Skip to content
Snippets Groups Projects
Select Git revision
  • 339d4a342c4da20e3f3a49ec6346e81b77e7fd5f
  • develop default protected
  • feature/kinetic-schemes
  • feature/reconstruction
  • feature/local-dt-fsi
  • feature/composite-scheme-sources
  • feature/composite-scheme-other-fluxes
  • feature/serraille
  • feature/variational-hydro
  • feature/composite-scheme
  • hyperplastic
  • feature/polynomials
  • feature/gks
  • feature/implicit-solver-o2
  • feature/coupling_module
  • feature/implicit-solver
  • feature/merge-local-dt-fsi
  • master protected
  • feature/escobar-smoother
  • feature/hypoelasticity-clean
  • feature/hypoelasticity
  • v0.5.0 protected
  • v0.4.1 protected
  • v0.4.0 protected
  • v0.3.0 protected
  • v0.2.0 protected
  • v0.1.0 protected
  • Kidder
  • v0.0.4 protected
  • v0.0.3 protected
  • v0.0.2 protected
  • v0 protected
  • v0.0.1 protected
33 results

pugs

Stéphane Del Pino's avatar
Stéphane Del Pino authored
The new class is meant to be much faster
- it requires an estimation of the number values per row
- if the number of stored values per line exceeds the reserved one, an
a special storage zone (per row) is used.

It is important to give a precise estimation of the number of values
per rows:
- if the provided expected number of values is too small with regard
to the real number of values stored for a given row, then building
performances can be noticeably affected
- if the provided expected number of values per row corresponds
exactly to the effective number of stored values, then the
construction of the CRSMatrix costs nothing.

In the later case, one benefits from a significant CPU and memory
costs reduction.
34459135
History
Name Last commit Last update