Why don't we trade Quebec for Alaska? Heheheh...

Ok, back to the ARMv6 vs v7a issue. I'm not sure what all these press releases about 'porting Ubuntu to ARMv7' is really about. I don't know a single Ubuntu (or Debian or any other distro) application which relies much on assembly. They're all written in C, or C++, or occasionally some other high level language.

So, what all of those applications rely on in order to get good performance out of a particular CPU is just this single question: "Did somebody add support for ARM v7a optimizations to GCC already?". If yes, then you can choose to optimize for your architecture of choice by using a particular compiler flag. And potentially make the executable incompatible with older (e.g. ARM v6) architectures.

The real porting to a new architecture goes on in A) the compiler suite, as mentioned above, and B) in the kernel. Not in the distro. Ubuntu is a distro. The kernel porting is done by all those folks (including Nokia, and other companies) contributing to the kernel in general, and linux-omap in particular (for our omap-based ARM devices).

So what does it mean when they say "porting Ubuntu to ARM"? It doesn't make sense. "Ubuntu contributing to ARM Linux kernel development", or "Ubuntu supporting work on the GNU Compiler Collection" would make some sense. "porting Ubunto to ARMv7" does not.

I think it mostly has to do with people thinking compiling something for ARM automatically makes it a mobile distro. See Deblet—it doesn't.

The conclusions being draw here are really somewhat outlandish for a new CPU target announcement. . . . :\

Well, there's the issue that a lot of applications (e.g. iceweasel) used to be broken on ARM, presumably due to alignment and/or type size issues (but IDK). Fixing that is legitimate app-level porting. Of course, now that Debian has them working for ARMv4 (I think?), tweak compile settings and everything's likely to work straight off on ARMv6. (Don't get me wrong, even just building everything optimized for a higher CPU, with no contributions to gcc, and putting it all in a repository so you can rootstrap up to a full, more-optimized, system is certainly good work, but it does seem a little over-hyped if that's all that's going on...)

Just noticed http://blog.canonical.com/?p=74 - "For now, builds will expect a minimum of an ARMv5 instruction set". Interesting.

Just to clarify especulations, this is not a covered approach of Ubuntu to Maemo. ARM and Ubuntu have probably good reasons to sign such agreement without needing to put Maemo in the equation. Of course the fact that Ubuntu goes ARM puts it closer to Maemo but that's it.

Note also that the announcement actually talks about "Full Ubuntu Desktop Experience" while "Ubuntu Mobile" is not mentioned. Probably for a reason?

As a note aside, personally I'm looking forward to the outcome of this announcement also for some reasons that hopefully will make our life easier: the ARM stakeholders getting more and more sensitive about open source and the Ubuntu fans learning more and more about why direct comparisons between ARM/mobile and x86/PC are not always applicable.

Arm targeting the netbook market probably:
http://www.engadget.com/2008/11/13/a...ook-mid-chips/

Edit: but surely you already knew that

I have to say I'm still more excited about a handheld device using ARM Cortex A8 than a netbook one. Although a netbook with "all-day battery life" will revolutionize the laptop industry as much as, if not more than, the original EeePC did.

EDIT: I'm also glad to see a big, serious effort to port the Ubuntu repositories to ARM, not the small scale, experimental effort of the Mojo team. I'm also really glad to see that they're targeting ARMv5+, which means we're in on the fun!

As I noted in the Mojo Ubuntu thread, once you step away from the hardware support and user experience of the x86 desktop / laptop, and look at the variety of software in the repositories, Ubuntu doesn't have a lot to differentiate itself from Debian Sid at the moment. I suspect they're going to focus on getting this ARM port to be as smooth and simple to install as it is on the x86 desktop, with lots pretty install wizards and good hardware support. That's been their strategy on the desktop, and it's worked to make them really stand out.

I'm also looking forward to seeing the results. The process interests me just as much as the results. Will they make a flexible system that tries to abstract away the details of a whole range of ARM systems? Will they target popular released systems with custom releases? Or has ARM contracted them to support specific unreleased machines? Any way you look at it, 2009 should be an interesting year for mobile Linux.

-John

A link to my blogpost to clarify things a little bit: http://idlethread.blogspot.com/2008/...on-arm_14.html

I think this was not posted yet:
http://www.ubuntu.com/news/arm-linux
They are talking a lot about targeting to armv7