HPC/Applications/namd - Revision history

Stern: Stern moved page HPC/Modules/namd to HPC/Applications/namd

2013-06-25T15:21:29Z

Stern moved page HPC/Modules/namd to HPC/Applications/namd

← Older revision	Revision as of 15:21, June 25, 2013
(No difference)

Stern: /* Using GPUs and MPI */

2012-11-26T19:30:18Z

Using GPUs and MPI

← Older revision		Revision as of 19:30, November 26, 2012
Line 39:		Line 39:
	mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \		mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \
	namd2'''-cuda +idlepoll''' ''file.namd''		namd2'''-cuda +idlepoll''' ''file.namd''
	* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>, increase up to <code>ppn='''12'''</code>, and/or increase <code>nodes > 1</code>		* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>, increase up to <code>ppn='''12'''</code>, and/or increase <code>nodes > 1</code>. Speedups from a GPU can be expected to be of order 4–5 ×.
	* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.		* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.

Stern: /* Using GPUs and MPI */

2012-11-26T19:23:17Z

Using GPUs and MPI

← Older revision		Revision as of 19:23, November 26, 2012
Line 39:		Line 39:
	mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \		mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \
	namd2'''-cuda +idlepoll''' ''file.namd''		namd2'''-cuda +idlepoll''' ''file.namd''
	* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>.		* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>, increase up to <code>ppn='''12'''</code>, and/or increase <code>nodes > 1</code>
	* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.		* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.

Stern: /* Using GPUs and MPI */

2012-11-09T17:59:19Z

Using GPUs and MPI

← Older revision		Revision as of 17:59, November 9, 2012
Line 19:		Line 19:
	module load fftw3		module load fftw3
	module load cuda		module load cuda
	module load namd~~/2.9plus-MPI-icc-3~~		module load namd
	<!--		<!--
	; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].		; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].

Stern at 21:42, November 8, 2012

2012-11-08T21:42:15Z

← Older revision		Revision as of 21:42, November 8, 2012
Line 1:		Line 1:
	~~== Introduction ==~~
	~~[http://www.ks.uiuc.edu/Research/namd/ NAMD] is a parallel, object-oriented molecular dynamics code designed~~
	~~for high-performance simulation of large biomolecular systems.~~
	~~NAMD is distributed free of charge and includes source code.~~

	~~Subject to [http://www.ks.uiuc.edu/Research/namd/license.html Univ. of Illinois Non-Exclusive, Non-Commercial Use License].~~

	~~The NAMD project is funded by the National Institutes of Health (grant number PHS 5 P41 RR05969).~~

	== MPI version ==		== MPI version ==
	=== Modules ===		=== Modules ===

Stern at 21:39, November 8, 2012

2012-11-08T21:39:36Z

← Older revision		Revision as of 21:39, November 8, 2012
Line 1:		Line 1:
	== Introduction ==		== Introduction ==

	[http://www.ks.uiuc.edu/Research/namd/ NAMD] is a parallel, object-oriented molecular dynamics code designed		[http://www.ks.uiuc.edu/Research/namd/ NAMD] is a parallel, object-oriented molecular dynamics code designed
	for high-performance simulation of large biomolecular systems.		for high-performance simulation of large biomolecular systems.
Line 9:		Line 8:
	The NAMD project is funded by the National Institutes of Health (grant number PHS 5 P41 RR05969).		The NAMD project is funded by the National Institutes of Health (grant number PHS 5 P41 RR05969).

	== ~~NAMD on Carbon~~ ==		== MPI version ==
	~~=== Required reading ===~~
	~~[[HPC/Getting_started]]~~
	=== Modules ===		=== Modules ===
	* As of NAMD-2.9, the ~~named~~ module requires [[../fftw3\|FFTW-3]]. You can load both modules together:		* As of NAMD-2.9, the namd module requires [[../fftw3\|FFTW-3]]. You can load both modules together:
	module load fftw3 namd		module load fftw3 namd

	=== ~~Special instructions~~ ===		=== Using MPI ===
	* NAMD uses its own MPI wrapper, called [http://charm.cs.uiuc.edu/ Charm++], but the default <code>mpirun</code> from OpenMPI works just fine.		* NAMD uses its own MPI wrapper, called [http://charm.cs.uiuc.edu/ Charm++], but the default <code>mpirun</code> from OpenMPI works just fine.
	* Use the [[HPC/Submitting and Managing Jobs/Example Job Script#Generic job scripts\|Generic job template]]		* Use the [[HPC/Submitting and Managing Jobs/Example Job Script#Generic job scripts\|Generic job template]]
Line 22:		Line 19:
	#!/bin/bash		#!/bin/bash
	…		…

	cd $PBS_O_WORKDIR		cd $PBS_O_WORKDIR
	mpirun -machinefile $PBS_NODEFILE -np $PBS_NP \		mpirun -machinefile $PBS_NODEFILE -np $PBS_NP \
	namd2 ~~inputfile~~.namd		namd2 file.namd
	</syntaxhighlight>		</syntaxhighlight>

	== ~~CUDA~~ ==		== Using GPUs and MPI ==
	As of module version namd/2.9plus-MPI-icc-3 the module provides two binaries, a regular one called <code>namd2</code> (as usual) and a GPU-eneabled binary called <code>namd2-cuda</code>.		As of module version namd/2.9plus-MPI-icc-3 the module provides two binaries, a regular one called <code>namd2</code> (as usual) and a GPU-eneabled binary called <code>namd2-cuda</code>. The latter requires the cuda module as additional prerequisite:
			module load fftw3
			module load cuda
			module load namd/2.9plus-MPI-icc-3
	<!--		<!--
	; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].		; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].
Line 48:		Line 47:
	…		…
	mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \		mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \
	namd2'''-cuda +idlepoll''' ''namd~~.conf~~''		namd2'''-cuda +idlepoll''' ''file.namd''
	* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>.		* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>.
	* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.		* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.

