Gregor Richards wrote:

> Don Clugston wrote:
>> Tomas Lindquist Olsen wrote:
>>> Kiriakos Alexoglou wrote:
>>>
>>>> Please Walter add support for x86_64!
>>>>
>>>> I use Suse Linux 10.0 and developing in Qt and
>>>> I want to try make Qt bindings for dmd.
>>>>
>>>> The bindings that exist for Qt right now,
>>>> can all work with x86_64 architectures.
>>>>
>>>> In my personal opinion x86_64 support is
>>>> more important than adding additional feutures to dmd.
>> 
>> vote--. That may be true, but do you have any idea how much work this would involve?? (Hint: begin by implementing a linker from scratch). We'd lose Walter for a year! Has to happen someday, but I don't think Walter should put any thought into it until GDC-64 is firmly established.
>> 
>> OTOH, I hope we get a working GDC-x64 very soon.
> 
> vote-- as well.
> 
> For the record, you're at 0 votes now :P
> 
>   - Gregor Richards

I would agree with this too, but I believe Walter has stated in the past that he needs to make DMC 64-bit compatible anyway.

-- 
Lars Ivar Igesund
blog at http://larsivi.net
DSource, #d.tango & #D: larsivi
Dancing the Tango

what's so bad about cross compilation?
compilers have intermediate code generation for decades, therefore
having different code generation and -optimization backends is a good
thing. am i wrong?

BLS wrote:
> Tomas Lindquist Olsen schrieb:
>> BLS wrote:
>>
>>
>>> Seems to be a good chance to start re-implementing the complete D
>>> Tool-Chain Development in D.
>>> (Instead of using C and ASM)
>>> IMO D 2.0 should be implemented in D (seperated from 1.x) , even if the
>>> 2.0 Backend is closed source.
>>> Bjoern
>>>
>>>
>>> Kiriakos Alexoglou schrieb:
>>>
>>>> Please Walter add support for x86_64!
>>>>
>>>> I use Suse Linux 10.0 and developing in Qt and
>>>> I want to try make Qt bindings for dmd.
>>>>
>>>> The bindings that exist for Qt right now,
>>>> can all work with x86_64 architectures.
>>>>
>>>> In my personal opinion x86_64 support is
>>>> more important than adding additional feutures to dmd.
>>>>
>>>> Right now I have to switch to the 32 bit version of Qt 4.2 and start making the bindings for it. I have no other choice.
>>>>
>>>> There are so many additional registers waiting to get used by all of us! :-)
>>>>
>>>> Thank you for the exciting D language
>>>> you offered to all of us!
>>>> Keep up the good work!!!
>>>>
>>>>
>>>>
>>>> *I think the D Logo with the Dolphin on it at http://www.sukimashita.com/d/ is very nice!!!
>>>>
>>>>
>>
>>
>> By using the LLVM backend D could be implemented in D.
>> LLVM is C++ but you can output a ASM-like text file instead. I'm not sure
>> how much this would hurt performance, but I'm guessing it's not that
>> much.
>> Also compared to what is gained it's a small price to pay.
>>
>> LLVM has a bytecode VM, JIT and some pretty neat optimisation technology.
>>
>> I think it could be interesting...
> 
> NO !
> NO VM,NET or D to WhatTheHeck cross compilation
> 
> I simply vote for a D implemented in D. Frontend,  Backend, Linker ....
> the complete Toolchain..... 32/64 bit at your choice.
> The impact of having D in D for Tools like IDEs is significant.
> And :
> I would prefer to have all  *D Tools implemented as DDL*  guess why ?
> Bjoern
> 
> 

Hi Jascha,

> what's so bad about cross compilation?

In case that you are talking about GCC,
1)fat bottom binaries
2)a never ending compile link cycle.

> compilers have intermediate code generation for decades....
Yep. And this is good for what ?

However, we are leaving the 64 bit discussion.
Bjoern

Jascha Wetzel schrieb:
> what's so bad about cross compilation?
> compilers have intermediate code generation for decades, therefore
> having different code generation and -optimization backends is a good
> thing. am i wrong?
> 
> BLS wrote:
> 
>>Tomas Lindquist Olsen schrieb:
>>
>>>BLS wrote:
>>>
>>>
>>>
>>>>Seems to be a good chance to start re-implementing the complete D
>>>>Tool-Chain Development in D.
>>>>(Instead of using C and ASM)
>>>>IMO D 2.0 should be implemented in D (seperated from 1.x) , even if the
>>>>2.0 Backend is closed source.
>>>>Bjoern
>>>>
>>>>
>>>>Kiriakos Alexoglou schrieb:
>>>>
>>>>
>>>>>Please Walter add support for x86_64!
>>>>>
>>>>>I use Suse Linux 10.0 and developing in Qt and
>>>>>I want to try make Qt bindings for dmd.
>>>>>
>>>>>The bindings that exist for Qt right now,
>>>>>can all work with x86_64 architectures.
>>>>>
>>>>>In my personal opinion x86_64 support is
>>>>>more important than adding additional feutures to dmd.
>>>>>
>>>>>Right now I have to switch to the 32 bit version of Qt 4.2
>>>>>and start making the bindings for it. I have no other choice.
>>>>>
>>>>>There are so many additional registers waiting to get used
>>>>>by all of us! :-)
>>>>>
>>>>>Thank you for the exciting D language
>>>>>you offered to all of us!
>>>>>Keep up the good work!!!
>>>>>
>>>>>
>>>>>
>>>>>*I think the D Logo with the Dolphin on it at
>>>>>http://www.sukimashita.com/d/ is very nice!!!
>>>>>
>>>>>
>>>
>>>
>>>By using the LLVM backend D could be implemented in D.
>>>LLVM is C++ but you can output a ASM-like text file instead. I'm not sure
>>>how much this would hurt performance, but I'm guessing it's not that
>>>much.
>>>Also compared to what is gained it's a small price to pay.
>>>
>>>LLVM has a bytecode VM, JIT and some pretty neat optimisation technology.
>>>
>>>I think it could be interesting...
>>
>>NO !
>>NO VM,NET or D to WhatTheHeck cross compilation
>>
>>I simply vote for a D implemented in D. Frontend,  Backend, Linker ....
>>the complete Toolchain..... 32/64 bit at your choice.
>>The impact of having D in D for Tools like IDEs is significant.
>>And :
>>I would prefer to have all  *D Tools implemented as DDL*  guess why ?
>>Bjoern
>>
>>

