Not so hARMless: ARM software development gets a major boost

Not so hARMless: ARM software development gets a major boost

Summary: Facebook, HP, Samsung and others are looking to jump-start development of software on ARM processors and in doing so are embarking on a risky, high-stakes road.

TOPICS: Cloud, Apps, Processors, ARM

What do Facebook, HP, AMD and Samsung have in common?

All of them are keen on using low-power processors to cut power consumption in chips, datacentres and mobile devices, and have helped found the Linaro Enterprise Group to invest resources in developing software for the energy-thrifty RISC (Reduced Instruction Set) chip architecture.

The goal of the group, whose formation was announced on Thursday at the ARM TechCon summit in Santa Clara, is to accelerate the rate of software development on ARM's RISC chips.

Though ARM's chips sit inside most of the world's phones, one of the major criticisms levelled (frequently by rival Intel) against their RISC chip architecture is that it lacks support for legacy applications for servers, desktops and netbooks. This is borne out in the real world, as Microsoft's Surface RT tablet ships with little legacy software due to the restraints imposed by the ARM platform.

Once ARM chips have a decent datacentre software ecosystem, it becomes a lot easier for companies to cut their electricity bills by transitioning from power-hungry Intel and AMD x86 chips to power-thrifty ARM processors.

The Linaro Enterprise Group's members represent a broad swathe of the technology industry, including chips (Applied Micro, ARM, Calxeda, Samsung), operating systems (Canonical, Red Hat), and datacentre gear (HP, IBM). 

"The significance of key industry players coming together like this to develop new aspects of the ecosystem is showing the transformation position the industry is now in," ARM's chief executive Warren East said in a statement. 

How much of a change?

However, though the industry could be in the midst of a 'transformation' - that is, fewer companies showing total allegiance to Intel - the question of how far it will transform, and into what, remains open.

For change to happen, the software needs to be developed. So far, this has been difficult

For change to happen, the software needs to be developed. So far, this has been difficult. RISC chip company Tilera launched a range of impressive 100-core processors in 2011 to take on Intel in the high-end datacentres, but judging by recent announcements this approach failed to see enough take-up and the company has now beat a retreat back down to embedded networking with a nine-core chip.

Similarly, RISC company Adapteva recently had to take to crowd-funding site Kickstarter for its multicore Parallela processor to try and get the money to develop a platform that would drum up developers for RISC chips.

Where Linaro may differ is in its trans-industry partner approach. This model has already worked for the open-source cloud platform OpenStack, which developed from the same type of motivation (a bunch of second-tier companies that wanted their own cloud, but didn't have the money to build an equivalent to Amazon or Microsoft on their own). 

Whether this approach works or not will become clear in the next year. For now, the enthusiasm is there - Facebook is testing ARM chips in its datacentre and AMD plans to license and make its own designs - but the evidence for future success is scant. The first server-grade 64-bit chips are due to come out next year and, at that point, we will find out if the Linaro Enterprise Group has what it takes to strongARM the industry into adopting them.

(It must be said that as Intel brings in chips made to new process technologies and ARM increases the server capabilities of its chips, the power consumption of both architectures will converge, but we are a few years yet away from that happening.)

Topics: Cloud, Apps, Processors, ARM

Jack Clark

About Jack Clark

Currently a reporter for ZDNet UK, I previously worked as a technology researcher and reporter for a London-based news agency.

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • I dont think we're a few years away from intel getting to low power.

    In fact I think intel will get there on the hw side before these clowns will get there on the software side. And its a lot easier to move that datacenter software to low power intel chips than to low power arm chips. Intels only a couple months away from haswell, and a year from another shrink. And theyre watching the data center ramifications very closely
    Johnny Vegas
    • "Clowns"?

      That single word totally destroyed your whole post, and your reputation (assuming you had one of course). If you want to see a clown, look in the mirror.

      Only insecure and petty minded shills/fan boys feel the need to disparage such a respectable group of companies.

      You should be ashamed of yourself, but I doubt you have the mental capacity to even go there.
    • Haswell helps but on the power side it may not be enough.

      Once we see A50 implementations (64 bit server line) we will get a good feel. I am guessing they will consume more power than the A15 but if implemented on a smaller process perhaps not a lot and perhaps none at all.

      Intel's Haswell will still be 5-10x more power hungry.

      Granted it will be faster, but many of the work loads these servers process are highly threaded, many short lifespan light-weight processes.

      I agree it will be a couple of years and as ARM adds capability the power consumption will go up and meanwhile Intel's power consumption will go down. Whether they intersect in 2 years - I doubt it.

      ARM will have some things going for it. Cheaper, low-power, fast enough for highly parallel lightweight loads.
  • Why not fund a toolchain for ARM?

    It seems there are no good free software development platforms for ARM. Why not develop that? Too obvious?
    • Re: Why not fund a toolchain for ARM?

      We already have the Android SDK and NDK, with which we can write code in native C or C++ with our choice of IDE (I use Emacs) and automate builds of arbitrary complexity with Ant.

      If you want to go to a lower level, you can directly use the same GCC cross compilers that the NDK is built on. Or there's LLVM/clang, if you prefer.

      Remember also that these tools are not just ARM-specific. There are some really cheap MIPS-based devices coming out of China. By working at the Android level, rather than the ARM machine level, you can develop code that runs on both architectures.
    • RE: Why not fund a toolchain for ARM?

      Here ya go (all open source):

      gcc 4.6:
      gcc Aarch64:

      Tons more on: