7 ms·
A Tourist’s Guide to the LLVM Source Code
- the_duke 10y ago100+ points, first place on front page, and not a single comment. Now that's rare. I assume everyone is excitedly browsing https://github.com/llvm-mirror/llvm https://github.com/llvm-mirror/llvm? ;) ( the official repo at http://llvm.org/git/llvm http://llvm.org/git/llvm is down)
- tomcam 10y agoGuilty as charged! This is so so welcome to an old school compiler guy
- pyvpx 10y agoas someone who has never written a compiler but has an intense interest in doing so, this is also very welcomed!
- tomcam 10y agoNow is is a great time to write your first compiler. The wealth of tools and docs these days is breathtaking. Give it a go!
- csmattryder 10y agoI bet you have, but if not, do yourself a favour and grab a copy of "Compilers: Principles, Techniques and Tools" aka "The Dragon Book". Still got my copy from uni, dog-eared and sticky-noted throughout.
- tom_mellior 10y agoI recall the classic 1986 edition of the Dragon Book spending waaaay too much time on boring details of lexing and parsing. I think it turned many people off the interesting aspects of compilers. It also had an interval-based formulation of data-flow analysis that nobody uses, possibly because it's (a) more complex than worklist iteration, and (b) not applicable to irreducible control flow graphs. Are newer editions better and more focused on interesting solutions to interesting problems?
- rwallace 10y agoTry the Tiger Book instead (a.k.a. 'Modern compiler implementation in C', or there are Java or ML versions if those are more to your taste). It has a better balance: more to the point treatment of lexing and parsing, leaving more space for a more comprehensive treatment of optimization and code generation.
- pjmlp 10y agoI second it, even if the book has a few issues here and there, it was one of the best books I read on the subject.
- astrange 10y agoI recommend making sure to avoid reading this book, it spends all its time making mountains out of molehills in the easy parts and leaves out the hard parts. How did they write so many pages about text parsing? The books by Appel and Louden (forgot the names) are good.
- akavel 10y agoI have a theory that since the introduction of "favorites" on HN, people more often fave something as a "Read Later [Totally Some Day, Really, Only Just... Not Today... But I Will Soo Come Back To It]" bookmark. Thus "lists of 100 interesting books", "algorithms you must know", "write yourself an OS in 99 days", etc. etc., I feel they have tendency to behave like that. (The theory is based partially on what I observed looking at myself ;). But please note it's totally just a loose theory, I don't have data or whatever, nor I'd judge this as bad or good. Only observing with curiosity.
- pjmorris 10y agoAnec-data: I favorite things as a "Read Later [Totally Some Day, Really, Only Just... Not Today... But I Will Soo Come Back To It]" bookmark. Well, mostly. I do read some percentage of the articles and comments as I go, and I do go back to see updates to the comments... and to, every once in awhile, read the fave'd article. I feel like I ought to have a side project that uses my HN favorites, pinboard links, comments, etc to seed a recommendation engine of 'other stuff you should probably read'.
- ChickeNES 10y ago> I feel like I ought to have a side project that uses my HN favorites, pinboard links, comments, etc to seed a recommendation engine of 'other stuff you should probably read'. I would totally use it. I considered writing something myself, but I have too many on-going side projects as it is unfortunately.
- pag 10y agoI always use this: https://code.woboq.org/llvm/llvm/ https://code.woboq.org/llvm/llvm/
- nrjdhsbsid 10y agoAwesome overview! The author threw a hilarious quip in at the end too: Target: the processor-specific parts of the backends live here. There are lots of TableGen files. As far as I can tell, you create a new LLVM backend by cloning the one for the architecture that looks the most like yours and then beating on it for a couple of years.
- gravypod 10y agoI've heard this many times that llvm targets are really hard to do. I don't really understand why. The code has already been parsed, optimized, and cleaned into a presumably clean format. What is the issue? Is the framework provided by llvm just crappy?
- david-given 10y agoA while back I got about half of an LLVM backend for the VideoCore IV done (since completed by someone else). I eventually ground to a halt due to intractable bugs; later, someone picked it up and completed it. This is from years ago, mind, but my recollection of the pain points were: - enormously steep learning curve; very little of this layer is documented and there's a lot of concepts hidden behind layers of abstractions hidden behind other layers of abstractions, all of which you need to understand, but they're all cyclic. There's nowhere to start; everything you look at depends on your understanding of something else. - very hard to debug --- the compiler's main reaction to things going wrong is to seg fault, which requires debugging through compiler internals (and understanding the compiler internals) to fix. - holes in the tools. After generating the IR node tree, the main piece of machinery in the compiler is a pattern matcher which finds machine code instructions for sets of nodes in the tree. This is automatically generated by tablegen, but there are some patterns that it can't match, which means that along with the tablegen matcher you also need to write a manual matcher in C++ for the other patterns, which is very verbose and difficult to keep track of. Also, the tablegen matcher is pretty brittle and has a tendency to either hang or match the wrong pattern if you don't get it just right. Understanding why it's doing what it's doing is non-trivial. The main problem, though, is that the pool of knowledge of how to write targets is restricted to a very small number of people --- it's all what an ex-boss of mine used to refer to as 'tribal knowledge'. So that when you run into a problem, you're reliant on finding one of these people who has enough time or incentive to help. Since the problems you run into tend to be complex, helping is hard, so getting good help isn't easy, and so on. I should add that gcc suffers from exactly the same set of problems, except possibly a little more so. gcc's get better target layer documentation, but the LLVM community is much more helpful; gcc is simpler, but LLVM I think has the edge of good design. They're both much of a muchness. It's frustrating, as compiler backends aren't that hard. There should be a compiler somewhere which combines easy porting with adequate code generation[1], for people who don't want to spend the resources needed for an LLVM or gcc port. [1] I'm actually working on this.
- zump 10y agoOnly problem is the CamelCase is incredibly verbose.
- swah 10y agoThat's critical for enterprise adoption! More seriously, I guess some "design patterns" really ought to appear in a project this big..
- eliben 10y agoI wouldn't call CamelCase naming a design pattern, but I agree that a large project should stick to consistent coding guidelines - which LLVM does [http://llvm.org/docs/CodingStandards.html http://llvm.org/docs/CodingStandards.html] Any particular style grows on you with time. It's more important to make sure the code is internally consistent.
- swah 10y agoSorry, I didn't realized the op was really criticizing the casing, I though it was about AbstractFactoryGenerator kinds of names.
- omouse 10y agoThis is the kind of doc that should be written about work projects as well. I always love seeing great docs that make the job of understanding a codebase way simpler.
- crb002 10y agoI've been putting it off, but I am definitely learning enough LLVM IR to code the 57 exercises for programmers. Anyone interested in banging out the LLVM IR solutions to http://rosettacode.org http://rosettacode.org? It looks like a few exist, but not many: http://rosettacode.org/wiki/99_Bottles_of_Beer/Assembly#LLVM http://rosettacode.org/wiki/99_Bottles_of_Beer/Assembly#LLVM
- nuclx 10y agoRegehr's blog is an invaluable source of practically relevant SE knowledge.