On Wed, 8 Jul 2020 11:32:09 -0500 Scott Cheloha <[email protected]> wrote:
> On Wed, Jul 08, 2020 at 09:36:03AM -0600, Theo de Raadt wrote: > > Ugly test program (not showing it) seems to show syncronized clocks on > > powerpc, or at least closely sycronized. It might be one register. > > I guess I'll share mine, then. > > With this program I can consistently catch an offset of -2 to -1 on my > G5. Such a small offset is probably fine, right? I'm curious if it > gets any larger on other machines. /sys/arch/macppc/macppc/cpu.c has code to "Sync timebase" of the cpus. One cpu does h->tb = ppc_mftb() + 100000; other does ppc_mttb(h->tb). Your -2 to -1 implies that time went backwards! I don't like this, but time also goes backwards in OpenBSD/amd64 and in other systems. (Microsoft Docs, "Acquiring high-resolution time stamps", says that stamps "from different threads", within a small range of uncertainty, have "ambiguous ordering".)
