On 13 July 2015 at 07:48, Paolo Bonzini <pbonz...@redhat.com> wrote: > > > On 13/07/2015 00:02, Peter Maydell wrote: >> (I'm wondering if we somehow manage to spend all our time >> trying to service the GUI and no time making forward >> progress in the guest, though I don't have a clear idea >> in mind of why this would be so > > Perhaps because Xlib calls (and hence GTK+ calls) are blocking? > >> or why this change would >> make things worse, except that presumably we now do more >> full-screen-invalidates.) > > Yes, then it makes sense for this patch to make a difference. VNC does > dirty rectangles and has multithreading, but GTK+ doesn't do either.
Yes, that would make sense. But why do we need to do more full invalidates now? > You can just add a memory_region_set_log call for the integratorcp and > versatilepb RAM regions. It seems a bit of a shame that we basically broke previously working configs for all our boards with embedded framebuffer in RAM :-( -- PMM
