llvm dev - Mar 2011 - [LLVMdev] Using LLVM to convert a language to another language

Hi,

Sorry for my newbies questions... but here it is...

My goal is to have a "shading language", it is very similar to the
"C" language but with special tokens.

I have a parser and a lexer done with Lex/bison, once I have lexed/parsed my
"shading language" I must create 3 new "source code" in
OpenCL.

It mean that by example for the following :

shader matte(float Kd, color c)
{
  Ci = diffuse(Kd * c);
}

I have to create 3 methods in OpenCL, by example

1 - matte_sampling
2 - matte_pdf
3 - matte_f

Each of theses method is in OpenCL (It is like C too).

So, I would like to use LLVM to create an optimized version of my shader code.
But once I have the LLVM byte code... how can I parse it "easily" to
create my shader code ?

NB: also, you can tell me if it is really a good idea !

Thanks

Probably the easiest way is to take the C backend and modify it to handle the
special OpenCL constructs that you generate.
> -----Original Message-----
> From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at
cs.uiuc.edu]
> On Behalf Of krys at polarlights.net
> Sent: Tuesday, March 08, 2011 12:48 PM
> To: llvmdev at cs.uiuc.edu
> Subject: [LLVMdev] Using LLVM to convert a language to another language
> 
> Hi,
> 
> Sorry for my newbies questions... but here it is...
> 
> My goal is to have a "shading language", it is very similar to
the "C"
> language but with special tokens.
> 
> I have a parser and a lexer done with Lex/bison, once I have
> lexed/parsed my "shading language" I must create 3 new
"source code" in
> OpenCL.
> 
> It mean that by example for the following :
> 
> shader matte(float Kd, color c)
> {
>   Ci = diffuse(Kd * c);
> }
> 
> I have to create 3 methods in OpenCL, by example
> 
> 1 - matte_sampling
> 2 - matte_pdf
> 3 - matte_f
> 
> Each of theses method is in OpenCL (It is like C too).
> 
> So, I would like to use LLVM to create an optimized version of my
> shader code. But once I have the LLVM byte code... how can I parse it
> "easily" to create my shader code ?
> 
> NB: also, you can tell me if it is really a good idea !
> 
> Thanks
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

On Mar 8, 2011, at 12:47 PM, krys at polarlights.net wrote:
> Hi,
> 
> Sorry for my newbies questions... but here it is...
> 
> My goal is to have a "shading language", it is very similar to
the "C" language but with special tokens.
> 
> I have a parser and a lexer done with Lex/bison, once I have lexed/parsed
my "shading language" I must create 3 new "source code" in
OpenCL.
> 
> It mean that by example for the following :
> 
> shader matte(float Kd, color c)
> {
>  Ci = diffuse(Kd * c);
> }
> 
> I have to create 3 methods in OpenCL, by example
> 
> 1 - matte_sampling
> 2 - matte_pdf
> 3 - matte_f
> 
> Each of theses method is in OpenCL (It is like C too).
> 
> So, I would like to use LLVM to create an optimized version of my shader
code. But once I have the LLVM byte code... how can I parse it
"easily" to create my shader code ?
At a high level, this sounds like like a task better suited for Clang. I'd
suggest asking on cfe-dev about source-to-source rewriting.

-Jim
> 
> NB: also, you can tell me if it is really a good idea !
> 
> Thanks
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev