NTAUPAR: Difference between revisions

From VASP Wiki
No edit summary
No edit summary
 
(12 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{TAGDEF|NTAUPAR|[integer]}}
{{TAGDEF|NTAUPAR|[integer]}}
{{DEF|NTAUPAR|1|for [[Low scaling ACFDT/RPA and GW algorithms|ACFDT/RPA]] and [[Low scaling ACFDT/RPA and GW algorithms|GW calculations]] for large systems}}
{{DEF|NTAUPAR|depends on {{TAG|MAXMEM}} |used in low scaling [[GW approximation of Hedin's equations#lowGW|GW]] and RPA/ACFDT calculations.}}


Description: {{TAG|NTAUPAR}} available as of VASP.6, specifies the number of processor groups sharing imaginary time grid points.
Description: {{TAG|NTAUPAR}} available as of VASP.6, specifies the number of MPI groups sharing same imaginary time grid points. The default value of NTAUPAR is set to the largest possible value supported on the compute nodes to speed up the GW or RPA calculation.
----
----
== Related Tags and Sections ==
NTAUPAR has the biggest impact on memory usage as well as total runtime for low-scaling GW and RPA calculations. If not found in the {{FILE|INCAR}}, NTAUPAR is set automatically based on the value of {{TAG|MAXMEM}} (the available memory for each rank on each compute node), such that the GW and RPA job fits in the RAM on each compute node.
 
If {{TAG|MAXMEM}} is not set, VASP looks in "/proc/meminfo" for "MemAvailable" to set {{TAG|MAXMEM}} internally, otherwise the code uses the value provided in the {{FILE|INCAR}}.
 
NTAUPAR={{TAG|NOMEGA}} is the maximum value possible, while NTAUPAR=1 is the smallest possible value.
== Related tags and articles ==
{{TAG|NOMEGAPAR}},  
{{TAG|NOMEGAPAR}},  
{{TAG|NOMEGA}}
{{TAG|NOMEGA}}
Line 10: Line 15:
{{sc|NTAUPAR|Examples|Examples that use this tag}}
{{sc|NTAUPAR|Examples|Examples that use this tag}}
----
----
[[The_VASP_Manual|Contents]]


[[Category:INCAR]] [[Category:GW]] [[Category:ACFDT]][[Category:Parallelization]]
[[Category:INCAR tag]][[Category:Many-body perturbation theory]][[Category:GW]][[Category:ACFDT]][[Category:Performance]][[Category:Parallelization]][[Category:Low-scaling GW and RPA]][[Category:Memory]]

Latest revision as of 08:55, 21 February 2024

NTAUPAR = [integer] 

Default: NTAUPAR = depends on MAXMEM used in low scaling GW and RPA/ACFDT calculations.

Description: NTAUPAR available as of VASP.6, specifies the number of MPI groups sharing same imaginary time grid points. The default value of NTAUPAR is set to the largest possible value supported on the compute nodes to speed up the GW or RPA calculation.


NTAUPAR has the biggest impact on memory usage as well as total runtime for low-scaling GW and RPA calculations. If not found in the INCAR, NTAUPAR is set automatically based on the value of MAXMEM (the available memory for each rank on each compute node), such that the GW and RPA job fits in the RAM on each compute node.

If MAXMEM is not set, VASP looks in "/proc/meminfo" for "MemAvailable" to set MAXMEM internally, otherwise the code uses the value provided in the INCAR.

NTAUPAR=NOMEGA is the maximum value possible, while NTAUPAR=1 is the smallest possible value.

Related tags and articles

NOMEGAPAR, NOMEGA

Examples that use this tag