jcduell_at_lbl_dot_gov
Date: Fri Jul 23 2004 - 14:05:08 PDT
On Fri, Jul 23, 2004 at 04:11:59PM -0400, Grigory Bronevetsky wrote: > Yes, the checkpoints are restarting. I just couldn't piece together why > they were restarting given that the application seems to malloc more state > than the size of the checkpoint. I don't know this application very well > so it may be that it just never touches some of its data pages. Would such > malloc-ed but untouched pages be saved by BLCR? It may depend on your malloc implementation, but if the pages have never been used before (i.e. weren't handed back from a free() call), chances are the VM system has them all pointing to a common zero page until they're written on, and we aren't saving them. -- Jason Duell Future Technologies Group <jcduell_at_lbl_dot_gov> Computational Research Division Tel: +1-510-495-2354 Lawrence Berkeley National Laboratory