hi all, my name is Jakob Praher. I am located in Linz, Austria. Currently I am finishing my master thesis, which is about developing an aspect oriented dynamic bytecode instrumentation framework for program analysis on top of LLVM. I am very interested to work in that area as a PHd student. From your website I know you are also working on an LLVM backend. PyPy seems to be a European effort, so my naive question is, whether you have any PhD projects in this area or could point me to some further information. I would be very interested to hearing from you. Thank you Jakob
Jakob Praher <jp@hapra.at> writes:
hi all,
my name is Jakob Praher. I am located in Linz, Austria. Currently I am finishing my master thesis, which is about developing an aspect oriented dynamic bytecode instrumentation framework for program analysis on top of LLVM. I am very interested to work in that area as a PHd student. From your website I know you are also working on an LLVM backend.
I'm not _entirely_ sure what you are saying here. We have interests in AOP and interests in LLVM but they are not particularly related, whereas your interests seem to be intertwined. But it's late and I am suffering from a stressful trip, so maybe the confusion is just at my end :)
PyPy seems to be a European effort,
Only by accident, mostly...
so my naive question is, whether you have any PhD projects in this area or could point me to some further information. I would be very interested to hearing from you.
Well, we certainly hope that there are interesting lines of investigation in PyPy. To be more specific, we probably need you to be more specific about your interests too. If you are asking "do we have pots of cash lying around to fund PhDs", then I'm pretty sure the answer is "no". PyPy's funding from the EU was not structured in this sort of way (AIUI). Cheers, mwh -- (FREE|OPEN) BSD: Shire horse. Solid, reliable, only occasionally prone to crushing you against a wall and then only because you've told it to without knowing. -- Jim's pedigree of operating systems, asr
Hi Jakob! Jakob Praher <jp@hapra.at> writes:
my name is Jakob Praher. I am located in Linz, Austria. Currently I am finishing my master thesis, which is about developing an aspect oriented dynamic bytecode instrumentation framework for program analysis on top of LLVM. I am very interested to work in that area as a PHd student. From your website I know you are also working on an LLVM backend.
Well, yes. But an LLVM backend in PyPy lingo is different from an LLVM backend in LLVM lingo :-). The latter would be a backend to LLVM that targets a certain CPU architecture. But it is really the other way round: we have a backend that produces _LLVM_ code out of our own IL. No clue about the Ph.D., sorry. Cheers, Carl Friedrich
Carl Friedrich Bolz schrieb:
Hi Jakob!
Jakob Praher <jp@hapra.at> writes:
my name is Jakob Praher. I am located in Linz, Austria. Currently I am finishing my master thesis, which is about developing an aspect oriented dynamic bytecode instrumentation framework for program analysis on top of LLVM. I am very interested to work in that area as a PHd student. From your website I know you are also working on an LLVM backend.
Well, yes. But an LLVM backend in PyPy lingo is different from an LLVM backend in LLVM lingo :-). The latter would be a backend to LLVM that targets a certain CPU architecture. But it is really the other way round: we have a backend that produces _LLVM_ code out of our own IL.
Yes :-) I just wanted to point out that I have some knowledge of the LLVM infrastructure and that I was using the JIT for my own project quite substantially. While there are many areas where still need a lot of work to understand, I am very interested in projects like yours and so I just wanted to ask. LLVM lacks symbolic information (hence lowlevel), which makes it impossible to be used as a first class IL for a project like pypy. I will have a look at your IL.
No clue about the Ph.D., sorry.
Never mind. Thanks for your reply.
Cheers,
Carl Friedrich _______________________________________________ pypy-dev@codespeak.net http://codespeak.net/mailman/listinfo/pypy-dev
hi Michael, Michael Hudson schrieb:
Jakob Praher <jp@hapra.at> writes:
hi all,
my name is Jakob Praher. I am located in Linz, Austria. Currently I am finishing my master thesis, which is about developing an aspect oriented dynamic bytecode instrumentation framework for program analysis on top of LLVM. I am very interested to work in that area as a PHd student. From your website I know you are also working on an LLVM backend.
I'm not _entirely_ sure what you are saying here. We have interests in AOP and interests in LLVM but they are not particularly related, whereas your interests seem to be intertwined.
I only wanted to give you some information, what I am doing right now. I am also interested in dynamic runtime systems. I have some experience with self and prototype based oo systems. I also had some experiences with Jalpanero (now JikesRVM), which is also selfhosting like PyPy. What I mean is that I am flexible. For me it is important to do open source research and that I could gain some experience for implementing large runtime systems.
But it's late and I am suffering from a stressful trip, so maybe the confusion is just at my end :)
PyPy seems to be a European effort,
Only by accident, mostly...
That is interesting :-)
so my naive question is, whether you have any PhD projects in this area or could point me to some further information. I would be very interested to hearing from you.
Well, we certainly hope that there are interesting lines of investigation in PyPy. To be more specific, we probably need you to be more specific about your interests too.
See above. My biggest concern is whether there could be some real colaboration with a university. It is quite hard to do this kind of research in Linz. Mostly because of a lack of support from the university side.
If you are asking "do we have pots of cash lying around to fund PhDs", then I'm pretty sure the answer is "no". PyPy's funding from the EU was not structured in this sort of way (AIUI).
No I just wanted to say that it would be very interesting to work in an area like dynamic object oriented runtimes. But there are many things why I am asking this out loud: * Is there any university involved in this project? * What area of PyPy needs some research in the size of a PhD project? Granted it is somewhat ignorant not to come up with an answer to the second question before writing to this list. But actually I wanted to hear some opinions from you. Anyways Linz is not the best place for hard core open source research. Before working on anything myself again, you have to understand that I have to do some questioning. -- Jakob
Hi Jakob, I think we can safely say that the Pypy project contains several areas that are suitable for PhD level research. Just to mention a few: - Modularisation of the source language, allowing expansion and contraction of the available syntax and semantics. - Adaption of the platform to handle multiple source languages. -Use of the Pypy architecture for distributed computing, multiprocessing and handling of security issues at interpreter level. - Various aspects of code generation including translation, JIT specialisation, garbage collection and backend production. - Pypy integration with existing platforms, like Java and CLI. måndag 03 april 2006 09.55 skrev Jakob Praher:
so my naive question is, whether you have any PhD projects in this area or could point me to some further information. I would be very interested to hearing from you.
Well, we certainly hope that there are interesting lines of investigation in PyPy. To be more specific, we probably need you to be more specific about your interests too.
See above. My biggest concern is whether there could be some real colaboration with a university. It is quite hard to do this kind of research in Linz. Mostly because of a lack of support from the university side.
If you are asking "do we have pots of cash lying around to fund PhDs", then I'm pretty sure the answer is "no". PyPy's funding from the EU was not structured in this sort of way (AIUI).
No I just wanted to say that it would be very interesting to work in an area like dynamic object oriented runtimes. But there are many things why I am asking this out loud:
* Is there any university involved in this project? * What area of PyPy needs some research in the size of a PhD project?
Heinrich Heine University of Duesseldorf is a member of the project, with professor Leuchel being formally responsible. In practice, it is Armin Rigo and Michael Hudson who are financed to be working on Pypy. As of this week Carl Friedrich Boltz will be doing his Bachelor of Science there, focusing on Pypy. While professor Leuchel is a very open minded person, who seems to enjoy giving people the liberty to pursue the kind of research they are interested in, I have no clue as to what the policy of the CS department or the university are concerning the acceptance of PhD students. I hope what I have said will help you get started on your own investigation. In the meantime, I hope you will have time to get acquainted with the code base. You are also welcome to join one of our sprints, where you will get a chance to collaborate with the developers on specific issues. The most convenient place for you in the reasonably near future will probably be the sprint right after Europython at CERN in Switzerland. We do have an open sprint in Japan before that, but it is a long way to travel and we will probably have as many newcomers as we can cope with. Best of luck Jacob Hallén
Hi Jacob, thank you for your support. Looks very interesting to hear that. What are you currently working on? AFAICT you are employed by strakt.com. Are you working on pypy as a research project? Jacob Hallen schrieb:
- Modularisation of the source language, allowing expansion and contraction of the available syntax and semantics.
- Adaption of the platform to handle multiple source languages.
That is insteresting, especially when supporting other dynamic languages.
-Use of the Pypy architecture for distributed computing, multiprocessing and handling of security issues at interpreter level.
Ah yes. I recently attended a guest lecture by Michael Franz. He is a professor at University Irvine in California. They are implementing MAC based security labels for virtual machines. He managed to bring the pefromance loss of labeling down to 6 % of the whole execution time. Given the fact that SeLinux for instance already supports MAC, MAC can be an interesting concept to bring to virtual machines as well.
- Various aspects of code generation including translation, JIT specialisation, garbage collection and backend production.
Interesting.
- Pypy integration with existing platforms, like Java and CLI.
Could be interesting too. Especially Java plans on integrating an invokedynamic instruction. This makes it possible to design more efficient code generators. But I am not really 100 % sure, how this will look like.
Heinrich Heine University of Duesseldorf is a member of the project, with professor Leuchel being formally responsible. In practice, it is Armin Rigo and Michael Hudson who are financed to be working on Pypy. As of this week Carl Friedrich Boltz will be doing his Bachelor of Science there, focusing on Pypy.
Thats interesting to hear.
While professor Leuchel is a very open minded person, who seems to enjoy giving people the liberty to pursue the kind of research they are interested in, I have no clue as to what the policy of the CS department or the university are concerning the acceptance of PhD students. I hope what I have said will help you get started on your own investigation.
Oh great. Thanks for the pointers. The problem I am facing is that PhD slots are pretty scarce and so there are often many people already waiting for one :-)
In the meantime, I hope you will have time to get acquainted with the code base. You are also welcome to join one of our sprints, where you will get a chance to collaborate with the developers on specific issues. The most convenient place for you in the reasonably near future will probably be the sprint right after Europython at CERN in Switzerland. We do have an open sprint in Japan before that, but it is a long way to travel and we will probably have as many newcomers as we can cope with.
I will surely look at the project code base, no matter how this turns out. Thank you.
Best of luck
Jacob Hallén
participants (4)
-
Carl Friedrich Bolz -
Jacob Hallen -
Jakob Praher -
Michael Hudson