should be
245 considered experimental since it lacks substantial testing. Also, GPU update is only supported with the GPU direct
246 communications and ``GMX_FORCE_UPDATE_DEFAULT_GPU`` variable should be set simultaneously with ``GMX_GPU_DD_COMMS``
247 and ``GMX_GPU_PME_PP_COMMS`` environment variables in multi-rank case. Does not override ``mdrun -update cpu``.
250 set in the same way as ``mdrun -gpu_id``, ``GMX_GPU_ID``
251 allows the user to specify different GPU IDs for different ranks, which can be useful for selecting different
252 devices on different compute nodes in a cluster. Cannot be used in conjunction with ``mdrun -gpu_id``.
255 set in the same way as ``mdrun -gputasks``, ``GMX_GPUTASKS`` allows the mapping
256 of GPU tasks to GPU device IDs to be different on different ranks, if e.g. the MPI
257 runtime permits this variable to be different for different ranks. Cannot be used
258 in conjunction with ``mdrun -gputasks``. Has all the same requirements as ``mdrun -gputasks``.
260 ``GMX_GPU_DISABLE_COMPATIBILITY_CHECK``
261 Disables the hardware compatibility check in OpenCL and SYCL. Useful for developers
262 and allows testing the OpenCL/SYCL kernels on non-supported platforms without source code modification.
264 ``GMX_IGNORE_FSYNC_FAILURE_ENV``
265 allow :ref:`gmx mdrun` to continue even if
269 when set to a floating-point value, overrides the default tolerance of
270 1e-5 for force-field floating-point parameters.
272 ``GMX_MAXCONSTRWARN``
273 if set to -1, :ref:`gmx mdrun` will
274 not exit if it produces too many LINCS warnings.
277 neighbor list balancing parameter used when running on GPU. Sets the
278 target minimum number pair-lists in order to improve multi-processor load-balance for better
279 performance with small simulation systems. Must be set to a non-negative integer,
280 the 0 value disables list splitting.
281 The default value is optimized for supported GPUs
282 therefore changing it is not necessary for normal usage, but it can be useful on future architectures.
285 use neighbor list and kernels based on charge groups.
288 when set, print detailed neighbor search cycle counting.
290 ``GMX_NBNXN_EWALD_ANALYTICAL``
291 force the use of analytical Ewald non-bonded kernels,
292 mutually exclusive of ``GMX_NBNXN_EWALD_TABLE``.
294 ``GMX_NBNXN_EWALD_TABLE``
295 force the use of tabulated Ewald non-bonded kernels,
296 mutually exclusive of ``GMX_NBNXN_EWALD_ANALYTICAL``.
298 ``GMX_NBNXN_SIMD_2XNN``
299 force the use of 2x(N+N) SIMD CPU non-bonded kernels,
300 mutually exclusive of ``GMX_NBNXN_SIMD_4XN``.
302 ``GMX_NBNXN_SIMD_4XN``
303 force the use of 4xN SIMD CPU non-bonded kernels,
304 mutually exclusive of ``GMX_NBNXN_SIMD_2XNN``.
306 ``GMX_NOOPTIMIZEDKERNELS``
307 deprecated, use ``GMX_DISABLE_SIMD_KERNELS`` instead.
309 ``GMX_NO_CART_REORDER``
310 used in initializing domain decomposition communicators. Rank reordering
311 is default, but can be switched off with this environment variable.
313 ``GMX_NO_LJ_COMB_RULE``
314 force the use of LJ paremeter lookup instead of using combination rules
315 in the non-bonded kernels.
317 ``GMX_NO_INT``, ``GMX_NO_TERM``, ``GMX_NO_USR1``
318 disable signal handlers for SIGINT,
319 SIGTERM, and SIGUSR1, respectively.
322 do not use separate inter- and intra-node communicators.
325 skip non-bonded calculations; can be used to estimate the possible
326 performance gain from adding a GPU accelerator to the current hardware setup -- assuming that this is
327 fast enough to complete the non-bonded calculations while the CPU does bonded force and PME computation.
328 Freezing the particles will be required to stop the system blowing up.
330 ``GMX_PULL_PARTICIPATE_ALL``
331 disable the default heuristic for when to use a separate pull MPI communicator (at >=32 ranks).
334 shell positions are not predicted.
336 ``GMX_NO_UPDATEGROUPS``
337 turns off update groups. May allow for a decomposition of more
338 domains for small systems at the cost of communication during update.
341 the ideal number of charge groups per neighbor searching grid cell is hard-coded
342 to a value of 10. Setting this environment variable to any other integer value overrides this hard-coded
345 ``GMX_PME_NUM_THREADS``
346 set the number of OpenMP or PME threads; overrides the default set by
347 :ref:`gmx mdrun`; can be used instead of the ``-npme`` command line option,
348 also useful to set heterogeneous per-process/-node thread count.
351 use P3M-optimized influence function instead of smooth PME B-spline interpolation.
353 ``GMX_PME_THREAD_DIVISION``
354 PME thread division in the format "x y z" for all three dimensions. The
355 sum of the threads in each dimension must equal the total number of PME threads (set in
356 :envvar:`GMX_PME_NTHREADS`).
359 if the number of domain decomposition cells is set to 1 for both x and y,
360 decompose PME in one dimension.
362 ``GMX_REQUIRE_SHELL_INIT``
363 require that shell positions are initiated.
365 ``GMX_REQUIRE_TABLES``
366 require the use of tabulated Coulombic
367 and van der Waals interactions.
370 should contain multiple masses used for test particle insertion into a cavity.
371 The center of mass of the last atoms is used for insertion into the cavity.
374 use graph for bonded interactions.
