* Mauro Mozzarelli ([email protected]) wrote: > > Are your AMD processors actually of a type that is > > supported in > > OpenSolaris for frequency scaling? If they are > > family type 15 or lower, > > then they aren't and likely will never be because of > > lacking invariant > > TSC support. > > Of course, family 15 exactly. Do you think that if I had a business > case for a recent and more expensive server I would install > OpenSolaris on it? It is expected that until OpenSolaris will not be > ready for production it will be run by testers and early adopters on > reclaimed hardware.
I'm not really understanding what you're trying to say here. What I will say is that this isn't really a business case issue. It's a technology issue (keeping in mind that I don't work on kernel code or the power management code in OpenSolaris, I do however follow those technology lists). If those processor families had invariant TSC support (like modern AMD processors and those from Intel) then OpenSolaris would have supported them. They don't, so they aren't supported. Now, there have been posts over the years on [email protected] about this lack of support and the answer usually comes down to "it's very hard to do *right* without invariant TSC support if at all and likely not worth the large effort required". If you'd like to see this support, then step up and start digging in to the code :-) This is "Open"Solaris after all. The following thread sums things up pretty well: http://opensolaris.org/jive/thread.jspa?messageID=448711&tstart=0 In a perfect world with no lack of resources I think you would find that Sun would love to implement everything and scratch everyone's itch. Sadly, we don't live in a perfect world and tradeoffs must be made. OpenSolaris supports powernow on AMD processors from the K10 series (barcelona and higher). They were introduced in 2007. Family 15/Turion was introduced between 2003-2006, pretty old in terms of technology age. > If what you say is correct it is unlikely that it will ever achieve a > great deal of a market share. I completely disagree. For many reasons. It already *has* market share. Think of all the Solaris 10 deployments (from which OpenSolaris was initially started from). Not to mention the people (like Joyent) who are running OpenSolaris *today* in *production*. Or all the people using series 7000 boxes to manage their data, running on OpenSolaris code. Just to name a few. OpenSolaris does a good job of supporting *recent* hardware (say stuff produced in the last 3-5 years). It doesn't do so well on stuff older than that (although your mileage may vary) and likely never will because there just isn't much return on an investment to support 'older' hardware. Now, if there were more in the community who wanted to "step up" and add support for this older type of hardware then they would be welcomed. Anyway, as I said, it's not likely that OpenSolaris will get support for powernow on older processor families other than what is currently supported unless the community decides to step up and do the work or an easier solution can be found (which from everything I've read doesn't appear to be possible). I suppose Oracle could decide to invest in this work if/when they take control of Sun, but I don't see the logic in it for them. Expending resources to support 'old' hardware isn't exactly a no-op. My .02. Cheers, -- Glenn _______________________________________________ opensolaris-discuss mailing list [email protected]
