On Wed, 2014-05-28 at 20:29 -0300, Mauricio Faria de Oliveira wrote: [...] > > Also most architectures now use several types of NO_HZ, so the main > > frequency is not really important anymore. > > With that, I disagree. > > The application of NO_HZ is only idle CPUs (given it's description, > tickless /idle/, and the rename to NO_HZ_IDLE with the introduction of > NO_HZ_FULL): > > """ > CONFIG_NO_HZ_IDLE: [formerly CONFIG_NO_HZ=y], this turns off the > periodic tick when a CPU enters idle mode. [4] > """ > > The target scenario for the systems running ppc64el is under-load, > not that idle. There, the lower interrupt rate helps. The target scenario - for all systems with these processors? I thought the point of OpenPower was to allow multiple manufacturers to support a wider range of applications. I don't see how you can justify such a broad claim about how this architecture will be used. > Also, again there's good reason for the defconfigs to have HZ=100, > and that to be kept as is by many distros out there. Well, the defconfig for x86 has HZ=1000, but we don't follow that either. We picked something in the middle of the range of options and we are consistent across architectures. > I wouldn't know all the reasons, but that's been thought by talented > people, for this platform. So I'd ask you to keep HZ=100 on Debian too. > > Is that ok? > > I can look for more evidence/explanations if you'd like. I think this does need more explanation. Ben. -- Ben Hutchings Absolutum obsoletum. (If it works, it's out of date.) - Stafford Beer
Attachment:
signature.asc
Description: This is a digitally signed message part