On 02/06/2014 10:15 AM, Jan de Mooij wrote:
On Thu, Feb 6, 2014 at 6:35 PM, Nicolas B. Pierron
<[email protected]> wrote:
As they are looking at making a working mobile device, I recommended that
they also add a MIPS Simulator (which would be added to v8 soon). The MIPS
Simulator will serve 2 purposes:
- Check the MIPS backend on Try.
- Make MIPS testing easier for other devs.
The problem I see with this is that MIPS is a tier-3 platform afaik
and we don't have real MIPS hardware to test on. So what do we do when
we break the MIPS simulator build?
I do not think we should care, as this is not a Tier-1 platform, but I think
that having this infrastructure is better to bisect what cause the
regression and thus what needs to be added to fix it.
What if we add a new feature that
requires MIPS assembler + simulator changes, something similar to the
float32 work that happened?
Then we should break it explicitly with a
MOZ_ASSERT_UNREACHABLE("Parity?");
Such as MIPS developer can come and search for obvious locations where the
parity is not respected.
We can do our best to apply
changes/refactorings etc also to the MIPS directory, but I don't think
we can keep MIPS working at all cost with no way to test it...
I agree, this is not a priority for us at the moment, and we should accept
to break it if we cannot improve it. But if nobody want to support it for a
while, then we should remove it.
--
Nicolas B. Pierron
_______________________________________________
dev-tech-js-engine-internals mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-tech-js-engine-internals
