NCORE - Revision history

Huebsch at 09:37, 18 March 2026

2026-03-18T09:37:02Z

← Older revision		Revision as of 09:37, 18 March 2026
Line 43:		Line 43:
	{{TAG\|LSCALAPACK}},		{{TAG\|LSCALAPACK}},
	{{TAG\|LSCAAWARE}},		{{TAG\|LSCAAWARE}},

	[[GPU ports of VASP]],		[[GPU ports of VASP]],
	[[Combining MPI and OpenMP]],		[[Combining MPI and OpenMP]],

	[[Optimizing the parallelization]],		[[Optimizing the parallelization]],
	[[~~:Category:~~Parallelization]],		[[Parallelization]],
	[[Energy cutoff and FFT meshes]]		[[Energy cutoff and FFT meshes]]

	{{sc\|NCORE\|HowTo\|Workflows that use this tag}}		{{sc\|NCORE\|HowTo\|Workflows that use this tag}}
	~~----~~

	[[Category:INCAR tag]][[Category:Performance]][[Category:Parallelization]]		[[Category:INCAR tag]][[Category:Performance]][[Category:Parallelization]]

Wolloch at 09:18, 18 March 2026

2026-03-18T09:18:51Z

Show changes

Wolloch at 08:20, 16 March 2026

2026-03-16T08:20:41Z

← Older revision		Revision as of 08:20, 16 March 2026
Line 64:		Line 64:
	{{TAG\|LSCALAPACK}},		{{TAG\|LSCALAPACK}},
	{{TAG\|LSCAAWARE}},		{{TAG\|LSCAAWARE}},
	[[~~OpenACC_GPU_port_of_VASP~~\|~~OpenACC~~ GPU ~~Port~~ of VASP]],		[[GPU ports of VASP\|GPU ports of VASP]],
	[[Combining MPI and OpenMP]]		[[Combining MPI and OpenMP]]

Huebsch at 09:23, 25 May 2022

2022-05-25T09:23:58Z

← Older revision		Revision as of 09:23, 25 May 2022
Line 65:		Line 65:
	{{TAG\|LSCAAWARE}},		{{TAG\|LSCAAWARE}},
	[[OpenACC_GPU_port_of_VASP\|OpenACC GPU Port of VASP]],		[[OpenACC_GPU_port_of_VASP\|OpenACC GPU Port of VASP]],
	[[~~Hybrid_MPI/OpenMP_parallelization\|Hybrid~~ MPI/OpenMP ~~parallelization~~]]		[[Combining MPI and OpenMP]]

	{{sc\|NCORE\|Examples\|Examples that use this tag}}		{{sc\|NCORE\|Examples\|Examples that use this tag}}

Karsai: /* Related Tags and Sections */

2022-04-08T14:47:21Z

Related Tags and Sections

← Older revision		Revision as of 14:47, 8 April 2022
Line 56:		Line 56:
	</pre>		</pre>

	== Related ~~Tags~~ and ~~Sections~~ ==		== Related tags and articles ==
	{{TAG\|NPAR}},		{{TAG\|NPAR}},
	{{TAG\|LPLANE}},		{{TAG\|LPLANE}},
Line 70:		Line 70:
	----		----

	[[Category:INCAR]][[Category:Performance]][[Category:parallelization]]		[[Category:INCAR tag]][[Category:Performance]][[Category:parallelization]]

Kresse at 17:24, 23 November 2021

2021-11-23T17:24:06Z

← Older revision		Revision as of 17:24, 23 November 2021
Line 5:		Line 5:
	----		----
	VASP currently offers parallelization and data distribution over bands and/or over plane wave coefficients, and as of VASP.5.3.2, parallelization over '''k'''-points (no data distribution, see {{TAG\|KPAR}}).		VASP currently offers parallelization and data distribution over bands and/or over plane wave coefficients, and as of VASP.5.3.2, parallelization over '''k'''-points (no data distribution, see {{TAG\|KPAR}}).
	To ~~obtain~~ high efficiency on massively parallel systems or modern multi-core machines, it is strongly recommended to use all parallelization options available. Most algorithms work with any data distribution (except for the single band conjugated gradient, which is ~~considered to be~~ obsolete).		To achieve high efficiency on massively parallel systems or modern multi-core machines, it is strongly recommended to use all parallelization options available. Most algorithms work with any data distribution (except for the single band conjugated gradient, which is obsolete).

