[Vm-dev] Re: [Pharo-dev] NB and assembly methods (WAS: a lot of non-constructive things I refuse to talk about :)

[Eliot Miranda]

Hi Igor,


On Tue, Feb 11, 2014 at 7:16 AM, Igor Stasenko <siguctua at gmail.com> wrote:

>
> On 11 February 2014 10:56, Esteban Lorenzano <estebanlm at gmail.com> wrote:
>
>> >
>> > PS. I am hacking on protobuf in Pharo 3.0 and noticed cool NB assembler
>> in some places - are such methods creeping into base libraries now? What
>> does that mean for other CPUs? Sorry, for changing subject - feel free to
>> reply under different subject.
>>
>> Hi Goran,
>>
>> Yes... since NB and Athens, there are some methods in assembler spread
>> around. Is my opinion that all assembler-containing methods should be
>> refactored into a more comprensible unit (and that can be changed by
>> another set of assembly instrucction when we achieve multiplatform-NB).
>> Today they are all along Athens, and if we continue this path they will be
>> all along Pharo soon. I would better modularise that.
>> But Igor maybe has something different to say :)
>>
>> Esteban
>>
>
> not much to add to what i already been said:
>  - have any other CPU which can run pharo JIT? No? Will talk when it will
> be available.
>

There's an unfortunate chicken-and-egg problem here.  The way we develop a
new platform with Cog is to build a simulator for the CPU and run the VM
simulator above the CPU simulator.  But, by making Pharo non-portable
because NB only works on x86 you've made it extremely difficult to use a
Pharo image to develop the new platform.  That will be really important
when x86-64 comes into the picture, which it shoudl sometime later this
year.  In fact its worse.  Because there is no simulator support for NB I
can't even start up a Pharo image on the current simulator and that's
directly affecting my ability to sup[port Clément in the Sista work.   Note
that we already have an ARM simulator and unfortunately Pharo can't run on
that simulator either.  Some way of making NB get out of the way except on
x86 until it is portable would be much appreciated.

Doing anything before that is just waste of time.
>
> --
> Best regards,
> Igor Stasenko.
>



-- 
best,
Eliot
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20140211/3fbfa43e/attachment.htm