[CakeML-dev] bootstrap timeout

[Ramana Kumar]

Excellent. Yes that confirmation helps. (I ended up finishing to_lab_x64
interactively, because I want to produce a new .S file quickly.)

On 10 April 2017 at 23:14, Yong Kiam <tanyongkiam at gmail.com> wrote:

> FWIW, it went past apply colour (par) for me non-interactively. (now at
> enc_sec (par))
>
> On Mon, Apr 10, 2017 at 2:53 AM, Ramana Kumar <Ramana.Kumar at cl.cam.ac.uk>
> wrote:
>
>> Running it interactively succeeded. So now I'm more confused about why or
>> where it was getting stuck. I'll try another non-interactive attempt.
>>
>> On 10 April 2017 at 11:20, Ramana Kumar <Ramana.Kumar at cl.cam.ac.uk>
>> wrote:
>>
>>> Hi all,
>>>
>>> I think the bootstrap is currently (cd43c3944) getting stuck in the
>>> "apply colour (par)" phase.
>>>
>>> I've run it for several hours and it doesn't finish (
>>> https://cakeml.org/bootstrap-timing.txt suggests it used to take <30
>>> mins).
>>>
>>> I also tried running it on some smaller sections of the whole program,
>>> e.g., the first 10 progs, the second 100, etc. and in those cases it does
>>> complete successfully and doesn't seem to leave anything unevaluated (I
>>> looked for vars and abstractions in the results). So it's not obvious
>>> what's going wrong...
>>>
>>> My guess is there's something unevaluated somewhere in the declarations
>>> I didn't look at explicitly yet. Currently I'm evaluating
>>> to_lab_x64Bootstrap on the whole program interactively to try to find it.
>>>
>>> If you know of anything that might have changed recently (within the
>>> last 2 weeks) that would affect the bootstrap evaluation of
>>> checking/applying the register allocator, let me know. Given that a
>>> "successful" regression test usually times out, this may have been
>>> happening for a while. I know the bootstrap was successfully completed at
>>> 32b749 (March 22), though, because I have a snapshot of the result for that
>>> commit.
>>>
>>> Cheers,
>>> Ramana
>>>
>>
>>
>> _______________________________________________
>> Developers mailing list
>> Developers at cakeml.org
>> https://lists.cakeml.org/listinfo/developers
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cakeml.org/pipermail/developers/attachments/20170410/6ded83ed/attachment.html>