ARMv4 vs ARMv5

thomasg thomas at gstaedtner.net
Thu Oct 16 01:40:51 CEST 2008


On Tue, Oct 14, 2008 at 12:58 PM, Graeme Gregory <graeme at openmoko.org>wrote:

> On Tue, 2008-10-14 at 11:52 +0100, Andy Green wrote:
> > Hi -
> >
> > At some point we will be working with s3c6410, which is ARMv5 capable as
> > well as existing ARMv4 capable.
> >
> > How is the distro side planning to come at that... are we just going to
> > treat it as ARMv4 so we can share packages on all the platforms, or is
> > there enough advantage in ARMv5 that we will basically build packages
> twice?
> >
> > What's the situation then for people who have an ARMv5 based distro on
> > their device for installing ARMv4 packages (which they can run) too?
> >
> I thought s3c6410 was armv6 with floating point?


According to the available datasheet[1] the core of the s3c6410 is a
ARM1176JZF-S which is armv6 [2].


> In 99% of cases users wont see much benefit from armv5 vs armv4
> packages. Some special cases like mplayer where the new instructions are
> used show real benefits.
>
> By default OE will build both armv4 and armv5 packages if you have two
> different machines. I'm not sure what we shall do yet, needs some
> discussion internally I think.
>

If using the newer instructions brings even the slightest acceleration it
should be used imho.
Also, the s3c6410 has an FPU (ARM VFP), so I don't think it makes any sense
to use the softfloat-armv4-packages.
No need to sacrifice performance to simplicity.

1: http://semicon.samsung.de/applicationprocessors/documents/6410_PTB.pdf
2: http://www.arm.com/products/CPUs/ARM1176.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.openmoko.org/pipermail/devel/attachments/20081016/dce305d5/attachment.htm 


More information about the devel mailing list