I suppose people might be wondering what I'm working on with the compiler at the moment. After all, why did I waste time implementing an inline assembler, when so much needs to be done with the core language?

The answer is the garbage collector. It is currently written in C++, and is single threaded. It's the main barrier to getting D to support multithreaded programming. To upgrade the garbage collector, it needs to handle exceptions.

D has great syntax for synchronizing and handling exceptions, but those do not fit in well with C++ exceptions. So, the garbage collector needs to be rewritten in D. Rewriting it in D needs, well, a few lines of inline assembler.

Now that the inline assembler is done (and I also put the instruction set support enhancements back into the C compiler), I am working on converting the garbage collector to D. My aim is to remove all the C and C++ code from the runtime library, and have it fully bootstrapped into D.

Writing the garbage collector in D proves that D can write system level code. (How many other pretender successors to C++ can you write a real gc in? <g>)

If you used the CLR from .NET you would get a garbage collector "for free". :-)

   Dan

"Walter" <walter@digitalmars.com> wrote in message news:a6cclp$1mqi$1@digitaldaemon.com...
> I suppose people might be wondering what I'm working on with the compiler
at
> the moment. After all, why did I waste time implementing an inline assembler, when so much needs to be done with the core language?
>
> The answer is the garbage collector. It is currently written in C++, and
is
> single threaded. It's the main barrier to getting D to support
multithreaded
> programming. To upgrade the garbage collector, it needs to handle exceptions.
>
> D has great syntax for synchronizing and handling exceptions, but those do not fit in well with C++ exceptions. So, the garbage collector needs to be rewritten in D. Rewriting it in D needs, well, a few lines of inline assembler.
>
> Now that the inline assembler is done (and I also put the instruction set support enhancements back into the C compiler), I am working on converting the garbage collector to D. My aim is to remove all the C and C++ code
from
> the runtime library, and have it fully bootstrapped into D.
>
> Writing the garbage collector in D proves that D can write system level code. (How many other pretender successors to C++ can you write a real gc in? <g>)
>
>

"J. Daniel Smith" <j_daniel_smith@HoTMaiL.com> wrote in message news:a6idtd$1am9$1@digitaldaemon.com...
> If you used the CLR from .NET you would get a garbage collector "for
free".
> :-)

Can't write system code in .NET.

It's your language, so you're of course free to do whatever you want with it.

But I still contend that by focusing on the "systems" aspect, you're loosing a lot of mindshare; I'd much rather see D as an "improved C++" (or "an alternative to C#") rather than a wholesale flat-out replacment for C/C++. Counting either lines-of-code or number of developers, I suspect only a very small amount of "systems" code is written these days.  Device drivers is the most obvious example where it is still necessary for a non-trivial number of programmers to get down to the "bare metal"; but D has the potential to be useful to a much larger audience than low-level bit-twidlers.

   Dan

"Walter" <walter@digitalmars.com> wrote in message news:a6j2nt$1jqe$1@digitaldaemon.com...
>
> "J. Daniel Smith" <j_daniel_smith@HoTMaiL.com> wrote in message news:a6idtd$1am9$1@digitaldaemon.com...
> > If you used the CLR from .NET you would get a garbage collector "for
> free".
> > :-)
>
> Can't write system code in .NET.
>
>

"J. Daniel Smith" <j_daniel_smith@HoTMaiL.com> wrote in message news:a6j877$1m4a$1@digitaldaemon.com...
> "Walter" <walter@digitalmars.com> wrote in message news:a6j2nt$1jqe$1@digitaldaemon.com...
> >
> > "J. Daniel Smith" <j_daniel_smith@HoTMaiL.com> wrote in message news:a6idtd$1am9$1@digitaldaemon.com...
> > > If you used the CLR from .NET you would get a garbage collector "for
> > free".
> > > :-)
> >
> > Can't write system code in .NET.
>
> It's your language, so you're of course free to do whatever you want with it.
>
> But I still contend that by focusing on the "systems" aspect, you're
loosing
> a lot of mindshare; I'd much rather see D as an "improved C++" (or "an alternative to C#") rather than a wholesale flat-out replacment for C/C++. Counting either lines-of-code or number of developers, I suspect only a
very
> small amount of "systems" code is written these days.  Device drivers is
the
> most obvious example where it is still necessary for a non-trivial number
of
> programmers to get down to the "bare metal"; but D has the potential to be useful to a much larger audience than low-level bit-twidlers.
>
>    Dan
>

I think Walter meant that D needs a system level garbage
collector, not some CLR interpreted one, and that such a
gc cannot be built with .NET?


--
Stijn
OddesE_XYZ@hotmail.com
http://OddesE.cjb.net
__________________________________________
Remove _XYZ from my address when replying by mail

"J. Daniel Smith" a écrit :

> It's your language, so you're of course free to do whatever you want with it.
>
> But I still contend that by focusing on the "systems" aspect, you're loosing a lot of mindshare; I'd much rather see D as an "improved C++" (or "an alternative to C#") rather than a wholesale flat-out replacment for C/C++. Counting either lines-of-code or number of developers, I suspect only a very small amount of "systems" code is written these days.  Device drivers is the most obvious example where it is still necessary for a non-trivial number of programmers to get down to the "bare metal"; but D has the potential to be useful to a much larger audience than low-level bit-twidlers.
>
>    Dan
>
> "Walter" <walter@digitalmars.com> wrote in message news:a6j2nt$1jqe$1@digitaldaemon.com...
> >
> > "J. Daniel Smith" <j_daniel_smith@HoTMaiL.com> wrote in message news:a6idtd$1am9$1@digitaldaemon.com...
> > > If you used the CLR from .NET you would get a garbage collector "for
> > free".
> > > :-)
> >
> > Can't write system code in .NET.
> >
> >
>
>
>

- If you can do system programming then that mean you are not limited by
the langage itself. That mean all the possibility remains
open,  witch is a good thing for a new langage. Lets restric it the
later possible.
- system programming is often complex and difficult to debug.
For that, a new tool like D is interresting.

roland

D is intended as the evolutionary next step for the kinds of programs that C and C++ targets. That is a great deal broader than what .net addresses.

From a pragmatic standpoint, D is new and has yet to prove itself. .net is brand new and has yet to live up to any of its promises. It seems to me that tying D to .net at this time will add unnecessary risk.

I think the bare metal bit twiddlers need a great language, too <g>.

"J. Daniel Smith" <j_daniel_smith@HoTMaiL.com> wrote in message news:a6j877$1m4a$1@digitaldaemon.com...
> It's your language, so you're of course free to do whatever you want with it.
>
> But I still contend that by focusing on the "systems" aspect, you're
loosing
> a lot of mindshare; I'd much rather see D as an "improved C++" (or "an alternative to C#") rather than a wholesale flat-out replacment for C/C++. Counting either lines-of-code or number of developers, I suspect only a
very
> small amount of "systems" code is written these days.  Device drivers is
the
> most obvious example where it is still necessary for a non-trivial number
of
> programmers to get down to the "bare metal"; but D has the potential to be useful to a much larger audience than low-level bit-twidlers.
>
>    Dan