	{{TAG\|NCORE}} is available from VASP.5.2.13 on, and is more handy than the previous parameter {{TAG\|NPAR}}. The user should either specify {{TAG\|NCORE}} or {{TAG\|NPAR}}, where {{TAG\|NPAR}} takes a higher preference. The relation between both parameters is		{{TAG\|NCORE}} is available from VASP.5.2.13 on, and is more handy than the previous parameter {{TAG\|NPAR}}. The user should either specify {{TAG\|NCORE}} or {{TAG\|NPAR}}, where {{TAG\|NPAR}} takes a higher preference. The relation between both parameters is

	NCORE =number-of-cores / NPAR		NCORE =number-of-cores /KPAR / NPAR

	{{TAG\|NCORE}} determines how many cores share the work on an individual orbital. The current default is {{TAG\|NCORE}}=1, meaning that one orbital is treated by one core. {{TAG\|NPAR}} is then set to the total number of cores. If {{TAG\|NCORE}} equals the total number of cores, {{TAG\|NPAR}} is set to 1. This implies data distribution over plane wave coefficients only: all cores will work together on every individual band, i.e., the plane wave coefficients of each band are distributed over all cores. This is usually very slow and should be avoided.		{{TAG\|NCORE}} determines how many cores share the work on an individual orbital. The current default is {{TAG\|NCORE}}=1, meaning that one orbital is treated by one core. {{TAG\|NPAR}} is then set to the total number of cores (divided by KPAR). If {{TAG\|NCORE}} equals the total number of cores, {{TAG\|NPAR}} is set to 1. This implies data distribution over plane wave coefficients only: all cores will work together on every individual band, i.e., the plane wave coefficients of each band are distributed over all cores. This is usually very slow and should be avoided.

	{{TAG\|NCORE}}=1 is the optimal setting for small unit cells, and platforms with a small communication bandwidth. It is also a good choice for up to 8 cores. However, this mode substantially increases the memory requirements, because the non-local projector functions must be stored entirely on each core. In addition, substantial all-to-all communications are required to orthogonalize the bands.		{{TAG\|NCORE}}=1 is the optimal setting for small unit cells, and platforms with a small communication bandwidth. It is also a good choice for up to 8 cores. However, this mode substantially increases the memory requirements, because the non-local projector functions must be stored entirely on each core. In addition, substantial all-to-all communications are required to orthogonalize the bands.

	On massively parallel systems and modern multi-core machines we strongly ~~urge~~ to set		On massively parallel systems and modern multi-core machines we strongly recommend to set

	NCORE = 2 up to number-of-cores-per-socket (or number-of-cores-per-node)		NCORE = 2 up to number-of-cores-per-socket (or number-of-cores-per-node)


	For large unit cells, we found that this can improve the performance by up to a factor four compared to the default, and it also significantly improves the stability of the code due to reduced memory requirements. Ideally, {{TAG\|NCORE}} should be a factor of the number-of-cores-per-socket (or number-of-cores-per-node), since this reduces communication between the sockets or nodes. The best value {{TAG\|NCORE}} depends somewhat on the number of atoms		For large unit cells, we found that this can improve the performance by up to a factor four compared to the default, and it also significantly improves the stability of the code due to reduced memory requirements. Ideally, {{TAG\|NCORE}} should be a factor of the number-of-cores-per-socket (or number-of-cores-per-node), since this reduces communication between the sockets or nodes. The best value {{TAG\|NCORE}} depends somewhat on the number of atoms in the unit cell. Values around 4 are usually ideal for 100 atoms in the unit cell. For very large unit cells (more than 400 atoms) values around 12-16 are often optimal. If you run extensive simulations for similar systems, make your own tests.
	in the unit cell. Values around 4 are usually ideal for 100 atoms in the unit cell. For very large unit cells (more than 400 atoms) values around 12-16 are often optimal. If you run extensive simulations for similar systems, make your own tests.

