Yes, a good linear solver involves a suitable choice of Krylov method and preconditioner.<br><br><div class="gmail_quote">On Mon, Jun 4, 2012 at 7:37 AM, w_ang_temp <span dir="ltr"><<a href="mailto:w_ang_temp@163.com" target="_blank">w_ang_temp@163.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div style="line-height:1.7;font-size:14px;font-family:arial"><div>Hello<br> As I become familiar with PETSc, I make a conclusion about the question and hope that it is right.<br>
In order to get satisfactory results when solving a linear system with KSP, the Krylov space methods</div>
<div>are typically used in conjunction with a preconditioner. So the user should select the effective Krylov</div>
<div>method(such as cg) and preconditioner(such as jacobi) for this linear system Ax=b. Or it maybe not</div>
<div>convergence.<br> Thanks.<br> Jim<br><br></div><div class="im">
<div></div>
<div></div>
<div><br></div>>At 2012-05-26 23:05:26,w_ang_temp <<a href="mailto:w_ang_temp@163.com" target="_blank">w_ang_temp@163.com</a>> wrote:<br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px 0px 0.8ex;PADDING-LEFT:1ex">
<div style="line-height:1.7;font-size:14px;font-family:arial">
<div>>Hello, Matt</div>
<div>> The results of the parallel and the serial are <span>identical by the "-ksp_rtol 1.0e-15 -ksp_converged_reason -ksp_monitor_true_residual".</span></div>
<div><span>> I will make a <span>careful</span> <span>analysis</span> of these <span>parameters next day because it is late night here now.</span></span></div>
<div><span><span>> Thanks again. Jim</span></span><br></div>>在 2012-05-26 22:26:20,"Matthew Knepley" <<a href="mailto:knepley@gmail.com" target="_blank">knepley@gmail.com</a>> 写道:<br>
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px 0px 0.8ex;PADDING-LEFT:1ex">>On Sat, May 26, 2012 at 2:22 PM, w_ang_temp <span dir="ltr"><<a href="mailto:w_ang_temp@163.com" target="_blank">w_ang_temp@163.com</a>></span> wrote:<br>
<div class="gmail_quote">
<blockquote style="BORDER-LEFT:#ccc 1px solid;MARGIN:0px 0px 0px 0.8ex;PADDING-LEFT:1ex" class="gmail_quote">
<div style="LINE-HEIGHT:1.7;FONT-FAMILY:arial;FONT-SIZE:14px">
<div>>Hello,Matt<br>> Thanks again.<br>> What you said("Use -ksp_type preonly -pc_type lu") means that I should use a direct solver.And I am puzzled.<br></div></div></blockquote>
<div><br></div>
<div>>Okay, you want to remove variables from the investigation. You say there is a difference between</div>
<div>>the parallel and serial solve. We do this all the time, so it is not a bug. It is an issue of understanding.</div>
<div>>First, take away the tolerance issues:</div>
<div><br></div>> -ksp_rtol 1.0e-15 -ksp_converged_reason -ksp_monitor_true_residual</div>
<div class="gmail_quote"><br>
<div>>and send output of the serial and parallel run.</div>
<div><br></div>
<div>> Matt</div></div></blockquote></div></blockquote></div></div><br><br><span title="neteasefooter"><span></span></span></blockquote></div><br>