Stern: /* CUDA */

2012-11-08T21:33:04Z

CUDA

← Older revision		Revision as of 21:33, November 8, 2012
Line 29:		Line 29:

	== CUDA ==		== CUDA ==
			As of module version namd/2.9plus-MPI-icc-3 the module provides two binaries, a regular one called <code>namd2</code> (as usual) and a GPU-eneabled binary called <code>namd2-cuda</code>.
			<!--
	; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].		; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].
	* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions		* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions
	~~<!--~~ * http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA (namd-2.7) -->		* http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA (namd-2.7) -->
	* http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html ~~– running with CUDA (namd current)~~
			The relevant paragraphs from the [http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html NAMD User Guide] are:
	<blockquote>		<blockquote>
	Energy evaluation is slower than calculating forces alone, and the loss is much greater in CUDA-accelerated builds. Therefore you should set outputEnergies to 100 or higher in the simulation config file. Some features are unavailable in CUDA builds, including alchemical free energy perturbation and the Lowe-Andersen thermostat.		Energy evaluation is slower than calculating forces alone, and the loss is much greater in CUDA-accelerated builds. Therefore you should set outputEnergies to 100 or higher in the simulation config file. Some features are unavailable in CUDA builds, including alchemical free energy perturbation and the Lowe-Andersen thermostat.
	<br>		<br>
	As this is a new feature you are encouraged to test all simulations before beginning production runs. Forces evaluated on the GPU differ slightly from a CPU-only calculation, an effect more visible in reported scalar pressure values than in energies.		As this is a new feature you are encouraged to test all simulations before beginning production runs. Forces evaluated on the GPU differ slightly from a CPU-only calculation, an effect more visible in reported scalar pressure values than in energies.
			<br.>… <br>
			Each namd2 thread can use only one GPU. Therefore you will need to run at least one thread for each GPU you want to use. Multiple threads can share a single GPU, usually with an increase in performance. NAMD will automatically distribute threads equally among the GPUs on a node.
	</blockquote>		</blockquote>

Line 44:		Line 49:
	mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \		mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \
	namd2'''-cuda +idlepoll''' ''namd.conf''		namd2'''-cuda +idlepoll''' ''namd.conf''
			* Test and optimize the <code>''N''</code> and <code>''PPN''</code> parameters for your situation. Start with <code>nodes=1:ppn=4</code>.
			* The <code>gpus=…</code> modifier refers to GPUs ''per node'' and currently must always be 1.

	== Documentation ==		== Documentation ==

Stern: /* CUDA */

2012-11-08T21:22:28Z

CUDA

← Older revision		Revision as of 21:22, November 8, 2012
Line 31:		Line 31:
	; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].		; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].
	* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions		* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions
	* http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA (namd-2.7)		<!-- * http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA (namd-2.7) -->
	* http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html – running with CUDA (namd current)		* http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html – running with CUDA (namd current)
			<blockquote>
			Energy evaluation is slower than calculating forces alone, and the loss is much greater in CUDA-accelerated builds. Therefore you should set outputEnergies to 100 or higher in the simulation config file. Some features are unavailable in CUDA builds, including alchemical free energy perturbation and the Lowe-Andersen thermostat.
			<br>
			As this is a new feature you are encouraged to test all simulations before beginning production runs. Forces evaluated on the GPU differ slightly from a CPU-only calculation, an effect more visible in reported scalar pressure values than in energies.
			</blockquote>

			To run on GPU nodes:
			#PBS -l nodes=''N'':ppn=''PPN''''':gpus=1'''
			…
			mpirun -machinfile $PBS_NODEFILE -np $PBS_NP \
			namd2'''-cuda +idlepoll''' ''namd.conf''

	== Documentation ==		== Documentation ==

Stern: /* CUDA */

2012-11-05T14:45:50Z

CUDA

← Older revision		Revision as of 14:45, November 5, 2012
Line 30:		Line 30:
	== CUDA ==		== CUDA ==
	; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].		; Update 2012-11-02: This section is in development. --[[User:Stern\|stern]].
	* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions.		* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions
	* http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA.		* http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA (namd-2.7)
	* http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html		* http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html – running with CUDA (namd current)

	== Documentation ==		== Documentation ==

Stern: /* CUDA */

2012-11-02T21:01:22Z

CUDA

← Older revision		Revision as of 21:01, November 2, 2012
Line 32:		Line 32:
	* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions.		* http://www.ks.uiuc.edu/Research/namd/development.html – build instructions.
	* http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA.		* http://www.ks.uiuc.edu/Research/namd/2.7/ug/node72.html – running with CUDA.
			* http://www.ks.uiuc.edu/Research/namd/2.9/ug/node88.html

	== Documentation ==		== Documentation ==