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

Tue Jul 5 18:08:30 CEST 2011

Hi all:
furthermore I believe is more challenging to defeat the other languages, but can you create the scenario, nobody will create it instead the own one creates for them. For instance how can you beat C++ in a 64 KB ram machine, I think is doable, this techniques have been applied and are doable in retrospective many years ago (think pdp-11 era) many of this were solved, but this piece of history was to do it once more nearer, unfortunately, this ideas came to nothing in current jobs market, which is under well understood, far more technical issues were solved and applied before, nowadays, this things come in webminars, where what is the idea behind that, to design solutions you don't have to do. It's just to solve the problem of half less memory with less than that, it might become available a superfaster computer substantially cheaper to buy, I think the last is the reality, very crude, this is a free-software lesson, you can co-create your own sector
 market research if it's well understood, if not go and make them happy shipping cheaper computers (meanwhile your problems aren't solved at all, you just buy it, now what)
Thanks in advance

--- El mar, 5/7/11, felipe valdez <felipevaldez at gmail.com> escribió:

De: felipe valdez <felipevaldez at gmail.com>
Asunto: Re: [M3devel] Google Benchmark - anyone interested in an Modula 3 version?
Para: "Daniel Alejandro Benavides D." <dabenavidesd at yahoo.es>
Fecha: martes, 5 de julio, 2011 10:28

such as, getting, a job, for instance...

On Tue, Jul 5, 2011 at 10:28 AM, felipe valdez <felipevaldez at gmail.com> wrote:

so, if I understand correctly  that would solve type inference, compiler speed, jvm code generation, equations in general, and constraint systems?

sounds like too much to take try at once.

why not just focusing one one particular problem?

how would resources be obtain in order to solve this problem (people-wise) ?

or do you intend to do so all by yourself, in your spare time?

I had the Idea that there were more pressing issues at hand...

On Tue, Jul 5, 2011 at 8:30 AM, Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es> wrote:

Hi all:
Additional understanding of seminal work might be some clues of what is about concurrent and distributed compilations, naturally, the case might be solved for those platforms but for the rest of the world, it would mean the execution time is perfect or reasonable in terms of theoretical performance. So the idea behind this is in those cases where the system needs to be expanded but it's also slower the ability to rule out RT penalties, now, when it comes to JIT, it's also true the performance might be overwhelming, but still the most difficult part is to type check, since majority of work has been done in theoretical grounds, there must be the case for the type-theoretical analysis of the system to try to close the gap between the acceptable and what it isn't, e.g we can't afford to have such large compilations on real-time but perhaps a balance
 between both performance and language capabilities, then it follows than not all the cases are so high degraded performance, but the cases where the idea is to diminish burden between the expected versus the actual implementation such as smaller footprint (traditionally hardware assisted interpretation, etc), but again if such resources are still limited then a shorter route might be the performance profiling with back end optimization, which is the harder to do since the burden could be too high but still the only route that might be taken, as for what I know, here are some cases you can have that, the most advanced perhaps is converting the type inference in a constraint satisfaction problem, which might solve it faster if enough information is given.

Similar approaches have been realized for Java systems, using Millner Calculus type system as a bases for an OO system
Now, I'm cautious because we then really might need measure and tuning of
 Constraint solver in the terms needed, which is possible to some extent, still this are very dynamic situations, but could be improve somehow the issue here.
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.47.1739

Such works like this can be applied to other domains, like equational reasoning for instance (see below p 10):
http://www.collectionscanada.gc.ca/obj/s4/f2/dsk2/tape16/PQDD_0009/NQ28028.pdf

Thanks in advance

--- El lun, 4/7/11, felipe valdez <felipevaldez at gmail.com> escribió:

De: felipe valdez <felipevaldez at gmail.com>
Asunto: Re: [M3devel] Google Benchmark - anyone interested in an Modula 3 version?
Para: "Daniel Alejandro Benavides D." <dabenavidesd at yahoo.es>

Fecha: lunes, 4 de julio, 2011 21:53

I
 don't understand the message.

