Hi!

I was wondering, is anyone aware if there been any efforts on using D as the foundation for a GPGPU language?

For example: Given that there is are LLVM based D and CUDA compilers, might there be a way to modify D to nVidia GPU-enabled DUDA similar to how C++ is modified to CUDA, without having to implement the entire compiler?

Keep up the interesting and good work!

Anders

I think CUDA is a completely pointless waste of time because of OpenCL.

On Mon, May 14, 2012 at 9:35 PM, <"Anders Sjögren\" <anders@sjogren.info>"@puremagic.com> wrote:

> Hi!
>
> I was wondering, is anyone aware if there been any efforts on using D as the foundation for a GPGPU language?
>
> For example: Given that there is are LLVM based D and CUDA compilers, might there be a way to modify D to nVidia GPU-enabled DUDA similar to how C++ is modified to CUDA, without having to implement the entire compiler?
>
> Keep up the interesting and good work!
>
> Anders
>



-- 
Bye,
Gor Gyolchanyan.

Hi Gor!

Then I guess you think C++ is a completely pointless waste of time because of C also? CUDA carries a subset of C++, while OpenCL does not, allowing for OO-programming, metaprogramming, etc. Adding to that, there is also strong IDE support (editing and debugging) for CUDA, while for OpenCL I have yet to find that (which would of course also be an issue for "DUDA"). Not _completely_ pointless yet, right?

Anyways, the original question was about GPU-oriented languages and D.

As a variant to CUDA in D (DUDA): If there was a way to use a subset of D (à la C++AMP) and have that turn into GPU and/or CPU (using the code-gen/backend of the LLVM CUDA compiler, or compile to OpenCL as a target for that matter), it'd be great. I too don't like vendor lock-in (CUDA=nVidia) and C++ has it's drawbacks as a language. The alternative of OpenCL is too much plain C for it to feel modern/productive. An alternative based on D could really have a potential of being great!?

To state it as an open question: What would be the D way of facilitating GPU(+CPU) programming?

Cheers

Anders

---

On Monday, 14 May 2012 at 18:24:26 UTC, Gor Gyolchanyan wrote:
> I think CUDA is a completely pointless waste of time because of OpenCL.
>
> On Mon, May 14, 2012 at 9:35 PM, <"Anders Sjögren\"
> <anders@sjogren.info>"@puremagic.com> wrote:
>
>> Hi!
>>
>> I was wondering, is anyone aware if there been any efforts on using D as
>> the foundation for a GPGPU language?
>>
>> For example: Given that there is are LLVM based D and CUDA compilers,
>> might there be a way to modify D to nVidia GPU-enabled DUDA similar to how
>> C++ is modified to CUDA, without having to implement the entire compiler?
>>
>> Keep up the interesting and good work!
>>
>> Anders

On Monday, 14 May 2012 at 18:24:26 UTC, Gor Gyolchanyan wrote:
> I think CUDA is a completely pointless waste of time because of OpenCL.

While I do agree it is better to use standard technologies, the truth is that CUDA wins out against OpenCL in most gamming and HPC scenarios, not to forget how the CUDA debugging tools are miles ahead from OpenCL ones.

As such, most GPGPU developers don't care about OpenCL.

--
Paulo

That's why CUDA is completely pointless: you're gonna wrap it anyway. Instead of wrapping a past-age vendor-locked pre-OpenCL technology, you could wrap modern portable technology.

The richness of the language is completely irrelevant, as it's gonna be
wrapped anyway.
Yes, CUDA is completely pointless.

On Mon, May 14, 2012 at 11:51 PM, <"Anders Sjögren\" <anders@sjogren.info>"@puremagic.com> wrote:

