VASP versus ABINIT for parallel calculations

  • Context: Graduate 
  • Thread starter Thread starter olgranpappy
  • Start date Start date
  • Tags Tags
    Calculations Parallel
Click For Summary
SUMMARY

This discussion focuses on the comparison between VASP 5.2 and ABINIT for conducting parallel calculations in computational materials science. The user, Adam, is considering transitioning from ABINIT, which he finds problematic despite being free, to VASP due to the similar input file structures. A key concern raised is the efficiency of running large parallel jobs with 100-1000 processors, highlighting potential communication bottlenecks among processors during calculations.

PREREQUISITES
  • Understanding of VASP 5.2 and its parallel calculation capabilities
  • Familiarity with ABINIT and its "paral_kgb" variables
  • Knowledge of Hamiltonian matrix diagonalization in computational physics
  • Experience with parallel computing concepts and challenges
NEXT STEPS
  • Research VASP 5.2 parallelization techniques and best practices
  • Explore ABINIT's "paral_kgb" settings for optimizing performance
  • Investigate communication bottlenecks in parallel computing environments
  • Learn about scaling parallel jobs effectively across multiple processors
USEFUL FOR

This discussion is beneficial for computational physicists, materials scientists, and researchers involved in parallel computing who are evaluating software options for large-scale simulations.

olgranpappy
Homework Helper
Messages
1,271
Reaction score
3
Does anyone here have experience using VASP (especially VASP 5.2) for parallel calculations? I'm currently using ABINIT (abinip, and abinip with the "paral_kgb" variables) and although the price is right (free), it has been giving me some problems lately.

The input files for VASP and ABINIT seem to have a fairly similar structure, so I was thinking about buying VASP. Anyone have any advice in this regard? For example, how does ABINIT compare to VASP as far as large parallel jobs (~100--1000 processors) go?

Thanks,

Adam
 
Physics news on Phys.org
olgranpappy,

Did I read you right? 100-1000 processors? That's a lot of processors. Sure each processor can diagnolize a portion of the Hamiltonian matrix, but each processor needs to communicate with each other, which will bottleneck your calculation since there are so many processors. The most processors I've ever used was 8 processors for about 125-atoms.
 

Similar threads

Graduate DFT Code for Finite-Temperature Calculations
  • · Replies 1 ·
Replies
1
Views
3K
Requesting suggestions for languages, libraries, and architectures for parallel (and sometimes non parallel) numerical and scientific computations
  • · Replies 8 ·
Replies
8
Views
5K
Graduate Understanding NOAA and Naval Observatory's Solar Calculations
  • · Replies 16 ·
Replies
16
Views
12K
MCNP6.2: F2 tally and problem with the area calculation
  • · Replies 2 ·
Replies
2
Views
5K
Undergrad Progress In Calculating The HLbL Contribution To Muon g-2
  • · Replies 0 ·
Replies
0
Views
3K
Graduate Calculating Thermo Properties of Water in Excel
  • · Replies 2 ·
Replies
2
Views
2K
MATLAB MATLAB GPU Cross-Correlation Code Fails with OOM on 8 GB Supercomputer
  • · Replies 4 ·
Replies
4
Views
8K
Graduate 1000 Basis functions limitation - Gaussian 09?
  • · Replies 2 ·
Replies
2
Views
2K
Graduating soon with a BS in Physics
  • · Replies 4 ·
Replies
4
Views
2K
Advice on Pursuing a Career in Physics After Graduating
  • · Replies 14 ·
Replies
14
Views
2K