Re: ARM EABI port: minimum CPU choice
+++ Martin Guy [06-05-12 18:49 +0200]:
> Hi again
> What feelings do people have about the choice we have to make about
> the minimum ARM CPU that will be able to run the forhcoming Debain ARM
> EABI arch?
> As I understand it, the choice is between ARMv4t and ARMv4 (such as
> If we target ARMv4t, we save 2 instructions in every function return.
> Is a slightly-universally-slower arch that includes StrongARM CPUs
> preferable to a slightly faster one that leaves them running the
> current Debian arm arch?
Debian emphasises supporting as much stuff as possible, over maximum speed.
I think we should support armv4 if at all possible, and it seems that it is
possible with the minor cost of a small slowdown (it would be useful to
measure if it is significant at some point), and carrying a gcc patch.
Strongarm support is important because currently a lot of debian buildd
machines use it, and there is a lot of strongarm hardware out there that
people might want to run Debian or debian-derived stuff on.
My understanding of the workaround is that it will not introduce any
incompatibility with binaries which do not include this bodge, so it can be
easily dropped at some point in the future when strongarm support is deemed
no longer necessary. Somone does need to actually patch GCC to do this and
test that it works. Anyone planning to do that?
I talked to Anthony Towns about this (at Debconf) and he agred with the
above analysis, and also that a new arch was appropriate (he had expressed
some reservations in the past).
Aleph One Ltd, Bottisham, CAMBRIDGE, CB5 9BA, UK Tel +44 (0) 1223 811679
work: http://www.aleph1.co.uk/ play: http://www.chaos.org.uk/~wookey/