Friday, October 22, 2010

PhD Thesis about PyPy's CLI JIT Backend

Hi all,

few months ago I finished the PhD studies and now my thesis is available, just in case someone does not have anything better to do than read it :-).

The title of the thesis is High performance implementation of Python for CLI/.NET with JIT compiler generation for dynamic languages, and its mainly based on my work on the CLI backend for the PyPy JIT (note that the CLI JIT backend is currently broken on trunk, but it's still working in the cli-jit branch).

The thesis might be useful also for people that are not directly interested in the CLI JIT backend, as it also contains general information about the inner workings of PyPy which are independent from the backend: in particular, chapters 5 and 6 explain how the JIT frontend works.

Here is the summary of chapters:
  1. Introduction
  2. The problem
  3. Enter PyPy
  4. Characterization of the target platform
  5. Tracing JITs in a nutshell
  6. The PyPy JIT compiler generator
  7. The CLI JIT backend
  8. Benchmarks
  9. Conclusion and Future Work

cheers, Anto


The Cannon Family said...


Eric van Riet Paap said...

Yes Anto, congratulations!

Lino said...

Impressive work, Antonio.


Anonymous said...

Very interesting stuff, still busily reading... could you write a short bibtex entry for citation? Thanks

glyph said...


(but when are we going to see it merged to trunk... ;-))

Antonio Cuni said...

thank you, guys :-)

@anonymous: here you can find the bibtex for the thesis, as wall as for other PyPy related papers:

@glyph: unfortunately, trunk has diverged a lot since the cli-jit branch, and merging is not an easy issue. There are also fundamental features that on CLI cannot be implemented as efficently as on x86. It's on my todo list, but no concrete plan so far :-(