Vaspmaster: /* Related Tags and Sections */

2021-02-15T17:14:36Z

Related Tags and Sections

← Older revision		Revision as of 17:14, 15 February 2021
Line 65:		Line 65:
	{{TAG\|LSCALAPACK}},		{{TAG\|LSCALAPACK}},
	{{TAG\|LSCAAWARE}},		{{TAG\|LSCAAWARE}},
	[[OpenACC_GPU_port_of_VASP\|~~The~~ OpenACC GPU~~-port~~ of VASP]],		[[OpenACC_GPU_port_of_VASP\|OpenACC GPU Port of VASP]],
	[[Hybrid_MPI/OpenMP_parallelization\|Hybrid MPI/OpenMP parallelization]]		[[Hybrid_MPI/OpenMP_parallelization\|Hybrid MPI/OpenMP parallelization]]

Vaspmaster: /* Related Tags and Sections */

2021-02-15T17:12:45Z

Related Tags and Sections

← Older revision		Revision as of 17:12, 15 February 2021
Line 64:		Line 64:
	{{TAG\|KPAR}},		{{TAG\|KPAR}},
	{{TAG\|LSCALAPACK}},		{{TAG\|LSCALAPACK}},
	{{TAG\|LSCAAWARE}}		{{TAG\|LSCAAWARE}},
			[[OpenACC_GPU_port_of_VASP\|The OpenACC GPU-port of VASP]],
			[[Hybrid_MPI/OpenMP_parallelization\|Hybrid MPI/OpenMP parallelization]]

	{{sc\|NCORE\|Examples\|Examples that use this tag}}		{{sc\|NCORE\|Examples\|Examples that use this tag}}

Miranda.henrique at 13:51, 17 December 2020

2020-12-17T13:51:14Z

← Older revision		Revision as of 13:51, 17 December 2020
Line 5:		Line 5:
	----		----
	VASP currently offers parallelization and data distribution over bands and/or over plane wave coefficients, and as of VASP.5.3.2, parallelization over '''k'''-points (no data distribution, see {{TAG\|KPAR}}).		VASP currently offers parallelization and data distribution over bands and/or over plane wave coefficients, and as of VASP.5.3.2, parallelization over '''k'''-points (no data distribution, see {{TAG\|KPAR}}).
	To obtain high efficiency on massively parallel systems or modern multi-core machines, it is strongly recommended to use all parallelization options ~~at the same time~~. Most algorithms work with any data distribution (except for the single band conjugated gradient, which is considered to be obsolete).		To obtain high efficiency on massively parallel systems or modern multi-core machines, it is strongly recommended to use all parallelization options available. Most algorithms work with any data distribution (except for the single band conjugated gradient, which is considered to be obsolete).

	{{TAG\|NCORE}} is available from VASP.5.2.13 on, and is more handy than the previous parameter {{TAG\|NPAR}}. The user should either specify {{TAG\|NCORE}} or {{TAG\|NPAR}}, where {{TAG\|NPAR}} takes a higher preference. The relation between both parameters is		{{TAG\|NCORE}} is available from VASP.5.2.13 on, and is more handy than the previous parameter {{TAG\|NPAR}}. The user should either specify {{TAG\|NCORE}} or {{TAG\|NPAR}}, where {{TAG\|NPAR}} takes a higher preference. The relation between both parameters is
Line 17:		Line 17:
	On massively parallel systems and modern multi-core machines we strongly urge to set		On massively parallel systems and modern multi-core machines we strongly urge to set

	NCORE = 2 - number-of-cores-per-socket (or number-of-cores-per-node)		NCORE = 2 up to number-of-cores-per-socket (or number-of-cores-per-node)


Line 31:		Line 31:
	<pre>		<pre>
	LPLANE = .TRUE.		LPLANE = .TRUE.
	NCORE = 2 - number-of-cores-per-node		NCORE = 2 up to number-of-cores-per-node
	LSCALU = .FALSE.		LSCALU = .FALSE.
	NSIM = 4		NSIM = 4

Kresse at 13:17, 14 October 2020

2020-10-14T13:17:56Z

← Older revision		Revision as of 13:17, 14 October 2020
Line 47:		Line 47:
	:Mind that you need at least a 100 Mbit full duplex network, with a fast switch offering at least 2 Gbit switch capacity to find useful speedups. Multi-core machines should be always linked by an Infiniband, since Gbit is too slow for multi-core machines.		:Mind that you need at least a 100 Mbit full duplex network, with a fast switch offering at least 2 Gbit switch capacity to find useful speedups. Multi-core machines should be always linked by an Infiniband, since Gbit is too slow for multi-core machines.

