Now that D 2.0 is fairly stable, are there any plans of writing the official DMD compiler with the D 2.0 language vs the present language of C++?

DMD 2.0 would have to be feature frozen and then DMD 3.0 could be written with the previous DMD 2.0 compiler.

What are your thoughts?

On 01/08/2013 10:48 AM, Tim Krimm wrote:
>
> Now that D 2.0 is fairly stable, are there any plans of writing the
> official DMD compiler with the D 2.0 language vs the present language of
> C++?
>
> DMD 2.0 would have to be feature frozen and then DMD 3.0 could be
> written with the previous DMD 2.0 compiler.
>
> What are your thoughts?

There is Denis Koroskin's ddmd:

  http://forum.dlang.org/thread/i4obl3$kgk$1@digitalmars.com?page=1

If this page is up to date, ddmd is currently at dmd 2.040's level:

  http://www.dsource.org/projects/ddmd

Ali

On Tue, Jan 8, 2013 at 8:18 PM, Ali Çehreli <acehreli@yahoo.com> wrote:

> On 01/08/2013 10:48 AM, Tim Krimm wrote:
>
>>
>> Now that D 2.0 is fairly stable, are there any plans of writing the official DMD compiler with the D 2.0 language vs the present language of C++?
>>
>> DMD 2.0 would have to be feature frozen and then DMD 3.0 could be written with the previous DMD 2.0 compiler.
>>
>> What are your thoughts?
>>
>
> There is Denis Koroskin's ddmd:
>
>   http://forum.dlang.org/thread/**i4obl3$kgk$1@digitalmars.com?**page=1<http://forum.dlang.org/thread/i4obl3$kgk$1@digitalmars.com?page=1>
>
> If this page is up to date, ddmd is currently at dmd 2.040's level:
>
>   http://www.dsource.org/**projects/ddmd<http://www.dsource.org/projects/ddmd>
>
> Ali
>

Isn't SDC also in D? (Bernard Helyer and friends)
https://github.com/bhelyer/SDC


Also, Timon Gehr spoke of his own front-end (assumed to be in D) in the past, but did not provide any link to it.


But, to answer the OP question: no, there are no plan to switch to D for the reference compiler in the near future, as far as I can tell.

On Tue, Jan 08, 2013 at 07:48:58PM +0100, Tim Krimm wrote:
> 
> Now that D 2.0 is fairly stable, are there any plans of writing the official DMD compiler with the D 2.0 language vs the present language of C++?
> 
> DMD 2.0 would have to be feature frozen and then DMD 3.0 could be written with the previous DMD 2.0 compiler.
> 
> What are your thoughts?

Philosophically, I like this idea. D should eat its own dogfood to prove its own worth. :)

However, having the D compiler itself written in D, means we will have trouble bootstrapping it on new platforms. The advantage of having a C++ implementation is that C/C++ compilers are almost the first thing that gets implemented on a new platform, so you can almost always count on their existence. So you can just compile DMD and away you go.

We *could* write a cross-compiler, of course, but it still requires that you first target the D compiler (written in D) to the new platform, and then cross-compile itself to that platform.  Whereas with DMD, you just use the target platform's C++ compiler and you're up and running.


T

-- 
What's a "hot crossed bun"? An angry rabbit.

On Tuesday, 8 January 2013 at 21:57:17 UTC, H. S. Teoh wrote:
> On Tue, Jan 08, 2013 at 07:48:58PM +0100, Tim Krimm wrote:
>> 
>> Now that D 2.0 is fairly stable, are there any plans of writing the
>> official DMD compiler with the D 2.0 language vs the present
>> language of C++?
>> 
>> DMD 2.0 would have to be feature frozen and then DMD 3.0 could be
>> written with the previous DMD 2.0 compiler.
>> 
>> What are your thoughts?
>
> Philosophically, I like this idea. D should eat its own dogfood to prove
> its own worth. :)
>
> However, having the D compiler itself written in D, means we will have
> trouble bootstrapping it on new platforms. The advantage of having a C++
> implementation is that C/C++ compilers are almost the first thing that
> gets implemented on a new platform, so you can almost always count on
> their existence. So you can just compile DMD and away you go.
>
> We *could* write a cross-compiler, of course, but it still requires that
> you first target the D compiler (written in D) to the new platform, and
> then cross-compile itself to that platform.  Whereas with DMD, you just
> use the target platform's C++ compiler and you're up and running.
>
>
> T

I think the OP implied that we could build DMD2 from its C++ source on any platform and then DMD3 from its D source with DMD2.

On Tuesday, 8 January 2013 at 21:57:17 UTC, H. S. Teoh wrote:
> We *could* write a cross-compiler, of course, but it still requires that
> you first target the D compiler (written in D) to the new platform, and
> then cross-compile itself to that platform.  Whereas with DMD, you just
> use the target platform's C++ compiler and you're up and running.

…except that you can't actually use that compiler for anything, because – wait for it – it still needs to be retargeted for the new platform. What kind of new system are you thinking of for which the first use case would be compiling x86 executables?

David

On Wed, Jan 09, 2013 at 01:33:35AM +0100, David Nadlinger wrote:
> On Tuesday, 8 January 2013 at 21:57:17 UTC, H. S. Teoh wrote:
> >We *could* write a cross-compiler, of course, but it still requires that you first target the D compiler (written in D) to the new platform, and then cross-compile itself to that platform.  Whereas with DMD, you just use the target platform's C++ compiler and you're up and running.
> 
> …except that you can't actually use that compiler for anything, because – wait for it – it still needs to be retargeted for the new platform. What kind of new system are you thinking of for which the first use case would be compiling x86 executables?
[...]

Heh, you're right. I appear to be having a streak of making a fool of myself today.


T

-- 
Today's society is one of specialization: as you grow, you learn more and more about less and less. Eventually, you know everything about nothing.

On 01/08/2013 06:38 PM, H. S. Teoh wrote:
> On Wed, Jan 09, 2013 at 01:33:35AM +0100, David Nadlinger wrote:
>> On Tuesday, 8 January 2013 at 21:57:17 UTC, H. S. Teoh wrote:
>>> We *could* write a cross-compiler, of course, but it still requires
>>> that you first target the D compiler (written in D) to the new
>>> platform, and then cross-compile itself to that platform.  Whereas
>>> with DMD, you just use the target platform's C++ compiler and you're
>>> up and running.
>>
>> …except that you can't actually use that compiler for anything,
>> because – wait for it – it still needs to be retargeted for the new
>> platform. What kind of new system are you thinking of for which the
>> first use case would be compiling x86 executables?
> [...]
>
> Heh, you're right. I appear to be having a streak of making a fool of
> myself today.
>
>
> T
>

I'm embarrassed to admit that I didn't understand the problem until I read your response...

I love the idea of a D compiler in D.
Walter might have issues with working on any other compiler backend other than digitalmars, wasn't that what was determined by previous threads?

On 01/08/2013 01:06 PM, Philippe Sigaud wrote:
> On Tue, Jan 8, 2013 at 8:18 PM, Ali Çehreli<acehreli@yahoo.com>  wrote:

>> There is Denis Koroskin's ddmd:

>>    http://www.dsource.org/**projects/ddmd<http://www.dsource.org/projects/ddmd>

> Isn't SDC also in D? (Bernard Helyer and friends)
> https://github.com/bhelyer/SDC

And Aziz Köksal's dil:

  https://github.com/azizk/dil

Ali

On Tuesday, 8 January 2013 at 21:57:17 UTC, H. S. Teoh wrote:
> Philosophically, I like this idea. D should eat its own dogfood


+11111!1!!1eleven1!!!!1!