EuroLLVM BoF session: Compilers in education

Hi all,

I'm organizing a BoF session during the upcoming EuroLLVM developers meeting. As the subject of this message already shows, this session will be on compilers in education. I'm currently looking for both participants to the discussion and input for the actual program of the session. I've already got some ideas which I'll introduce below.

At our university, we mostly come from a hardware and processor design background. In the past we managed to design all kinds of crazy architecture ideas and, since we could program them by hand for our demonstration cases, we usually assumed that compilers could just 'do the trick'. However, over the years we learned that this usually wasn't the case. Looking around for what was actually being taught in compiler design we found that the course at our university had been teaching mostly the classical frontend parts (parsing etc) and was discontinued several years ago. We decided it was time for us to step up and start teaching our students again what is and isn't possible in compilers. This time, also with extensions into the layers that relate more closely to the hardware so that the course would match our processor architecture backgrounds more closely.

In general, the contents of this course are more or less as follows:
  - Backend organization (optimizations, lowering, scheduling, register allocation)
  - Code optimization (code analysis, loop optimizations, auto vectorization, Polly)
  - Using the compiler (writing optimizer friendly code)
  - Heterogeneous systems and high-level languages (OpenCL, OpenMP, Halide)

Looking further we also found that there aren't many courses that cover these topics (while we think that they should be interest to the companies hiring our students).

During this BoF, I plan to introduce the topics which we now cover in our course and the assignments that we give to our students to go with that.

My long term goal is to make most of the materials for this course available to interested parties. One idea was to transform this into some kind of MOOC structure when after we get the lectures tested on our own students (currently there are some 70 students attending this course). We've already automated much of the testing and grading for the exercises (though improvement is still possibly there).

So, my questions to you:
- Would you be interested in attending this session?
- For industry people:
   * what kind of skills would you like your future (or current) employees to have?
   * do you have any educational materials you would be willing to share?
- For other educators: what kind of topics do you cover in your lectures?
   * would you be interested in presenting something about your approach?
- And in general: are there any other things you think that should be discussed?

Ok, that's it for now. I've already heard from a few people that they would be interested but I'd love to get some more input!

Best regards,
  Roel Jordans

Yes, please. That sounds very useful.

Joerg

- Would you be interested in attending this session?

Yes, I’m looking forward to it.

- For industry people:
* what kind of skills would you like your future (or current) employees to have?
* do you have any educational materials you would be willing to share?
- For other educators: what kind of topics do you cover in your lectures?
* would you be interested in presenting something about your approach?

I’d be happy to talk briefly about the course that I teach, which is the Masters’ compiler course at Cambridge. I have some resources online:

The focus of the course is on real problems and techniques for late-bound dynamic languages (things a bit like JavaScript and parallel languages (things a bit like OpenCL C / CUDA). The students have to do three shortish exercises. The first is a small amount of analysis, the second two each involve one of the toy languages: typically something like adding something like polymorphic inline caching or type specialisation to MysoreScript (a milder, less bitter JavaScriptl-like language) and adding either automatic paralellisation or vectorisation to the cellular automata DSL (which has an mostly parallel abstract machine, but a reference implementation that is purely sequential). They then have to do a miniproject (80% of the mark, producing something in the structure of a research paper) on something real. One of my students will be at EuroLLVM presenting a poster about his miniproject, so he can talk about what he’s learned. In previous years I’ve had people do things like adding small integer support to Pyston, evaluating the LLVM code generation of barriers on ARM (EuroLLVM 2014), or optimising memory allocation for the PImpl pattern (EuroLLVM 2015).

David

Hi David,

Great that you're joining as well. I'd like to hear from your student about his experiences, that sounds interesting.

In our course we currently have three main exercises:
- With the AVR backend we have an exercise that fixes some bits of instruction selection and adds support for GCC's __builtin_avr_delay_cycles intrinsic so that the students can improve the code generation for the blinking LED example program.

- In the optimization layer we have an exercise with code analysis, estimating instruction-level parallelism by implementing a list scheduling algorithm at the IR level.

- Using the compiler to get a small image processing application to optimize efficiently. The main goal here is to get insight in the autovectorization capabilities of compilers and what needs to be done to allow the compiler to efficiently optimize C code.

I'll make sure to bring some examples also of these assignments. For us the assignments make up 40% of the mark and we plan to have an interactive exam for the remaining 60%, in parallel there's a few bonus assignments which can (in total) give up to 20% of compensation credits for those who ran into the steep learning curve of the assignments.

  Roel

- Would you be interested in attending this session?

Yes, I’m looking forward to it.

- For industry people:
* what kind of skills would you like your future (or current) employees to have?
* do you have any educational materials you would be willing to share?
- For other educators: what kind of topics do you cover in your lectures?
* would you be interested in presenting something about your approach?

I’d be happy to talk briefly about the course that I teach, which is the Masters’ compiler course at Cambridge. I have some resources online:

http://compilerteaching.github.io

The focus of the course is on real problems and techniques for late-bound dynamic languages (things a bit like JavaScript and parallel languages (things a bit like OpenCL C / CUDA). The students have to do three shortish exercises. The first is a small amount of analysis, the second two each involve one of the toy languages: typically something like adding something like polymorphic inline caching or type specialisation to MysoreScript (a milder, less bitter JavaScriptl-like language) and adding either automatic paralellisation or vectorisation to the cellular automata DSL (which has an mostly parallel abstract machine, but a reference implementation that is purely sequential). They then have to do a miniproject (80% of the mark, producing something in the structure of a research paper) on something real. One of my students will be at EuroLLVM presenting a poster about his miniproject, so he can talk about what he’s learned. In previous years I’ve had people do things like

adding small integer support to Pyston, evaluating the LLVM code generation of barriers on ARM (EuroLLVM 2014), or optimising memory allocation for the PImpl pattern (EuroLLVM 2015).

Hello,

This looks really interesting! Speaking for those of us that won’t be able to attend the session in person: it would be great if you could share what was discussed afterwards!

Hi Débora,

That shouldn't be a problem. There will be notes taken during the session.

Cheers,
  Roel

FYI, we intend to have notes collaboratively taken in etherpad and shared for all BoF sessions.

The etherpad links (to yet empty notes) are already available on the EuroLLVM web page with the BoF abstracts: http://www.llvm.org/devmtg/2016-03/#BoFsAbstracts

Kind regards,
Arnaud