Hey, On 12/20/2013 12:29 PM, Victor Martinez wrote:
Hello, I'd like to start a thread talking about the new support for aarch64.
+1 Well, first I reorganized your questions in the order that I think more appropriate.
Gcc seems to give support to it [1] and this could be very interesting for the next year, when probably we'll start seeing first devices with aarch64 support (armv8) with backwards compatibility and not relay only in ARM emulator (Foundationv8).
Generally I like the idea and also I think we will soon see someone with an ARMv8 device wondering if CRUX-ARM can run on it.
Could be interesting to start a new git repository (toolchain-aarch64) to start playing with the new armv8 arch?
Sure, anticipate to this sounds good to me.
This could be the next step in CRUX-ARM where we can start developing a newer toolchain with 64bit support for this architecture. It can be researched to add this support in the next 3.1 release (or at least to develop until we can put hands on devices with this support) Any thoughts about this?
I agree with start playing for aarch64 but I honestly do not think we should include it in our future plans for 3.1 for now. As you say, that should happen when we have enough ARMv8 devices. On the other hand, I not want to be negative but realistic. We are a very small team and we have to concentrate our efforts on something concrete but very well done and not try to cover many things we could leave incomplete or failing to maintain properly. Besides, ,I think that start developing for toolchain-aarch64 means that we will again have 2 main releases for our product. In 2.7 we had the softfp main branch which was widely supported by all devices, and introduced a development hardfp branch. In 2.8 we keep de same branches: softp as the main one and hardfp as development branch. In 3.0 we switch our mainline to hardfp which becomes stable and decided to leave the softfp branch. So far so good, but the incorporation of aarch64 means more when you consider the steps taken in CRUX to move from i686 to x86_64, or what is the same, from x32 to x64. You thought we would have to take that product? Purelib or multilib? This things will decide our for first steps fro toolchain-aarch64. I would bet for a version that had its purelib overlays, but also think about the state of the software that can run aarch64. I'm not talking about the kernel or gcc toolchain and all, but for example firefox, chromium, etc. A repo compat32 will also be needed? I think we need to plan this well before anything else. Since the difficulty is not to create a Makefile for toolchain-aarch64 and see that is functional, but is usable, and we can keep all that good work. Regards, Victor.
xmas!! and best regards, -- Jose V Beneyto | http://sepen.it.cx/