Paul Mackerras the powerpc maintainer for the linux kernel A lot of these notes are on Paulus' slides why merge ppc and ppc64 -- two seperate arch directories -- reduce maintenence effort - lots of similar code - signals code was similar on both -- 64-bit embedded systems coming - all infrastructure for both (drivers etc) canbe used xmon debuggger used to not handle smp systems well, paulus madde it work with 64bit but it never got back into 32 bit smp systems, just an eample of lack of sync between trees. The new ARCH=powerpc supports all 64 bit systems, and 32bit supports powermac and CHRP machines, powermac no longer supported in ARCH=ppc some of the 32 bit embedded systems are still under development for the new ARCH= The merge happened fast, in January last year Paulus (maintainer) said that is a silly idea why would you want to do that, at OLS in July they all talked it out and said lets do it. By 10 January 2006 it was pretty much all done A side effect of this change, they decided to have all data about the platform/chip/system avaailable in a cmmon format, such as mac addresses, memory size, etc. They insist that all this be provided to them wth a single format for their device tree. Dave Gibson wrote a tool dtc (device tree compiler) that when given an easily understood text form of something it will have the blob created that the kernel will recognise and have all the available information there. Need not list things the kernel can discover This simplifies embedded devices well. Making the blob as small as possible and as easy as possible to provide/implement On to the mm changes, 9,7,7,9 4 level page lookup 64 kb pages option can be less efficient, buggy applications may break