On 08.07.2008 17:44, Jordan Crouse wrote:
On 08/07/08 08:21 -0700, ron minnich wrote:
I would be willing to bet that a lot of that patch activity was me, and I've been pulled away to other things. We need more people involved. It is true that the path of least resistance is v2 -- it's here and it works. But we need people to take the hard path and get on v3. I can tell you, that once you get going on it, you won't want to go back. v3 is really nice.
I guess v2 works too well :-)
Here's the problem. Based on the number of email and IRC requests, it is clear that RS690/SB600 code is in great demand. And right behind that we hear clamoring for RS780/SB700 and Barcelona B3. We currently have two choices - we can go for V3, which would give us big style points but our fans will probably be disappointed while we spend the time to iron out the wrinkles. Or we can go for V2 and run the risk of continuing to prop up the legacy code. Our hope is that eventually v3 will catch up with us, and we can seamlessly switch to v3 for some future chipset or processor.
Unfortunately, we don't have the resources to do both, and you will forgive us if we go the path of least resistance for our customers. I think that most every other commercial vendor in this project faces the same dilemma.
And I am immensely thankful that AMD is creating support for new chipsets. More supported chipsets mean more users which means more developers (to some extent). This indirectly benefits v3, so a big "thank you" from me.
The solution is to recruit a new generation of developers who are willing to take the time and effort of moving our primary processors and chipsets to v3 and stabilizing it to the point where the customers can reliably move. This has already happened for the most part for the Geode LX with positive results. Once we solve the few remaining warts, then there would be no reason to use V2 for LX.
Fully agreed.
Speaking strictly for myself, I do think that v3 is the way to go, and we look forward to when we can recommend it to our customers. But we just can't justify the extra time and effort at this point, and judging from the emails asking for SB600 code, neither can you.
I'm happy with the priorities you outline and hope we'll reach the "customer recommendable" milestone soon.
PS: And I apologize personally for doing more to damage v3 and LAR then any other three developers - I mean well, I promise! :)
Stirring things up is a great way to get people thinking about the grand scheme of things. You invested a lot of time analyzing LAR and the SELF proposal was a great way to advance our understanding of it. Although no commits resulted from it, I see your actions as very beneficial, not damaging. I hope to prepare a proposal incorporating your SELF design together with a slightly modified LAR design in the next few weeks.
Regards, Carl-Daniel