For the forward solve I use ASM+ILU in the same manner as for the adjoint problem. <div>The ASM not a bottleneck per se. Typically we see the adjoint problem taking the same amount of time as the non-linear problem for well-behaved flows, and the adjoint is shorter for less well-behaved flows. </div>
<div><br></div><div>The real problem I am having is for certain RANS cases, the frozen turbulence adjoint is extremely difficult to solve --- requiring GMRES subspace sizes on the order of 400-500 to converge. That's why I was investigating alternative preconditioning methods that could help solve these problems more efficiently. </div>
<div><br></div><div>Gaetan<br><br><div class="gmail_quote">On Mon, Apr 29, 2013 at 10:16 AM, Jed Brown <span dir="ltr"><<a href="mailto:jedbrown@mcs.anl.gov" target="_blank">jedbrown@mcs.anl.gov</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im">Gaetan Kenway <<a href="mailto:gaetank@gmail.com">gaetank@gmail.com</a>> writes:<br>
<br>
> The problems I am looking at are steady state or quasi-steady state (time<br>
> spectral approach). The example I sent before was steady state. The<br>
> nonlinear solver uses FAS multigrid (only used for full multigrid and<br>
> start-up on the fine grid) followed by an inexact Newton-Krylov method.<br>
<br>
</div>Is NK preconditioned by linear MG?<br>
<br>
Is the ASM preconditioner for the adjoint problem a bottleneck compared<br>
to the forward solve?<br>
</blockquote></div><br></div>