Tuesday, 16 December 2014

Re: [RFC] CONFIG_ARM64_64K_PAGES=y for vivid?

On Mon, Dec 15, 2014 at 7:08 PM, Adam Conrad <adconrad@ubuntu.com> wrote:
> On Mon, Dec 15, 2014 at 05:36:25PM -0700, Dann Frazier wrote:
>>
>> There's the question of whether or not we would be penalizing the
>> performance of other classes of workloads people want to run on arm64.
>> If there are some representative tests we should be looking at, please
>> let me know.
>
> So, this is the obvious concern, and it only makes logical sense that
> many workloads will perform better with a smaller page size, as if you
> don't actually need the giant loads, you'll be flushing fewer caches
> for nothing. But, I'll put my money where my mouth is and see if I can
> find benchmarks to back up that hand-waving.
>
>> I'm not sure how much interest there is for ARMv7 compat in Ubuntu.
>> There is the developer use case of using the same hardware for armhf
>> and arm64 porting - which, theoretically, we could also achieve by
>> rebuilding those ports w/ an updated binutils. But otherwise, I'm not
>> aware of many legacy ARMv7 apps that users are likely to want to bring
>> over to Ubuntu/arm64 that couldn't be rebuilt.
>
> So, retroactively rebuilding all our already released dists is clearly
> not an option, and I hadn't even thought about this armv7 compat issue.
>
> This does impact our previous master plan for scrapping our armv7 build
> infrastructure and building armhf on arm64 in chroots, like we do for
> all our other 64/32 port pairs.
>
> Now, if qemu-system-arm --enable-kvm happens to work on arm64 (with 64k
> pages), this might be a non-issue, as we can run fully accelerated v7
> VMs on our v8 kit, and that would serve just as well as 32-on-64 with
> chroots. Is that possible? Does it work today? If not, is there any
> one working on making it work?

There appears to be work to address 32-on-64 KVM upstream, but it does
not work today. While testing this I discovered a (very likely
platform-specific) issue on my test system that causes kvm to fail to
init on 64K hosts. I'll investigate that in the background.

--
ubuntu-devel mailing list
ubuntu-devel@lists.ubuntu.com
Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel