Il 09/04/2013 18:06, Fabien Chouteau ha scritto: > As I already explained to this mailling list, we (AdaCore) use QEMU to run a > lot of short executable (~2 secs) on many different targets. In that > configuration the SMP restriction on Windows (QEMU runs only on 1 core) was > not > efficient on our 16 cores servers. > > Few months ago I asked on this mailing list if the restriction could be > removed. There was no clear answer because nobody was able to tell if it would > work. We decided to give it a try. > > During the last months we only found one problem, it was a dead lock between > the TCG and IO thread, this is fixed by the second patch. > > We also found that suspendThread() and resumeThread() were not synchronous on > SMP systems, so we added a busy loop to wait for effective suspension. > > Since then, we run thousands of QEMU on our SMP Windows servers each day > without any problem. And of course we greatly improve the execution time of > our > test-suites.
Awesome! Thanks guys. Reviewed-by: Paolo Bonzini <pbonz...@redhat.com> Paolo