On Sun, Jul 3, 2011 at 8:51 PM, Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es> wrote:

Hi all:

it
would mean to compete in such a way in a super computer an alternative
to run without garbage collection (not unsafe but ESC/Modula-3 RTError
and RTCollector disabled but for safe automatically collected objects,
again if we are strictly speaking of a naive implementation algorithm,
if parallelization is allowed under recursive or nested paradigms we
are set to compete too but not absolutely).
I mean, given the effort
they took to run the compiler and tune it for other languages and would
get same compiler effort plus ESC modelling to subsume the collector
for perhaps dynamic scripting like say e.g Obliq, we don't want to
compete with Obliq, but if we do I guess we can't compile JIT it since
all RT structures are interpreted not plain executed, so perhaps this
would be the logic ideal competence (again the type inference could
save us some time but dynamic typing is another thing, it must run its
structures dynamically even without those checks the code is not just
plain executed, if so, then maybe the main and ideal platform would be
again super computers anyway).
If there is such possibility, also we would potentially execute several tests for running smoothly, such as performance penalty run for the given platform (lets say the debugger is able to give you such tests if so)

Let me know if other requirements implications are something of relevance here.. Thanks in advance

--- El sáb, 2/7/11, felipe valdez <felipevaldez at gmail.com> escribió:

De: felipe valdez <felipevaldez at gmail.com>

Asunto: Re: [M3devel] Google Benchmark - anyone interested in an Modula 3 version?
Para: "Daniel Alejandro Benavides D." <dabenavidesd at yahoo.es>

Fecha: sábado, 2 de julio, 2011 13:54

s/doing/making/ this available 

On Fri, Jul 1, 2011 at 7:22 PM, Daniel Alejandro Benavides D. <dabenavidesd at yahoo.es> wrote:

Hi all:

Indeed there is one besides here:

http://www.hackernewsers.com/users/mhd.html

http://news.ycombinator.com/item?id=2615096

http://news.ycombinator.com/item?id=2615599

I think there should be something like that, is certainly a good initiative, still I don't know the rules (or other's rules) to apply for a submission, the requirements, and what will give us.

For one thing I knew the programming marathon in my undergraduate with and there was one of such interesting (for me I can say) contest system on-line, the obvious problem of that platform was there were just some languages available (one sent code, in my case C to server side and got results relatively quick on screen), but I liked the spirit it was for an Algorithms project and we had some categories (so everybody was working very hard to get the best results which I don't criticize at least in one, I did get a good timing but with a very simple program, I mean in loc). It was Valladolid UVA system.

As we have some native and gcc-based backend (in gcc lang, e.g another front end, but low-level, there is another one based on Scale in Java) I guess you would have to have access to that, or us to offer a cm3-ide instance open to you so you can compile it, I would like compile time optimizations (that is not updated in the cm3 source tree), but not so much back-end or linker enabled optimization (just if they are not possible in compile time, but just for that reason), I mean aggressive optimization where code doesn't do what it needs to do is suspicious in my way of thinking (but maybe this is too much burden, even for Modula-3 is a lot of work to configure anyway this).

Ok, let me know, the important info, and if a public instance of the compiler is needed I can provide one, but my hardware options are not too open (we could compete with some retro supercomputers, but this could slower things down for us, I mean, if more platforms are needed we can make an arrangement, I'm sure there should be people open to do that, if there is such wide interest to do so in contest organizer, etc end, maybe even gcc hackers as well).

Thanks everybody for doing this available and tell us please that for us on how to do that, thanks in advance

--- El vie, 1/7/11, Michael Franz <mvfranz at gmail.com> escribió:

> De: Michael Franz <mvfranz at gmail.com>

> Asunto: [M3devel] Google Benchmark - anyone interested in an Modula 3 version?

> Para: "m3devel" <m3devel at elegosoft.com>

> Fecha: viernes, 1 de julio, 2011 16:41

> Hi,

>

> A co-worker pointed out this benchmark

> http://www.readwriteweb.com/hack/2011/06/cpp-go-java-scala-performance-benchmark.php

> . I use Scala at work, and have used Java for years.

> I would be

> interested in the outcome of Modula 3 port.

> Unfortunately I don't

> have enough experience with M3 to do it myself (guess it

> could be a

> learning experience, but then I can complain about a lack

> of time).

>

> I also follow the development of D and there is a port

> going on in

> their community.

> http://www.digitalmars.com/d/archives/digitalmars/D/Port_a_benchmark_to_D_137420.html

>

>

> Michael

>

-- 
312-444-2124Skype: f3l.headhunterCasa: 8043901

-- 
312-444-2124Skype: f3l.headhunter

Casa: 8043901

-- 
312-444-2124Skype: f3l.headhunter

Casa: 8043901

-- 
312-444-2124Skype: f3l.headhunterCasa: 8043901

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://m3lists.elegosoft.com/pipermail/m3devel/attachments/20110705/13b8ce92/attachment-0002.html>