<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; ">For PS, look in psPromotionManager.cpp, here the kernel you need to trivially tweak:<div><br></div><div><font face="Monaco" style="font-size: 11px;">template <class T> void PSPromotionManager::process_array_chunk_work(<br> oop obj,<br> int start, int end) {<br> assert(start <= end, "invariant");<br> T* const base = (T*)objArrayOop(obj)->base();<br> T* p = base + start;<br> T* const chunk_end = base + end;<br> while (p < chunk_end) {<br> if (PSScavenge::should_scavenge(p)) {<br> claim_or_forward_depth(p);<br> }<br> ++p;<br> }<br>}</font></div><div><br><div><div>Like Tony and Thomas said before, you'll still be seeing "surprises" due to array chunking and work stealing. Those, I guess, you'll just have to live with.</div><div><br></div><div>igor</div><div><br></div><div>On Sep 4, 2013, at 1:34 PM, Aleksey Shipilev <<a href="mailto:aleksey.shipilev@oracle.com">aleksey.shipilev@oracle.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite">Here you have it, thanks Igor. <br>Any reference to the relevant block of code?<br>I can probably try to fix this in background.<br><br>-Aleksey.<br><br>On 05.09.2013, at 0:00, Igor Veresov <<a href="mailto:iggy.veresov@gmail.com">iggy.veresov@gmail.com</a>> wrote:<br><br><blockquote type="cite">Yup, that's a depth-first array-scanning quirk. The work-stealing is done using stacks, so in order to have the first fields followed first the references need to be put of stack in reverse. That's done for regular objects but for arrays it's not. <br><br>igor<br><br>On Sep 4, 2013, at 12:51 PM, Aleksey Shipilev <<a href="mailto:aleksey.shipilev@oracle.com">aleksey.shipilev@oracle.com</a>> wrote:<br><br><blockquote type="cite">Hi Jon,<br><br>On 09/04/2013 10:19 PM, Jon Masamitsu wrote:<br><blockquote type="cite">I haven't followed this thread carefully enough but the ParallelGC <br>collector uses a depth-first traversal while the other collectors use<br>a breadth-first. Would that explain the difference?<br></blockquote><br>The referenced objects in the array are the leaves in reachability<br>graph. I thought there is no difference in depth- vs. breadth-first in<br>this case? It looks more like we record the traversed objects on some<br>LIFO structure, which polls the elements in the reverse order.<br><br>-Aleksey.<br></blockquote><br></blockquote></blockquote></div><br></div></body></html>