On 03.04.2009 03:41, Corey Osgood wrote:
On Thu, Apr 2, 2009 at 8:33 PM, Carl-Daniel Hailfinger wrote:
Although the breakage itself is not something to be celebrated, we now have a list of boards using ROMCC by default. Now if we could convert them to CAR, we'd win big time.
As good as that would be, we know the ROMCC code works, and we don't know if we can find people with boards to test new ports for all of them. That, and CAR is so much easier to work with in v3, so let's concentrate our efforts there. My 2 cents, at least.
Absolutely agreed about ease of use of v3 CAR. My hope was to simplify porting to v3 by switching v2 targets to CAR. But v2 CAR is horribly complicated, so your point about concentrating efforts on v3 is very valid. Now if some targets can be converted easily to v2 CAR (less than 30 minutes per board), my idea still might make sense, if only to differentiate between hard/impossible and easy CAR conversions from the hardware POV.
And to be honest, I want to test how well Urbez' gcc+xmmstack solution works out for the hard targets.
Regards, Carl-Daniel