View mode: basic / threaded / horizontal-split · Log in · Help
July 02, 2008
Re: DMD back end dev-kit
Reply to Bill,

> It's been talked about before, but exactly how much money do the
> people who own the copyright to DMD's backend want in order to make it
> freely available freely -- as at least a lib?  It's hard to believe
> anyone would appraise that thing with a high dollar value given its
> less than stellar optimization abilities, and since everyone knows you
> can't make much money off dev tools these days anyway.
> 
> I'd happily chip in $100 US if it meant finally having an official D
> compiler that anyone can recompile for themselves just by typing "make
> dmd".
> 
> Or is it more a matter of not being able to find the copyright holders
> in question to ask them?  If so, give us the names and whatever info
> you have, and I bet someone will be able to track 'em down.
> 
> --bb
> 

I'd chip in somthing, $20 at least, maybe more, can't say exactly how much 
right now... (man I whish Marx hadn't been wrong).
July 02, 2008
Re: DMD back end dev-kit
nazo Wrote:
> Also opensource backend will be needed in phobos for runtime assembler 
> and runtime D compiling feature :)
> 
> Runtime generations:
> C# has System.Reflection.Assembly that is runtime assembler.
> Xbyak library is runtime assembler for C++.
> tcc can be used as runtime C compiler.
> but D?

IIRC, Burton Radons was able to do just that... but with the DMD as an
executable, not a lib. See:

http://www.digitalmars.com/webnews/newsgroups.php?art_group=digitalmars.D.announce&article_id=11436
July 02, 2008
Re: DMD back end dev-kit
John Reimer Wrote:

> Hello Robert,
> 
> > John Reimer Wrote:
> > 
> >> Hello Bill,
> >> 
> >>> It's been talked about before, but exactly how much money do the
> >>> people who own the copyright to DMD's backend want in order to make
> >>> it freely available freely -- as at least a lib?  It's hard to
> >>> believe anyone would appraise that thing with a high dollar value
> >>> given its less than stellar optimization abilities, and since
> >>> everyone knows you can't make much money off dev tools these days
> >>> anyway.
> >>> 
> >>> I'd happily chip in $100 US if it meant finally having an official D
> >>> compiler that anyone can recompile for themselves just by typing
> >>> "make dmd".
> >>> 
> >>> Or is it more a matter of not being able to find the copyright
> >>> holders in question to ask them?  If so, give us the names and
> >>> whatever info you have, and I bet someone will be able to track 'em
> >>> down.
> >>> 
> >>> --bb
> >>> 
> >> I'd give another $100.
> >> 
> >> -JJR
> >> 
> > "Economic Stimulus Act" FTW.
> > 
> > Seriously, though, there's GDC. Maybe LLVMDC, too. So while I wouldn't
> > say no to a compiling lib of the backend, is it really needed?
> > 
> 
> 
> LLVMDC is only partially operational.  gdc hasn't proven that it can "keep 
> up".  They are both very important projects, but having the complete dmd 
> compiler means even more potential for active fixing of elusive bugs in the 
> reference compiler, code generator optimization and updates for more recent 
> instructions sets, improved object file format support on win32 (coff), shared-lib 
> fixes for linux(?), experimentation, improved tool support, compiler embedding, 
> immediate bug hunting expeditions for library developers... and easier distribution 
> of the compiler in general -- many things that Walter doesn't have time to 
> fix.  It's not likely to happen even with the economic incentive.  But having 
> a completely open reference compiler can mean a lot when it comes to getting 
> D adopted.  This has been one of D's major handicaps, I think, for a long 
> time.
> 
> -JJR

It appears I have been officially "pwned" here. You & Bill are right; 
GDC/LLVMDC isn't sufficient.
July 03, 2008
Re: DMD back end dev-kit
Hello Robert,


>> 
> It appears I have been officially "pwned" here. You & Bill are right;
> GDC/LLVMDC isn't sufficient.
> 


Well, GDC or LLVMDC might have been the way to go... if Walter were using/developing 
them as a reference compiler. :)

But that is something he apparently cannot do lest he "taint" the code.

-JJR
July 03, 2008
Re: DMD back end dev-kit
Robert Fraser wrote:
> It appears I have been officially "pwned" here. You & Bill are right; 
> GDC/LLVMDC isn't sufficient.

Or we could politely ask Walter to switch to llvm or gcc for the backend 
for dmd. Hey, he actually changed opEquals to bool...
July 03, 2008
Re: DMD back end dev-kit
Hello Christopher,

> Robert Fraser wrote:
> 
>> It appears I have been officially "pwned" here. You & Bill are right;
>> GDC/LLVMDC isn't sufficient.
>> 
> Or we could politely ask Walter to switch to llvm or gcc for the
> backend for dmd. Hey, he actually changed opEquals to bool...
> 


Similar has been suggested on several occasions.  If you take a peek at previous 
postings in this group on the topic you will see his answer which invariably 
is "no" with a short explanation about "tainting" (or something similar). 
He doesn't want to look at any other compiler code (opensource or otherwise). 
It's a self-imposed rule to simplify his defense in any potential legal 
disputes, as I understand it, since he appears to still be in the compiler 
business.