> In case that you are talking about GCC,
> 1)fat bottom binaries
> 2)a never ending compile link cycle.

i wasn't specifically talking about any compiler.
it's probably true that GCC has the properties you describe, but i don't
understand why they are caused by intermediate code generation or cross
compilation.

>> compilers have intermediate code generation for decades....
> Yep. And this is good for what ?

- easier/faster optimization
- machine independent optimization
- portability. not just different platforms, but also different feature
sets and versions

maybe that's just theory. i never implemented an optimizing compiler myself. but that's what the dragonbook and friends say. i'm very interested in more practical knowledge about these things. right now i just can't see why ICG should cause any trouble.

BLS wrote:
> Hi Jascha,
> 
>> what's so bad about cross compilation?
> 
> In case that you are talking about GCC,
> 1)fat bottom binaries
> 2)a never ending compile link cycle.
> 
>> compilers have intermediate code generation for decades....
> Yep. And this is good for what ?
> 
> However, we are leaving the 64 bit discussion.
> Bjoern
> 
> Jascha Wetzel schrieb:
>> what's so bad about cross compilation?
>> compilers have intermediate code generation for decades, therefore
>> having different code generation and -optimization backends is a good
>> thing. am i wrong?
>>
>> BLS wrote:
>>
>>> Tomas Lindquist Olsen schrieb:
>>>
>>>> BLS wrote:
>>>>
>>>>
>>>>
>>>>> Seems to be a good chance to start re-implementing the complete D
>>>>> Tool-Chain Development in D.
>>>>> (Instead of using C and ASM)
>>>>> IMO D 2.0 should be implemented in D (seperated from 1.x) , even if
>>>>> the
>>>>> 2.0 Backend is closed source.
>>>>> Bjoern
>>>>>
>>>>>
>>>>> Kiriakos Alexoglou schrieb:
>>>>>
>>>>>
>>>>>> Please Walter add support for x86_64!
>>>>>>
>>>>>> I use Suse Linux 10.0 and developing in Qt and
>>>>>> I want to try make Qt bindings for dmd.
>>>>>>
>>>>>> The bindings that exist for Qt right now,
>>>>>> can all work with x86_64 architectures.
>>>>>>
>>>>>> In my personal opinion x86_64 support is
>>>>>> more important than adding additional feutures to dmd.
>>>>>>
>>>>>> Right now I have to switch to the 32 bit version of Qt 4.2 and start making the bindings for it. I have no other choice.
>>>>>>
>>>>>> There are so many additional registers waiting to get used by all of us! :-)
>>>>>>
>>>>>> Thank you for the exciting D language
>>>>>> you offered to all of us!
>>>>>> Keep up the good work!!!
>>>>>>
>>>>>>
>>>>>>
>>>>>> *I think the D Logo with the Dolphin on it at http://www.sukimashita.com/d/ is very nice!!!
>>>>>>
>>>>>>
>>>>
>>>>
>>>> By using the LLVM backend D could be implemented in D.
>>>> LLVM is C++ but you can output a ASM-like text file instead. I'm not
>>>> sure
>>>> how much this would hurt performance, but I'm guessing it's not that
>>>> much.
>>>> Also compared to what is gained it's a small price to pay.
>>>>
>>>> LLVM has a bytecode VM, JIT and some pretty neat optimisation technology.
>>>>
>>>> I think it could be interesting...
>>>
>>> NO !
>>> NO VM,NET or D to WhatTheHeck cross compilation
>>>
>>> I simply vote for a D implemented in D. Frontend,  Backend, Linker ....
>>> the complete Toolchain..... 32/64 bit at your choice.
>>> The impact of having D in D for Tools like IDEs is significant.
>>> And :
>>> I would prefer to have all  *D Tools implemented as DDL*  guess why ?
>>> Bjoern
>>>
>>>

vote++
The lack of x86_64 support got me mad when i tried to use derelict. :/

Jascha Wetzel wrote:
>
>>> compilers have intermediate code generation for decades....
>> Yep. And this is good for what ?
> 
> - easier/faster optimization
> - machine independent optimization
> - portability. not just different platforms, but also different feature
> sets and versions
> 
> maybe that's just theory. i never implemented an optimizing compiler
> myself. but that's what the dragonbook and friends say. i'm very
> interested in more practical knowledge about these things. right now i
> just can't see why ICG should cause any trouble.

Me either.  I'd think it would be far preferable to the alternatives.


Sean