Skip to content

Memory mapping on disk
Compare
Choose a tag to compare
@abelcarreras abelcarreras released this 15 Jul 10:17
· 622 commits to master since this release
1.12
8939d29

Added new feature to use memory mapped on disk for most heavy arrays. RAM memory consumption decreases considerably but calculation is slower.
Added new feature to calculate the full power spectrum projected into one atom (no eigenvector necessary). Also works for DOS and thermal properties calculation.
Changed trapezoidal integration (using numpy) to Simpson integration (using scipy) along all the code.
Many bugs fixed including:

  • Force constants are no longer necessary to compute full power spectrum
  • Wrong results when also request "wave vector projected power spectrum" and "peak analysis" together in the same run.
Assets 2
Loading