Table 8 Iteration counts and time measurements per time step for comparison of physics-based block preconditioners and the newly proposed hybrid preconditioner in the Coriolis flow meter example
From: A hybrid interface preconditioner for monolithic fluid–structure interaction solvers
Mesh ID | # of linear iterations | Solver time | Setup time | ||||||
|---|---|---|---|---|---|---|---|---|---|
cor1 | cor2 | cor3 | cor1 | cor2 | cor3 | cor1 | cor2 | cor3 | |
M | 14 | 28 | 56 | 14 | 28 | 56 | 14 | 28 | 56 |
BGS(AMG) | 226 | 458 | 597 | \({14.7}\,{\hbox {s}}\) | \({44.9}\,{\hbox {s}}\) | \({66.6}\,{\hbox {s}}\) | \({9.8}\,{\hbox {s}}\) | \({11.1}\,{\hbox {s}}\) | \({11.3}\,{\hbox {s}}\) |
H-BGS(AMG) | 176 | 223 | 415 | \({14.6}\,{\hbox {s}}\) | \({28.3}\,{\hbox {s}}\) | \({59.1}\,{\hbox {s}}\) | \({11.8}\,{\hbox {s}}\) | \({11.7}\,{\hbox {s}}\) | \({14.6}\,{\hbox {s}}\) |
Savings (%) | 22.1 | 51.3 | 30.5 | 0.5 | 37.0 | 11.3 | − | − | − |
AMG(BGS) | 92 | 307 | 428 | \({23.2}\,{\hbox {s}}\) | \({136.8}\,{\hbox {s}}\) | \({158.3}\,{\hbox {s}}\) | \({10.1}\,{\hbox {s}}\) | \({11.5}\,{\hbox {s}}\) | \({11.4}\,{\hbox {s}}\) |
H-AMG(BGS) | 85 | 161 | 223 | \({23.0}\,{\hbox {s}}\) | \({77.8}\,{\hbox {s}}\) | \({90.3}\,{\hbox {s}}\) | \({10.7}\,{\hbox {s}}\) | \({12.1}\,{\hbox {s}}\) | \({13.5}\,{\hbox {s}}\) |
Savings (%) | 7.6 | 47.6 | 47.9 | 0.8 | 43.1 | 43.0 | − | − | − |