Excited about arm port

I have noticed that the haiku arm port has been getting a lot of attention lately. Let me just say that I am really excited about this. Anyhow, thanks ithamar, for working so very diligently on this. I may just have to pick up a beagleboard-xm if things continue :stuck_out_tongue: .

The ARM processor vary alot … yes there are 2.0Ghz processors in the wall wart computers for example but they have rather high Cycles Per Instruction and do not execute out of order. The 1Ghz omap4430 is probably much faster as it is multi-core,out of order and superscalar (multi instrucion issue per cycle). Remeber Ghz means nothing ask for standardized benchmarks like specrate etc… or linpack those are still rather synthetic but still actually mean something wheareas Ghz, Mips and Flops mean nothing.

Actually windows programs are portable as well…its would just be a matter of MS adding support for building binaries for X architecure such as Alpha and MIPS back in the day.

pandaboard.org << check it out

Would be nice to test with a 100 core cpu: http://tilera.com/

Or even a Mac Pro with only 12 cores!

But seriously, I have a little project I am working on that will use as many cores as haiku supports (currently set at 16). If anyone has a PC with 8+ cores, please let me know.

Tilera is MIPS and uses a modified Open64 toolchain IIRC…

Yes, The ARM port moves quite fast, but I think you still have a little time to gather the necessary funds…

So what is the expected “time until boot”, Ithamar?

I’m very interested in ARM CPUs, although I haven’t used any for computing. I feel that just having the x86 architecture is a little boring, and I have seen devs talk about how it’s complicated, irregular and so on. It’d be great if we could one day get good laptops and desktop computers using ARM CPUs.

They are bragging about how they have 2+Ghz dual core CPUs ready, but when we will actually see a computer with this still seems to be a mystery. When any of that hardware becomes available to users though, being ready with a Haiku port would be a really smart move. Neither Linux, nor Windows have anything major to offer (Only cellphone OS’s and whatnot).

Haiku also has an advantage in that pretty much the only applications still being developed are open source, meaning that they can be ported to ARM. Windows (And to some extent, Linux) would have to do something about all the old proprietary x86-only software that many users rely on which will not work on a foreign architecture.

Well, you seem to know a lot more about CPUs than me. I just want a really powerful one. Is there any simple way to compare ARM CPUs to my current CPU (Intel Core 2 Duo E7300)?

That’s absolutely true of course, but what I’m talking about is that most apps are not open-source, and thus you need the authors to port them. They might not even be interested in other platforms, and so oftentimes you need to abandon a lot of apps when moving on to other OS’s or CPU architectures. Open-source apps are not that big of an issue, just those that use assembly are tricky.

Looks nice, but it’s not a computer is it?


Look for benchmarks. I will post a few links to get you started.

The Cortex A9 CPU should give similar performance to Atom. Forget about comparing to Core2 Duo as these are more powerful.

ARM Cortex A9 pluses:
-very small size chip
-very low power consumption
-equal integer performance to Atom
-onboard HD decoder? (lacking on Atom boards which require add-in graphics card)

-fairly weaker FPU (will affect gaming, javascript, encoding, etc.).
-higher cost to Atom system?
-no Windows 7/XP, Haiku (stuck with Linux, Windows CE or Android)
-stuck with non-Windows games and software

ARM are better for tablets and smartbooks/netbooks to maximize battery use.

CoreMark is one benchmark used:




I don’t have an EfikaMX but if Haiku ran on it, I’d definitely get the smartbook.

Well, it boots already (using the verdex PXA270 target in qemu), it just doesn’t boot to completion :wink:

Since I’m doing this work in my spare time (which I accidentally had a little more of recently), I’m not giving any timelines whatsover out.

The question is, on what hardware? Haiku, Inc might be providing some real hardware for me, and when that arrives, that’ll be my only hardware target to focus on.

As is said in the comments here already, the different ARM-based CPUs are very different, there’s much more ‘driver’ style work to do per CPU then we’re used to in x86 land. Partly because these CPUs are actually SoCs (System-On-Chips), containing both peripherals as well as the CPU core, plus that the ARM CPU architecture does not handle things like interrupt handling really, the interrupt controller is just another peripheral where SoC manufacturers can pick and choose the solution they want to use…

Anyway, it very much looks like the TI OMAP range will get the most focus in the near future, once I get the basics working under QEMU using the verdex target (which makes development just so much easier at the initial stage then real hardware)…

Hey Ithamar, can you tell me are you looking at the TI OMAP3 range or OMAP4 range?
Or do you only have to focus on the TI OMAP rage as a whole?

Perhaps the BeagleBoard (OMAP3530, ARM Cortex-A8) or the updated Pandaboard (OMAP4430, Dual core Cortex-A9) would make good targets for development? Both are opensource and have attracted many opensource projects.



The question is, on what hardware? [/quote]Well, in terms of hardware that’s in reach of the consumer there are the Handheld PC’s and Pocket PC’s in the early 2000’s that use ARM processors (but natively run Windows CE 3 or 4). Their clock speed is a bit short but it would make a good starting point, give life to these devices.