ARM Delivers 64-bit ARMv8 Linux Kernel Support (AArch64)

[phoronix]

Phoronix: ARM Delivers 64-bit ARMv8 Linux Kernel Support (AArch64)

ARM has today posted their set of patches that implements core Linux kernel support for AArch64, the ARM 64-bit architecture...

http://www.phoronix.com/vr.php?view=MTEzNDg

[plonoma]

There is no hardware platform available at this point. From a kernel perspective, the aim is to minimise (or even completely remove) the platform code from the architecture specific directory. FDT is currently mandated and there are ongoing discussions for ACPI support.

We need a uniform mechanism for hardware discovery.
Currently this is a mess on tablets and smartphones with arm as far as I know.

Having ACPI would make everything a lot easier for software development and shorten the time for hardware integrators and platform makers to release new designs.
We want ACPI, we want ACPI!

Having ACPI on mobile phones and tablets would be awesome from a hardware point of view.
Plus it makes things easier to do, much easier to do.

(Heard that Microsoft is already asking ACPI for things that run windows 8.)
Seems like it's best to add ACPI anyway.

[89c51]

What can someone expect performance wise from these chips once actual HW hits the market?? Ie. are they going to be comparable to an Atom or something better than that.

[ldesnogu]

What can someone expect performance wise from these chips once actual HW hits the market?? Ie. are they going to be comparable to an Atom or something better than that.

Cortex-A9 already is faster clock for clock than Atom, so 64-bit ARM machines will most probably be faster than current Atom. Next gen Atom is a different story of course.

As far as 64-bit ARM support goes, we now have gcc and the kernel. But binutils are MIA

[curaga]

Cortex-A9 already is faster clock for clock than Atom

Huh? Last benchmarks I read on here showed a _dual_ A9 matching an old _single core_ Atom, in multithreaded benches. That would put it at about half clock per clock, no?

[curaga]

http://www.phoronix.com/scan.php?pag...4_armfeb&num=1

[ldesnogu]

1. N270 has HT which benefits Atom a lot given its dumb in-order archictecture.
2. Some of the benchmarks in your link aren't MT. Pick Byte nbench. Don't forget the A9 is running at 1.2 GHz vs 1.6 GHz N270.
3. FFmpeg probably is using a unoptimized codec.

So I stand to my point: at the same clock, A9 is faster than Atom

[curaga]

7-zip is multithreaded, as are the NAS fortran tests and video encoding tests, and the A9 does worse there.

[ldesnogu]

7-zip is multithreaded, as are the NAS fortran tests and video encoding tests, and the A9 does worse there.

Did 7-zip ARM version receive as much attention as the x86 one? It should also be noted that it's very likely this is memory bandwidth constrained and many existing A9 SoC have poor memory bandwidth (and this is not a property of A9 core itself).

How old is ARM FORTRAN port?

As far as encoding goes, the answer is the same as FFmpeg: probably hitting some ARM unoptimized path.

[droidhacker]

http://www.phoronix.com/scan.php?pag...tem&px=MTEzMTk

How quickly people forget about current news.
"2011-11-29" is an N270.
"xoom linaro" is a qualcomm MSM8255
See how it gets thoroughly whooped by a 1-core armv7 @ 1GHz?