On Tue, May 21, 2013 at 09:54:56PM +0200, Daniel Vetter wrote:
> Now this was broken in pretty fundamental ways:
> - M1/M2 have been consistently off by 2 and used doc values instead of
> the two less registers values our code expects.
> - M/N limits often were too small by seemingly arbitrary am
Now this was broken in pretty fundamental ways:
- M1/M2 have been consistently off by 2 and used doc values instead of
the two less registers values our code expects.
- M/N limits often were too small by seemingly arbitrary amounts. I
suspect this started to work around issues due to the wrong
