[LLVMdev] SoC proposal: HLVM Python front-end

Reid Spencer rspencer at reidspencer.com
Wed Mar 21 16:13:12 PDT 2007

Hi Rooslan,

On Thu, 2007-03-22 at 03:03 +0500, Rooslan S. Khayrov wrote:
> Hi,
> I am senior student at Tashkent University of Information Technologies
> and I am highly interested in programming language design. I want to
> develop Python front-end for HLVM as a part of LLVM participation in
> Google Summer of Code project.

Sounds good.

> I am in long love with Python programming language for its simplicity
> and effectiveness. Last year I noticed ShedSkin Python to C++
> translator among SoC projects and idea of Python high-performance
> compiler fascinated me. Later I learned about LLVM compiler
> infrastructure project. It was looking as a great target for such
> compiler offering much more flexibility than C without pain of dealing
> with assembly/machine code generation. (Well, folks from PyPy were
> thinking in the same direction much earlier.) And finally I found the
> related HLVM project aiming to help in exactly what I want to do. Not
> as robust as LLVM at that time but very promising.

Work will be starting again on HLVM in a few weeks.

> (I hope it's all ok to suggest an HLVM-related project as I have
> already seen an interest in such project here in this mail-list.)

Yes, that's fine.
> HLVM roadmap sets Python support after Ruby and Scheme ones, but I
> think this is not essential. 

It is not. When the roadmap was designed, that was the interest of the
parties involved. The actual work will get done in the order that those
contributing want it to get done. I have no issue with putting Python

> Python "pros":
> — I know it better.
> — It's in wider use today, particularly in desktop applications where
> HLVM back-end features (high performance, lesser code size) will be
> more useful.
> — It's tightly related with my BS thesis, so I'll start working
> whether or not application will be accepted, but having Google to pay
> for it will be really nice :-)

All good reasons :)

> Chris Lattner in his recent presentation set an ambitious goal of
> developing common representation and type inference infrastructure for
> dynamic languages. I want to see HLVM evolving into such framework and
> will be very glad to make my own contribution on this.

I would like to see this too.

> Sorry for coming up with this a bit late but it's better than never.

No worries.

Your proposal will be considered.

Thanks for submitting it!

Best Regards,



More information about the llvm-dev mailing list