PyPy sprint announcement: PyCon 2006, Texas, Feb 27st - March 2nd

Hello LLVM-ers,

The next sprint of PyPy will be held in Dallas, Texas, at the PyCon
conference. Most of you know about the LLVM back-end of PyPy. So far,
we use mostly the static compilation features of LLVM, but as we are
progressing on the JIT side we are considering starting sometime soon
working on just-in-time machine code generation backends. Clearly, LLVM
might prove to be a good target for that.

The full announcement follows. Although LLVM-JIT is not explicitly
mentioned, as it is still an experimental idea for us, feel free to show
up. Not all PyPy core developers will be present, so we will focus on
whatever the people present are interested in. There will be people
knowledgeable in both PyPy and LLVM around, ready to help newcomers to
work on this topic.

                             - * - * -

Post-PyCon PyPy Sprint: February 27th - March 2nd 2006

Hi Armin,

The next sprint of PyPy will be held in Dallas, Texas, at the PyCon
conference. Most of you know about the LLVM back-end of PyPy. So far,
we use mostly the static compilation features of LLVM,

How are you using the static compilation features? Are you able to do this via API calls or generating code that will be processed by tools?

Thanks,

Marcel

Hi Marcel,

How are you using the static compilation features? Are you able to do this via API calls or generating code that will be processed by tools?

For static compilation, currently pypy's genllvm emits ll code and the cfe is also used for external system calls, where the c code is shared (or borrowed) from pypy's genc. There are some half baked ideas wrt to using the api instead of emitting ll code. Also there is work going on in rctypes (see Armin's original mail) which may eliminate the need for the front end compiler.

Cheers,
Richard