This is completely understandable, but I think it hampers D's progress to 
some degree.  I'm not so sure we can ever get the backend by slapping down 
some money, but I'd agree that it would be great if something were done... 
moving to an opensource backend of one sort or another would be extremely 
good for D.

Whether or not our request is successful, the statement made here (about 
the donation) at least expresses how important it is to this community to 
see an active opensource backend as part of the reference compiler.  But 
we probably shouldn't get our hopes up too much: the issue hasn't budged 
one iota for quite a long time.  Yet, you never know. :)

-JJR
July 03, 2008
Re: DMD back end dev-kit
John Reimer wrote:
> Hello Christopher,
> 
>> Robert Fraser wrote:
>>
>>> It appears I have been officially "pwned" here. You & Bill are right;
>>> GDC/LLVMDC isn't sufficient.
>>>
>> Or we could politely ask Walter to switch to llvm or gcc for the
>> backend for dmd. Hey, he actually changed opEquals to bool...
>>
> 
> 
> Similar has been suggested on several occasions.  If you take a peek at 
> previous postings in this group on the topic you will see his answer 
> which invariably is "no" with a short explanation about "tainting" (or 
> something similar). He doesn't want to look at any other compiler code 
> (opensource or otherwise). It's a self-imposed rule to simplify his 
> defense in any potential legal disputes, as I understand it, since he 
> appears to still be in the compiler business.
> 
> This is completely understandable, but I think it hampers D's progress 
> to some degree.  I'm not so sure we can ever get the backend by slapping 
> down some money, but I'd agree that it would be great if something were 
> done... moving to an opensource backend of one sort or another would be 
> extremely good for D.
> 
> Whether or not our request is successful, the statement made here (about 
> the donation) at least expresses how important it is to this community 
> to see an active opensource backend as part of the reference compiler.  
> But we probably shouldn't get our hopes up too much: the issue hasn't 
> budged one iota for quite a long time.  Yet, you never know. :)

IIRC, earlier this year, Walter said he was still trying to get it 
released, and did expect to be successful eventually.

I doubt that the problem has much to do with money. Murky legal issues 
could well be a major part of it. I recently got some code from my 
university days open-sourced, and it was extremely painful and slow -- 
even after getting in-principle agreement from *everyone* who could 
possibly be a copyright owner, none of them were certain that they had 
authority to approve its release.
July 03, 2008
Re: DMD back end dev-kit
Hello Don,


> IIRC, earlier this year, Walter said he was still trying to get it
> released, and did expect to be successful eventually.


Hmmmm, I must have missed or forgotten that post.  That's better news than 
I expected.


> I doubt that the problem has much to do with money. Murky legal issues
> could well be a major part of it. I recently got some code from my
> university days open-sourced, and it was extremely painful and slow --
> even after getting in-principle agreement from *everyone* who could
> possibly be a copyright owner, none of them were certain that they had
> authority to approve its release.
> 


I think a few (now) opensource projects have been that way.  OpenWatcom, 
I think, had a long process of source cleanup and legal preparation/licensing 
before they could release the code.  It's probably the same with Java and 
OpenSolaris to some degree.

-JJR
July 03, 2008
Re: DMD back end dev-kit
Reply to John,

> Hello Christopher,
> 
>> Robert Fraser wrote:
>> 
>>> It appears I have been officially "pwned" here. You & Bill are
>>> right; GDC/LLVMDC isn't sufficient.
>>> 
>> Or we could politely ask Walter to switch to llvm or gcc for the
>> backend for dmd. Hey, he actually changed opEquals to bool...
>> 
> Similar has been suggested on several occasions.  If you take a peek
> at previous
> postings in this group on the topic you will see his answer which
> invariably
> is "no" with a short explanation about "tainting" (or something
> similar).
> He doesn't want to look at any other compiler code (opensource or
> otherwise).
> It's a self-imposed rule to simplify his defense in any potential
> legal
> disputes, as I understand it, since he appears to still be in the
> compiler
> business.
> 
> This is completely understandable, but I think it hampers D's progress
> to some degree.  I'm not so sure we can ever get the backend by
> slapping down some money, but I'd agree that it would be great if
> something were done... moving to an opensource backend of one sort or
> another would be extremely good for D.
> 
> Whether or not our request is successful, the statement made here
> (about the donation) at least expresses how important it is to this
> community to see an active opensource backend as part of the reference
> compiler.  But we probably shouldn't get our hopes up too much: the
> issue hasn't budged one iota for quite a long time.  Yet, you never
> know. :)
> 

We could do it in reverse; compile the GCC backend as a lib, and ship that 
and the minimal headers to Walter and then he could develop DMD with that 
without ever seeing a line of (executable) GCC code. He might even be able 
to build and ship both GCC and DMC versions of DMD.
July 03, 2008
Re: DMD back end dev-kit
Reply to don,

> IIRC, earlier this year, Walter said he was still trying to get it
> released, and did expect to be successful eventually.
> 
> I doubt that the problem has much to do with money. 

The money still works as a "statement of interest" of sorts and could be 
used as "incentive" to any reluctant parties.
1 2 3
Top | Discussion index | About this forum | D home