[M3devel] JIT [WAS: Google Benchmark - anyone interested in an Modula 3 version?]

[Hendrik Boom]

If understand correctly, these results are about figuring out what types 
the programmer meant if he failed to specify any.  My view is that the 
programmer probably knew full well what he meant, and he might as well 
say so, in the interest of clarity if nothing else.  I find programs 
with secret types to be incomprehensible, even while I'm writing them.

And getting error messages if what I wrote doesn't match what I 
intended, and getting better run-time, well, those are wonderful 
consequences.

-- hendrik

On Tue, Jul 05, 2011 at 01:31:12AM +0100, Daniel Alejandro Benavides D. wrote:
> Hi all:
> In fact there is some reqwork to make things happen faster in Abadi-Cardelli:
> http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.36.110
> 
> Never looked that paper, if I may say so, the approach by Tian Zhao omitted it, but as for your worries, this could make it work faster.
> I guess there should be some Turing awards waiting for somebody who can run even lower, but this geniuses come counted by the fingers of one hand, but you can't say this is the end, that's for sure, but then there's even worse situations if you add the self-object extension like for JS, and this is a certainly an open problem (even that I tried but for a good type theorist could be not that hard, as I consider myself just a follower of them).
> But Abadi-Cardelli's work was a bit hit, and it's quite practical in terms of research results I believe. Yet this is not the last work in this are.
> Thanks in advance
> 
> --- El lun, 4/7/11, Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es> escribió:
> 
> > De: Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es>
> > Asunto: Re: [M3devel] JIT [WAS: Google Benchmark - anyone interested in an Modula 3 version?]
> > Para: m3devel at elegosoft.com, "Hendrik Boom" <hendrik at topoi.pooq.com>
> > Fecha: lunes, 4 de julio, 2011 19:16
> > Hi all:
> > I think yes, any other full bloated object oriented
> > language has that floor, of course, you can help the
> > compiler by giving some annotations, but for the theoretical
> > results, this is what is proved to be truth.
> > 
> > Thanks in advance
> > 
> > --- El lun, 4/7/11, Hendrik Boom <hendrik at topoi.pooq.com>
> > escribió:
> > 
> > > De: Hendrik Boom <hendrik at topoi.pooq.com>
> > > Asunto: Re: [M3devel] JIT [WAS: Google Benchmark -
> > anyone interested in an Modula 3 version?]
> > > Para: m3devel at elegosoft.com
> > > Fecha: lunes, 4 de julio, 2011 19:13
> > > On Mon, Jul 04, 2011 at 06:08:54PM
> > > +0100, Daniel Alejandro Benavides D. wrote:
> > > > 
> > > > So this time I'm happy to be wrong, we can do
> > indeed
> > > JIT but we should 
> > > > measure the cost (as for type inference could
> > for
> > > **imp**ç run O(n^3) 
> > > > and currently run O(n^5), it might be necessary
> > do
> > > decompiling plus 
> > > > naturally compiling and forth inc ase of NetObj,
> > etc
> > > 
> > > O(n^3)? O(n^5)? If that's performance in practical
> > > situations, instead 
> > > of just a class of theoretical examples it's not what
> > we
> > > need for 
> > > everyday use.  I'd call it useless.  It's
> > > defnietely not what I'm 
> > > thinking of. 
> > > 
> > > -- hendrik
> > >
> >