376 ``GMX_VERLET_BUFFER_RES``
377 resolution of buffer size in Verlet cutoff scheme. The default value is
378 0.001, but can be overridden with this environment variable.
381 Not strictly a |Gromacs| environment variable, but on large machines
382 the hwloc detection can take a few seconds if you have lots of MPI processes.
383 If you run the hwloc command :command:`lstopo out.xml` and set this environment
384 variable to point to the location of this file, the hwloc library will use
385 the cached information instead, which can be faster.
388 the ``mpirun`` command used by :ref:`gmx tune_pme`.
391 the :ref:`gmx mdrun` command used by :ref:`gmx tune_pme`.
393 ``GMX_DISABLE_DYNAMICPRUNING``
394 disables dynamic pair-list pruning. Note that :ref:`gmx mdrun` will
395 still tune nstlist to the optimal value picked assuming dynamic pruning. Thus
396 for good performance the -nstlist option should be used.
398 ``GMX_NSTLIST_DYNAMICPRUNING``
399 overrides the dynamic pair-list pruning interval chosen heuristically
400 by mdrun. Values should be between the pruning frequency value
401 (1 for CPU and 2 for GPU) and :mdp:`nstlist` ``- 1``.
403 .. _opencl-management:
407 Currently, several environment variables exist that help customize some aspects
408 of the OpenCL_ version of |Gromacs|. They are mostly related to the runtime
409 compilation of OpenCL kernels, but they are also used in device selection.
411 ``GMX_OCL_NOGENCACHE``
412 If set, disable caching for OpenCL kernel builds. Caching is
413 normally useful so that future runs can re-use the compiled
414 kernels from previous runs. Currently, caching is always
415 disabled, until we solve concurrency issues.
418 Enable OpenCL binary caching. Only intended to be used for
419 development and (expert) testing as neither concurrency
420 nor cache invalidation is implemented safely!
422 ``GMX_OCL_NOFASTGEN``
423 If set, generate and compile all algorithm flavors, otherwise
424 only the flavor required for the simulation is generated and
427 ``GMX_OCL_DISABLE_FASTMATH``
428 Prevents the use of ``-cl-fast-relaxed-math`` compiler option.
431 If defined, the OpenCL build log is always written to the
432 mdrun log file. Otherwise, the build log is written to the
433 log file only when an error occurs.
436 If defined, it enables verbose mode for OpenCL kernel build.
437 Currently available only for NVIDIA GPUs. See ``GMX_OCL_DUMP_LOG``
438 for details about how to obtain the OpenCL build log.
440 ``GMX_OCL_DUMP_INTERM_FILES``
442 If defined, intermediate language code corresponding to the
443 OpenCL build process is saved to file. Caching has to be
444 turned off in order for this option to take effect (see
445 ``GMX_OCL_NOGENCACHE``).
447 - NVIDIA GPUs: PTX code is saved in the current directory
448 with the name ``device_name.ptx``
449 - AMD GPUs: ``.IL/.ISA`` files will be created for each OpenCL
450 kernel built. For details about where these files are
451 created check AMD documentation for ``-save-temps`` compiler
455 Use in conjunction with ``OCL_FORCE_CPU`` or with an AMD device.
456 It adds the debug flag to the compiler options (-g).
459 Disable optimisations. Adds the option ``cl-opt-disable`` to the
462 ``GMX_OCL_FORCE_CPU``
463 Force the selection of a CPU device instead of a GPU. This
464 exists only for debugging purposes. Do not expect |Gromacs| to
465 function properly with this option on, it is solely for the
466 simplicity of stepping in a kernel and see what is happening.
468 ``GMX_OCL_DISABLE_I_PREFETCH``
469 Disables i-atom data (type or LJ parameter) prefetch allowing
472 ``GMX_OCL_ENABLE_I_PREFETCH``
473 Enables i-atom data (type or LJ parameter) prefetch allowing
474 testing on platforms where this behavior is not default.
476 ``GMX_OCL_FILE_PATH``
477 Use this parameter to force |Gromacs| to load the OpenCL
478 kernels from a custom location. Use it only if you want to
479 override |Gromacs| default behavior, or if you want to test
482 ``GMX_OCL_SHOW_DIAGNOSTICS``
483 Use Intel OpenCL extension to show additional runtime performance
486 Analysis and Core Functions
487 ---------------------------
490 used by :ref:`gmx do_dssp` to point to the ``dssp``
491 executable (not just its path).
493 ``GMX_DIPOLE_SPACING``
494 spacing used by :ref:`gmx dipoles`.
497 sets the maximum number of residues to be renumbered by
498 :ref:`gmx grompp`. A value of -1 indicates all residues should be renumbered.
500 ``GMX_NO_FFRTP_TER_RENAME``
501 Some force fields (like AMBER) use specific names for N- and C-
502 terminal residues (NXXX and CXXX) as :ref:`rtp` entries that are normally renamed. Setting
503 this environment variable disables this renaming.
506 name of X11 font used by :ref:`gmx view`.
509 the time unit used in output files, can be
510 anything in fs, ps, ns, us, ms, s, m or h.
513 name of the ``multiprot`` executable, used by the
514 contributed program ``do_multiprot``.
517 number of CPUs to be used for Gaussian QM calculation
520 name of the ``total`` executable used by the contributed
521 ``do_shift`` program.
524 make :ref:`gmx energy` and :ref:`gmx eneconv`
528 where to find VMD plug-ins. Needed to be
529 able to read file formats recognized only by a VMD plug-in.
532 base path of VMD installation.
535 sets viewer to ``xmgr`` (deprecated) instead of ``xmgrace``.