If you have to stop "early" please get through the simple DCE and LVN passes.
A solid basic-block level DCE & LVN optimization along with a judicious "compile time evaluator" (simple constant folding) will let you produce code at around 80% of the level of LLVM.
Obviously, you're not going to get the magic that comes of LLVM's scalar analysis pass, but you're not going to be embarrassed with your performance, either.
I'm interested. I'm writing a compiler-interpreter now but my broad compiler-construction chops are spotty and going through this systematically might be useful.
I would love to do so! I'm a masters student who has been wishing there was an advanced compilers course in my program, there's only the undergrad level one, so this is perfect! Let me know how I can get in contact.
I’m interested. I may not have the most relevant background, though — I’m a data scientist with an econometrics education. The topic really does interest me, though.
Okay, I just created this on discord https://discord.gg/RBPmmdWg. Considering it might be more accessible for people. Wondering if I should post this out of the thread.
I am a member of one of the projects that got awarded. We desperately are in need of better infrastructure for our development and testing purposes. Also we plan on collaborating face-to-face the coming year and the award definitely takes care of the expense. One of our team members is going to work full time on the project and we will be able to support her for at least a couple of months. We are all so excited! :)
Btw, I am a part of the project PeARS.
Our github account is here: https://github.com/PeARSearch/
We are not yet release ready. We plan on making our first release later this year. Please feel free to hangout at our IRC channel: #pears on Freenode.
We are at a very early stage of development and we have got a lot of work to do. Any kind of contribution would be highly appreciated.
Thanks, very exciting project. I already made another post in this comments section before reading this (I was that excited :P). I see there's plans for a FF-plugin which is great news. My mind is a little blown as I have never really thought about P2P-search as a possibility. Less centralized power over search is a great and important cause (imo).