Thursday, July 5, 2007

Committed to LLVM

Yesterday I got accepted to the inner circle of LLVM developers by becoming a committer. For now I have very humble plans, such as periodically plowing through the sources and sanitizing a bit here and there. This may change to a more active role, but probably not in the form of contributing to llvm-core.

At least version 2.0 already does contain some stuff from me (improved memory efficiency).

My first contact to LLVM happened when I started a new backend for the d2c compiler. I found out that the Dylan language and LLVM would be a good match, and that d2c was modular enough to nicely marry these two worlds. I did produce some llvm output, but got distracted soon and the project never became functional. Ah well...

1 comment:

pet-ro said...

What were the main issues why you stopped the d2c / llvm project? (I expect time and no collaborators.)

LLVM is a Apple supported project. Dylan once were once also Apple supported. Have you ever discussed with LLMM people to check the value of LLVM for dynamic languages via dylan?

Did you also evalute a opendylan / llvm project?

I would expect that OD's HARP architecture could simplify a Dylan frontend project for LLVM.
I would appreciate a comment related to a HARP/LLVM form you d2c/LLVM project.

May be you can convince someone from the LLVM people to comment / evaluate this, too.

As the LLVM project is interested to have also comfortable frontend -- currently there is a new c-frontend project a LLVM - a OpenDylan/LLVM would make sense.

In my opinion a OpenDylan/LLVM project is more easier than a d2c/LLVM project. But I can not proof it. May be your tasklist (see below) can show what is necessary to proof this.

Wouldn't a OpenDylan/LLVM not a greater value than a d2c project because of the IDE (including a debugger)?

I would a appreciate if you come up with a tasklist for a "Dylan LLVM frontend project"?
Why not hacking such a "Tasklist: Dylan LLVM frontend" page to the wiki at