Dissertation Status
  2023-01-11

things are looking brighter regarding my project. in the past days i posted my proposal on lobsters [1] and on compilers@lists.iecc.com [2] and recieved feedback from many people. some feedback was good, some was bad, what matters is that my idea reached more people. some of the people are very important in the field of compilers, for example David Chisnall, Nuno Lopes or Anton Ertl. it's very nice that i had the ocassion to bounce ideas with them. what is even better is that Nuno agreed to work with me on this project. he has very much experience of working with UB related issues in LLVM and his expertise will help the project a lot. only by chatting with him for a bit i realised that the project is more complex than i thought. many places in Clang/LLVM use UB to make assumptions that are easier for the compiler. for example clang on one hand uses UB to make assumptions, then it passes the information with LLVM and the IR also makes UB assumptions independent of the ones from Clang. finally, the backend for generating assembly code also uses this kind of assumptions, but not that much. it's hard to demonstrate that after a period of time in which i try to disable all ub from the compiler i can certainly say that no more ubs are to be encountered. the difficulty here is that the code base is *huge* and cannot be understood by a single person. what is more interesting is that if Clang/LLVM are written in C/C++ they might also use UB in their code. which means that we stumble upon the problem of trusting trust, which sucks. however, i entered in to many technical details for this post. what i intended to say is that i finally found a research coordinator that has experience in this field and is able to allocate more time to discuss issues related to the project. what's even cooler is that this spring Euro LLVM will be held in UK and if i have something to present until then i could go with my work there. that means that i will meet other people from this field and will talk to them about my plans. i need to work hard for this to happen. now the presentations for the dissertation in my faculty will become a simple formality as the work will become much more complex and interesting. [1] https://lobste.rs/s/b820uv/proposal_analysis_speedup_gain [2] https://sympa.services.net/sympa/arc/compilers/2023-01/msg00009.html