That was sort of my guess. Which is mildly ironic, because it suggests that C4 might run very well on top of Xen directly (because Xen provides a batch PTE update mechanism). In fact, I wonder if the VM support for LinuxVM might not provide a relevant and useful API here.
But I've also been pondering Zing, which apparently doesn't need these changes. I haven't seen any quantitative comparison of the two collectors. I'll see if I can get Gil to tell me something about that. Finally, I'm wondering if (and if so, how well) the whole thing works with weakly ordered memory systems. On Tue, Oct 1, 2013 at 11:55 PM, Florian Weimer <[email protected]> wrote: > * Jonathan S. Shapiro: > >> Does anybody know the details of how C4 wanted to manipulate page tables >> that Linux can't handle in it's current form? Why did they introduce >> patches? > > Someone who looked at the MRI source code told me that they use kernel > patches to batch mmap/mprotect requests, so it's just a performance > optimization (which could still be critically important). Whether > that applies to C4 as well, I don't know. > _______________________________________________ > bitc-dev mailing list > [email protected] > http://www.coyotos.org/mailman/listinfo/bitc-dev > _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
