[Beowulf] Julia Language

John Hearns hearnsj at googlemail.com
Tue Sep 19 13:13:28 PDT 2017


Justin,  to be honest I think Julia parallelism is in an early stage.
I think you would be better to talk directly with the language founders.
A good place for discussions is at:
https://discourse.julialang.org/

On 18 September 2017 at 18:22, Justin Y. Shi <shi at temple.edu> wrote:

> Hi John:
>
> Thanks for your quick reply. I have been working on fault tolerant
> computing for a long while. Given Julia's recent successes, I am curious
> about how it's designer had tackled the scaling challenge.
>
> In other words, we knew that end-to-end computing paradigms are
> fundamentally unstable under the lens of extreme scale computing. I am
> always interested in the extremes when the envelope MTBF (mean time between
> failures) can be pushed.
>
> Justin
>
> On Mon, Sep 18, 2017 at 4:22 AM, John Hearns via Beowulf <
> beowulf at beowulf.org> wrote:
>
>> Justin,  I do not know the answer to this question.
>> Woudl you kindly elaborate further - do you mean is it necessary to have
>> checkpointing with Julia,
>> or are you working on checkpointing software?
>>
>> Not really an answer to this question. Julia uses the LLVM compiler and
>> something called multiple dispatch.
>> At first sight as a scientist/engineer you will probably throw your hands
>> up in horror at the though of multiple dispatch,
>> and start wailign that it is wasteful of machine resources and CPU
>> cycles. But no, wait and think on... computers are
>> pretty powerful these days and disk space (to a first approcxilmation) is
>> plentiful. SO read on with an open mind...
>>
>> https://armchairecology.blog/2017/07/10/julia-in-ecology-why
>> -multiple-dispatch-is-good/
>> http://www.stochasticlifestyle.com/type-dispatch-design-
>> post-object-oriented-programming-julia/
>>
>> So for any function Julia creates separate OPTIMISED code paths for every
>> combination of types the function can operate on.
>> (ie those types which do not have operators which are able to work on
>> them are ruled out).
>> Read that again - instead of creating huge code blocks which work with
>> any type which you give it,
>> separate code blocks with optimised code are produced. (Th is is my
>> understanding of how it works).
>>
>> A consequence of this is that first time through, a Julia program is
>> slow. You are cautioned to remember this when benchmarking.
>> I have never shot a gun, but if I may bowrrow a term from target shooting
>> there is a pretty tight grouping around the C bullseye here
>> https://julialang.org/benchmarks/
>>
>>
>>
>> Before anyone throws their teddies away, it is of course perfectly
>> possible to produce compiled Julia code. My reading of this is that this is
>> not
>> a slick process at the moment, and there are efforts ongoing to make this
>> easy.
>> Though this looks pretty slick to me:
>> https://github.com/JuliaComputing/static-julia
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> On 18 September 2017 at 04:08, Justin Y. Shi <shi at temple.edu> wrote:
>>
>>> Hi John:
>>>
>>> Do you need to add checkpoints for Julia programs?
>>>
>>> Just curious.
>>>
>>> Thanks!
>>>
>>> Justin
>>>
>>> On Wed, Sep 13, 2017 at 7:43 AM, John Hearns via Beowulf <
>>> beowulf at beowulf.org> wrote:
>>>
>>>> I see HPCwire has an article on Julia.  I am a big fan of Julia, so
>>>> though it worth pointing out.
>>>> https://www.hpcwire.com/off-the-wire/julia-joins-petaflop-club/
>>>> Though the source of this seems old news - it is a presentation from
>>>> this year's JuliaCon
>>>>
>>>> JuliaCon 2018 will be talking place at UCL in London so mark your
>>>> diaries. Yours truly should be there.
>>>>
>>>>
>>>> _______________________________________________
>>>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin
>>>> Computing
>>>> To change your subscription (digest mode or unsubscribe) visit
>>>> http://www.beowulf.org/mailman/listinfo/beowulf
>>>>
>>>>
>>>
>>
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>> To change your subscription (digest mode or unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.beowulf.org/pipermail/beowulf/attachments/20170919/f24c5659/attachment.html>


More information about the Beowulf mailing list