<html><head><meta http-equiv="Content-Type" content="text/html charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><br class=""><div><blockquote type="cite" class=""><div class="">On Dec 29, 2016, at 5:25 PM, Mark Adams <<a href="mailto:mfadams@lbl.gov" class="">mfadams@lbl.gov</a>> wrote:</div><br class="Apple-interchange-newline"><div class=""><div dir="ltr" class=""><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br class="">
BTW: Why use --download-fblaslapack on cray? [and not system blas - like cray-libsci]<br class=""></blockquote><div class=""><br class=""></div><div class="">Excellent question. Same answer as how did I spend a week preintf debugging in November. The LAPACK invert method in MKL (at NERSC) used in Plex gaves differences (the ones that I saw were small) that resulted the nonlinear solver stagnating. NERSC said they reported it.</div><div class=""><br class=""></div></div></div></div></div></blockquote><div><br class=""></div><div>Mark,</div><div><br class=""></div><div>I also had a similar problem on our XC40 Shaheen when using MKL . I fixed my problem by using an older MKL distribution. Assuming the same path, you can find it in <span style="background-color: rgb(255, 255, 255);" class="">/opt/intel/composer_xe_2015.1.133/mkl/lib/intel64/</span></div></div><br class=""><div class="">Stefano</div></body></html>