	*Massively parallel machines (Cray~~, Blue Gene~~):		*Massively parallel machines with dedicated network (maybe Cray):
	:On ~~many~~ massively parallel machines one is sometimes forced to use a large number of cores. In this case load balancing problems and problems with the communication bandwidth are likely to be experienced. In addition the local memory is fairly small on some massively parallel machines (too small keep the real space projectors in the cache using any reasonable VASP setting). Therefore, we recommend to set {{TAG\|NCORE}} on these machines to √''# of cores'' (explicit timing can be helpful to find the optimum value). The use of {{TAG\|LPLANE}}=.TRUE. is only recommended if the number of nodes is significantly smaller than {{TAG\|NGX}}, {{TAG\|NGY}} and {{TAG\|NGZ}}.		:On massively parallel machines one is sometimes forced to use a large number of cores. In this case load balancing problems and problems with the communication bandwidth are likely to be experienced. In addition the local memory is fairly small on some massively parallel machines (too small keep the real space projectors in the cache using any reasonable VASP setting). Therefore, we recommend to set {{TAG\|NCORE}} on these machines to √''# of cores'' (explicit timing can be helpful to find the optimum value). The use of {{TAG\|LPLANE}}=.TRUE. is only recommended if the number of nodes is significantly smaller than {{TAG\|NGX}}, {{TAG\|NGY}} and {{TAG\|NGZ}}.

	:In summary, the following setting is recommended		:In summary, the following setting is recommended

← Older revision		Revision as of 09:23, 25 May 2022
Line 65:		Line 65:
	{{TAG\|LSCAAWARE}},		{{TAG\|LSCAAWARE}},
	[[OpenACC_GPU_port_of_VASP\|OpenACC GPU Port of VASP]],		[[OpenACC_GPU_port_of_VASP\|OpenACC GPU Port of VASP]],
	[[~~Hybrid_MPI/OpenMP_parallelization\|Hybrid~~ MPI/OpenMP ~~parallelization~~]]		[[Combining MPI and OpenMP]]

	{{sc\|NCORE\|Examples\|Examples that use this tag}}		{{sc\|NCORE\|Examples\|Examples that use this tag}}

← Older revision		Revision as of 17:14, 15 February 2021
Line 65:		Line 65:
	{{TAG\|LSCALAPACK}},		{{TAG\|LSCALAPACK}},
	{{TAG\|LSCAAWARE}},		{{TAG\|LSCAAWARE}},
	[[OpenACC_GPU_port_of_VASP\|~~The~~ OpenACC GPU~~-port~~ of VASP]],		[[OpenACC_GPU_port_of_VASP\|OpenACC GPU Port of VASP]],
	[[Hybrid_MPI/OpenMP_parallelization\|Hybrid MPI/OpenMP parallelization]]		[[Hybrid_MPI/OpenMP_parallelization\|Hybrid MPI/OpenMP parallelization]]

← Older revision		Revision as of 17:12, 15 February 2021
Line 64:		Line 64:
	{{TAG\|KPAR}},		{{TAG\|KPAR}},
	{{TAG\|LSCALAPACK}},		{{TAG\|LSCALAPACK}},
	{{TAG\|LSCAAWARE}}		{{TAG\|LSCAAWARE}},
			[[OpenACC_GPU_port_of_VASP\|The OpenACC GPU-port of VASP]],
			[[Hybrid_MPI/OpenMP_parallelization\|Hybrid MPI/OpenMP parallelization]]

	{{sc\|NCORE\|Examples\|Examples that use this tag}}		{{sc\|NCORE\|Examples\|Examples that use this tag}}