Difference between revisions of "VASP best practices (NSC-UPPMAX January 2015)"

From SNIC Documentation
Jump to: navigation, search
(Seminar overview)
m (Seminar overview)
 
(6 intermediate revisions by one other user not shown)
Line 1: Line 1:
 
{{training event info
 
{{training event info
|description=Seminar on running VASP efficiently at Triolith and Beskow (the new Cray XC40 at PDC)
+
|description=Seminar on running VASP efficiently at Triolith and Beskow (the new Cray XC-40 at PDC)
 
|fields=Computational materials science
 
|fields=Computational materials science
 
|event type=Seminar
 
|event type=Seminar
Line 30: Line 30:
  
 
Lecture 1:  
 
Lecture 1:  
# parallel efficiency and running economically  
+
# Parallel efficiency and running economically  
# out of memory  
+
# Out of memory  
# typical bottlenecks for DFT codes and what it means when choosing hardware
+
# Typical bottlenecks for DFT codes and what it means when choosing hardware
  
Lecture 2: NPAR/NCORE etc
+
Lecture 2: NPAR, NSIM, KPAR etc
  
Lecture 3: Running big VASP calculations in parallel, some case studies from Beskow
+
Lecture 3: Running big VASP calculations in parallel, some case studies from Beskow and Triolith
  
 
== Trainer ==
 
== Trainer ==
The seminar is given by Peter Larsson from NSC
+
The seminar is given by [[User:Peter Larsson (NSC)|Peter Larsson]] from [[NSC]]
  
 
== Venue ==
 
== Venue ==
 
Lecture room Å80101 at the Ångström Laboratory
 
Lecture room Å80101 at the Ångström Laboratory

Latest revision as of 10:05, 28 November 2014

Name VASP best practices (NSC-UPPMAX January 2015)
Description Seminar on running VASP efficiently at Triolith and Beskow (the new Cray XC-40 at PDC)
Type of event Seminar
Location Uppsala
Start date 2015-01-15
End date 2015-01-15

Seminar overview

Schedule
Thursday, January 15
13.15-14.00 General performance aspects of DFT codes
14.15-15.00 Influential VASP settings
15.00-15.30 Coffee break
15.30-16.30 VASP in parallel - special focus on PDC's new Cray system
16.30-17.00 Discussions/Q&A

Lecture 1:

  1. Parallel efficiency and running economically
  2. Out of memory
  3. Typical bottlenecks for DFT codes and what it means when choosing hardware

Lecture 2: NPAR, NSIM, KPAR etc

Lecture 3: Running big VASP calculations in parallel, some case studies from Beskow and Triolith

Trainer

The seminar is given by Peter Larsson from NSC

Venue

Lecture room Å80101 at the Ångström Laboratory