> Hi Gor!
>
> Then I guess you think C++ is a completely pointless waste of time because of C also? CUDA carries a subset of C++, while OpenCL does not, allowing for OO-programming, metaprogramming, etc. Adding to that, there is also strong IDE support (editing and debugging) for CUDA, while for OpenCL I have yet to find that (which would of course also be an issue for "DUDA"). Not _completely_ pointless yet, right?
>
> Anyways, the original question was about GPU-oriented languages and D.
>
> As a variant to CUDA in D (DUDA): If there was a way to use a subset of D (à la C++AMP) and have that turn into GPU and/or CPU (using the code-gen/backend of the LLVM CUDA compiler, or compile to OpenCL as a target for that matter), it'd be great. I too don't like vendor lock-in (CUDA=nVidia) and C++ has it's drawbacks as a language. The alternative of OpenCL is too much plain C for it to feel modern/productive. An alternative based on D could really have a potential of being great!?
>
> To state it as an open question: What would be the D way of facilitating
> GPU(+CPU) programming?
>
> Cheers
>
> Anders
>
> ---
>
>
> On Monday, 14 May 2012 at 18:24:26 UTC, Gor Gyolchanyan wrote:
>
>> I think CUDA is a completely pointless waste of time because of OpenCL.
>>
>> On Mon, May 14, 2012 at 9:35 PM, <"Anders Sjögren\" <anders@sjogren.info>"@puremag**ic.com <http://puremagic.com>> wrote:
>>
>>  Hi!
>>>
>>> I was wondering, is anyone aware if there been any efforts on using D as the foundation for a GPGPU language?
>>>
>>> For example: Given that there is are LLVM based D and CUDA compilers,
>>> might there be a way to modify D to nVidia GPU-enabled DUDA similar to
>>> how
>>> C++ is modified to CUDA, without having to implement the entire compiler?
>>>
>>> Keep up the interesting and good work!
>>>
>>> Anders
>>>
>>
>
>


-- 
Bye,
Gor Gyolchanyan.

Again, grammar is irrelevant, since it's gonna be wrapped anyway.
And debuggers are just a matter of time to get right.
But the redundant vendor-locked obsolete technology is not gonna change.

On Tue, May 15, 2012 at 12:44 AM, Paulo Pinto <pjmlp@progtools.org> wrote:

> On Monday, 14 May 2012 at 18:24:26 UTC, Gor Gyolchanyan wrote:
>
>> I think CUDA is a completely pointless waste of time because of OpenCL.
>>
>
> While I do agree it is better to use standard technologies, the truth is that CUDA wins out against OpenCL in most gamming and HPC scenarios, not to forget how the CUDA debugging tools are miles ahead from OpenCL ones.
>
> As such, most GPGPU developers don't care about OpenCL.
>
> --
> Paulo
>



-- 
Bye,
Gor Gyolchanyan.

On Mon, 14 May 2012 15:45:54 -0500, Gor Gyolchanyan <gor.f.gyolchanyan@gmail.com> wrote:

> That's why CUDA is completely pointless: you're gonna wrap it anyway.
> Instead of wrapping a past-age vendor-locked pre-OpenCL technology, you
> could wrap modern portable technology.
>
> The richness of the language is completely irrelevant, as it's gonna be
> wrapped anyway.
> Yes, CUDA is completely pointless.

CUDA, or more specifically the PTX byte code is an open standard with backends for x86, ATI and NVIDIA. And although we are providing a mapping from D to an intermediate representation, that representation should be able to express as many of the higher level concepts as possible in order for the JIT to generate the most optimal code. Technologically, OpenCL/DirectCompute are at a level below CUDA 1.0 (we are now at 4.x) since both standards were written with the lowest common denominator in mind. Furthermore, at least on the HPC side, until another manufacturer decides to make a GPU with ECC support, NVIDIA is the only game  in town.

"Anders Sjögren" <anders@sjogren.info>" <Anders =?UTF-8?B?U2rDtmdyZW4i?= <anders@sjogren.info>> wrote in message news:ixkdokrmwmnkmciehqog@forum.dlang.org...
>
> DUDA

Thanks, now I can't get "Camptown Races" out of my head!

;)

On Tuesday, 15 May 2012 at 02:36:25 UTC, Nick Sabalausky wrote:
> Thanks, now I can't get "Camptown Races" out of my head!
>
> ;)

;-)

Am 15.05.2012 10:22, schrieb "Anders Sjögren" <anders@sjogren.info>":
> On Tuesday, 15 May 2012 at 02:36:25 UTC, Nick Sabalausky wrote:
>> Thanks, now I can't get "Camptown Races" out of my head!
>>
>> ;)
>
> ;-)

Doo-dah! doo-dah!
;)