Post what you're working on and chat and help out others.
I'm making my frst real compiler, one that actually outputs assembly rather than targetting another high level language. It's just going to be a simple scheme compiler. I started on it today, wrote a parser/read and tested it on its own source code.
What are you all making?
Name:
Anonymous2014-05-19 20:34
I'm designing a sublanguage for Symta, which would allow using markov model as a mean of defining a function. Should be useful in generating random maps for Warcraft 2, because you can train with already existing maps and it will produce high-quality maps, especially if you use some fractal-like models, where top-level markov analyzes high level features (like number of islands), while bottom layers various details, like the shape of these islands.
I know it will work, because I've some experience with map generation (https://github.com/saniv/wc2gen), and the rule-based expansion I used in my previous project could be parametrized with markov models.
>>2 OP here. Symta is really cool, was super impressed with your compiler thing and the graphics looks great.
Name:
Anonymous2014-05-19 21:31
I'm playing around with a program to implement the recent proposal of Lispweb. I ended up writing a simple specification in the back of an envelope. I haven't made much progress and I don't think this will ever be the real Lispweb.
>>2-sama Is the map generation an iterative process? ie, does it make several passes, or just one pass, applying constraints each time? Any screenshots?
>>1-sama That sounds technically impressive and involved. Is your project an exercise, or do you intend to use your scheme compiler generally? Are you going to implement all the syntax-rules stuff, or just define-macro?
>>12-hime, I find the output of the above program very compelling, so that is my main motivation for exploring the possibilities of this architecture, and getting practice representing things this way. Interestingly, people who have worked on projects related to this have gone a bit nuts:
He has since baleeted everything and has disappeared from anything even vaguely computer or AI related. His Fluid Analogies Engine thesis is missing also. Wacky.
Name:
Anonymous2014-05-20 1:32
>>13 What the fuck? This guy made a Bong-tard problem solver, and now he thinks his AI is too dangerous? Are you sure he's not Mentishit?
>>22,23 When you upload to archive.org, it automatically creates those md5 files. They aren't part of what was uploaded. The relevant files are only copycatFiles.zip and Tutorial.zip.
Name:
Anonymous2014-05-20 21:37
>>17 It was good to talk to you too, scheme compiler-san!
Name:
Anonymous2014-05-21 0:58
I'm always working on like 10 different things at the same time, because I have a hard time using things made by anyone else. Since I use C, most of the time I can write everything I need from scratch, and that's what ends up happening.
I was writing an command line application for Omegle, and I wrote my own http library for it (surprisingly simple, the hardest part was handling chunk encoding), but Omegle implements the chat with JSON, so I had to write a JSON parser, and then I realized I needed something curses-like, but I hate curses and it wasn't sufficient for what I wanted, so I wrote my own curses library using XTerm Control Sequences, and then because of that I also need to write my own input library using termios and ioctl for the framebuffer...
A project that could have taken a few days is taking ages because short of writing the OS and compiler itself I'm writing everything from zero.
I wrote my own curses library using XTerm Control Sequences, and then because of that I also need to write my own input library using termios and ioctl for the framebuffer...
I really, really want to see this.
Name:
Anonymous2014-05-21 1:09
>>27 I'm not really done with it yet, but here's some alternatives I found when I first started trying to figure out how it worked:
http://asm.sourceforge.net/articles/rawkb.html -- An article on how to use raw input mode for the Linux framebuffer, it's in ASM but it can be converted into C since it just uses syscalls. The hardest part about this is that not every keyboard has the consistent scancodes, and this also ignores higher level keyboard configuration, such as regional keyboard layout settings (you're basically reading straight from the keyboard driver).
I have been reading about register coloring and wrote a program (in go) that generates random graphs for testing algorithms out. I will probably try out the Welsh–Powell algorithm or Brelaz. It looks like this might be useful too http://en.wikipedia.org/wiki/Sethi–Ullman_algorithm I'm just not sure how function calls inside arithmetic expressions come into play yet.
>>1-sam here, I've been stuck for a few days on register allocation by graph coloring. Finally figured out how to compile arbitrary expressions with only one register! Wrote up a prototype in haskal and tested it a bunch - it works perfectly. It's really fun
the idea is this, to compile X+Y into register r (where X and Y might be complex expressions too) you compile X into memory location [X] and then compile Y into register r, add r, [x]. To compile it into a memory location you just do a mov after.
This generates lots of local variables of course, I can minimize the number of locals by graph coloring but I'd like to maximize use of registers too - haven't figured that part out...
Name:
Anonymous2014-05-24 10:15
I'm making my Emacs config work on Windows, mastering logic programming and writing devops scripts for some start-up.
