<div>Thanks, the synopsys TCAD product announced they use 128 bit float for numerically stable reason.<br>I will try some more robust methods first.<br>float128 seems not compatible with most external direct solvers.<br> <br></div><blockquote style="padding-left:5px;margin-left:5px;border-left:#a0c6e5 2px solid;margin-right:0px">On Sun, Sep 18, 2011 at 17:07, Gong Ding <span dir="ltr"><<a href="mailto:gdiso@ustc.edu">gdiso@ustc.edu</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div>Do you mean GMRES-like method will be <font>benefited</font> from float 128?</div></blockquote></div><br><div>It could be, especially if you use a finite difference Jacobian. If you suspect precision issues, you can try more robust orthogonalization -ksp_gmres_modifiedgramschmidt or -ksp_type gcr, or you can just try using float128.<br><br><br></div>
</blockquote><span ></span>