ARM chippies cooperate on Linux Does the Linux operating system need yet another distro? No. But a bunch of people interested in the ARM RISC processors used in mobile computers and netbooks — and hopefully someday soon inside of servers just to scare the hell out of Intel — are ganging up to create a unified foundation for ARM-based distros called Linaro. …
As a heavy user of Linux on ARM embedded processors, I welcome a distro project focused on that (huge) family of processors. At the moment, Linux on ARM is fragmented among several different development communities, depending upon architectural details (e.g. MMU vs MPU, internal vs external RAM, execute from Flash vs RAM, etc.) and the target market (e.g. multimedia, network device, smartphone, automotive, deeply embedded, etc.) for the specific variant. Hopefully this will bring some coherence to what has become the second most popular Linux target.
Actually, that's exactly what they are doing...
Got to admire El Reg, probably just read the headline of the press release and made up the rest of the article from that. Actually if you look even very briefly at the linaro web site, the FAQs make it quite plain this is not just another Linux distro:
"Is it a distribution?
No. It is a common software foundation and set of tools for other distributions to use."
From linaro.org:
"Linaro was set up as a not for profit company to focus and align work from a variety of companies, organizations and the open source community with the aim to output a validated kernel, low level software and tools release every six months, available through member silicon suppliers."
So it seems to be about producing optimised and validated Linux kernels and the associated tools.
A full distribution, such as MeeGo on ARM, can benefit from this work as can any other ARM based distribution.
Sounds like a good plan to concentrate the low level tools and ARM kernel development among those who actually produce the silicon, leaving the creators of the end-user distributions (Google/Android, Nokia/MeeGo, HP/WebOS) free to get on with doing what they do best.
This has got to be more efficient use of resources than each of the silicon manufacturers plus ARM spreading themselves thinly across so many ARM Linux distributions when all they care about is the low level stuff and none of the eye candy.
You could indeed use Debian ARM, but it is built targeting ARMv4 and above. The current revision of the ARM architecture is ARMv7. This means it misses out on all of the potential optimisations for modern hardware (NEON SIMD, VFP etc) in order to maintain backwards compatibility. This results in crap performance compared to what is possible with a distro targeting a more modern ARM chip such as OE/Angstrom/Maemo/Android/Meego/Ubuntu etc.
The article is more than a little misleading.
Linaro is NOT another distro, but rather a common software foundation and set of tools for other distributions to use. It is driven by silicon suppliers (the folks who make ARM-core chips), and provides a way for them to test their board support packages and make them available to device manufacturers, distributions and software developers.
All of the application software will be developed in the appropriate upstream open source project, and written for the target distro (Debian, Ubuntu, MeeGo, LiMo) that lands on top of Linaro.
There are companies like MontaVista who take a "free, open source" Linux and commercialise it and then make it as hard as possible to get hold of the result without paying them oodles of money for source or binaries (seemingly somewhat in contravention of the spirit if not the letter of the GPL). If this means these companies have a slightly harder time because others are more "open" and easier to do business with, the results will actually be beneficial to the ARM ecosystem.
You guys need to do some homework. Linaro is NOT a distribution. From the linaro.org FAQ:
Q2. Is it a distribution?
A2. No. It is a common software foundation and set of tools for other distributions to use.
So there you have it, there isn't ANOTHER Linux distribution in the market, but rather a project to further the goal of Linux distributions (like Ubuntu, MeeGo, etc) on ARM devices.
Lets get this right.
"The Isle of Man is an internally self-governing dependency of the British Crown and its people are British citizens. The Crown has ultimate responsibility for the good government of the Island and acts on the advice of Ministers of the UK Government, in their capacity as Privy Councillors. Her Majesty the Queen as Lord of Mann is Head of State" - source " Isle of Man Government website.
Re: "(Just a question: Why not call it Linarmux? Or Linarmour?)"
Because ARM is a registered trade mark of ARM Ltd...
"Prohibited Uses of ARM Trademarks
1. Do Not Use ARM Trademarks as the Name of Your Company, Products, or Services, or As A Domain Name. You may not use or register in any jurisdiction any ARM Trademark as all or part of your company, product, or service name without ARM's prior written permission. For example, you shall not use or register ARM, ARMADILLO or DISARM. Similarly, you may not use, or register, in any jurisdiction a domain name that incorporates any ARM Trademark. "
http://www.arm.com/about/trademark-usage-guidelines.php
Come on, Reg... engage brain...
The OpenEmbedded Linux distro (http://www.openembedded.org), which is Debian-based, IIRC, provides excellent ARM support, and its Angstrom offshoot (http://www.angstrom-distribution.org/) has specific support for TI's OMAP processor (most notably the insanely cool TI-inspired BeagleBoard single-board computer), as well as the über-miniature Gumstix SBCs.
There's a massive amount of software available in these embedded distros, and they're extensively used in embedded applications. One hopes that common sense and support for existing development ecosystems doesn't fall victim to the not-invented-here syndrome. Concerted support from major industry players for a particular set of tools could really help things along.
"The idea is to have Linaro work on the low-level stuff that makes the best use of ARM processor features to squeeze performance out of the Linux kernel and let Linux distros and vendors who make their own distros for their own devices do the higher level stuff where they can differentiate."
all well and good but Will they finally get around to actually write code that optimises all the ARM Kernel and support general purpose library's For SIMD NEON use , to remind you, IBM and all the other PPC/Altivec based Linux core dev's etc didn't do anything Altivec worth mentioning so masses of potential generic PPC Linux speed was lost for no good reason, especially compared to the PPC mac-OS fully Altivec SIMD optimised library's, IS the ARM A8/9 128bit NEON SIMD also going to be wasted at this key lower Linux level too like Altivec SIMD before it ?
see the proof of concept code and results and now apparently near abandoned project http://www.freevec.org/content/libfreevec_ng
"The idea behind libfreevec is not restricted to AltiVec anyway. I have proven that glibc, the #1 libc used on Linux, is totally unoptimized even for common platforms (such as x86 and x86_64), and there are performance gains that could/should materialize if someone took the effort to do it."
"the IEEE754 math functions used in glibc/libm. I've often mentioned that these are slow as molasses on ALL platforms, and now I can prove it, here are some results:"
in case its not clear even unoptimised apps can then take full speed advantage of any SIMD optimised key Arm NEON library's and move the bottle neck somewhere easier and higher up to fix later....
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
