<div dir="ltr">Hi Vladis,<div><br></div><div>Thanks for this response.</div><div>I was trying to use the Clavis scheduler: <a href="http://clavis.sourceforge.net/" target="_blank">http://clavis.<wbr>sourceforge.net/</a></div><div>In this scheduler they go through an important effort to migrate only currently used pages of a thread (using AMD IBS hardware counters). This could be implemented more neatly using page tables at the kernel level...</div><div><br></div><div>Regards,</div><div><br></div><div class="gmail_extra"><br><div class="gmail_quote">2018-01-19 14:26 GMT-05:00 <span dir="ltr"><<a href="mailto:valdis.kletnieks@vt.edu" target="_blank">valdis.kletnieks@vt.edu</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span>On Fri, 19 Jan 2018 13:57:18 -0500, Karaoui mohamed lamine said:<br>
><br>
> Does the "migrate_pages" syscall migrate memory pages in a *lazy* manner?<br>
> if not, is there another syscall or patch that does it?<br>
<br>
</span>man 2 migrate_pages says:<br>
<br>
RETURN VALUE<br>
On success migrate_pages() returns the number of pages that could not be moved (i.e., a return of zero means<br>
that all pages were successfully moved). On error, it returns -1, and sets errno to indicate the error.<br>
<br>
which implies that it doesn't return until an attempt has been made to migrate all relevant<br>
pages. And that rules out lazy migration.<br>
<br>
What problem are you trying to solve by doing it in a lazy manner?<br>
</blockquote></div><br></div></div>