Start of a new day. I'll get that release ready; shouldn't spend too much time on it. As compilers go, I doubt there will be a lot of interest in a pre-release of a bootstrap compiler.
Removed most debug prints from the code. Left the type inference debug prints since I'll work on that for the coming period. Removed parsing of series statements, since it looks like it actually increases the number of bugs made by programmers (me).
It takes some time to build a proper bootstrap. Also, I guess I should first use it myself a bit before releasing it. If anything, the gain will be that at least I will be using nicely numbered editions of the bootstrap compiler. There are a number of small things in it which I want to fix anyway before rewriting the typing code.
Monday, August 16, 2010
Sunday, August 15, 2010
Log 081510
I decided that the biggest problem hindering further development at the moment is just a good chair.
Worked a bit on the Google site. Looked at name registrars. Since when do I need to become a valued reseller? I just want a simple redirect for a few bucks?
Just before I went with the wrong registrar, found out it can be done easy within Google apps. So, did that, registered "hi-language.org" (Would have preferred hic.org, or something shorter, but -ah well,- good enough.)
Tinkered around for a few hours to get the site to become publicly viewable, where's the "make this site public for all meta-button?", but that worked out. Filling the site with content.
End of day. Site is filled. Not entirely happy with it, could have been more formal at points and clearer at other points, but I guess it'll just do. Next task: clean-up the bootstrap compiler and build a release.
As far as Google Apps/Sites goes. There are a few quirks, and blogger seems to handle cosmetics better than Sites at the moment, but to get a website up and running in under a day, cool! Kudos to Google.
Just before I went with the wrong registrar, found out it can be done easy within Google apps. So, did that, registered "hi-language.org" (Would have preferred hic.org, or something shorter, but -ah well,- good enough.)
Tinkered around for a few hours to get the site to become publicly viewable, where's the "make this site public for all meta-button?", but that worked out. Filling the site with content.
End of day. Site is filled. Not entirely happy with it, could have been more formal at points and clearer at other points, but I guess it'll just do. Next task: clean-up the bootstrap compiler and build a release.
As far as Google Apps/Sites goes. There are a few quirks, and blogger seems to handle cosmetics better than Sites at the moment, but to get a website up and running in under a day, cool! Kudos to Google.
Saturday, August 14, 2010
Log 081410
Finished the html snippet tool, it works. Read some of the comments on the P!=NP proof, there's no separation in his construction. That's all secondary.
Real questions are: What licence? What to publish? Everything, or just the C sources of the bootstrap?
Real questions are: What licence? What to publish? Everything, or just the C sources of the bootstrap?
Friday, August 13, 2010
Log 081310
Didn't do a lot except for tinkering with a mock-up for a new website. Which immediately gave a new problem: How and where to host? Static website, dynamic website, what software?
This is not something I want to spend a lot of time on. Looked at Google pages, maybe that's a good answer?
So, I tried it out. I, of course, couldn't get the intended design into one of Google's templates, but that might have been a good thing. So, working on the Google site for now.
Also, Google doesn't support uploading or using of javascript. So, I am writing a simple translator application in Hi to emit HTML snippets from the lexer. The compile, of course, failed somewhere.
Hmpf, I am going to rewrite each and every function till it's so painstakingly obvious and fast until I drop dead somewhere.
(It turns out I should keep better track of which compiler I am using. It parses, but doesn't support, series of statements like "E0;E1" yet, I use `unnamed' lets like "_=E0;E1" for that usually. I actually think I am going to get rid of series statements for the moment, it is convenient but clutters the bootstrap. Exception throwing works.)
This is not something I want to spend a lot of time on. Looked at Google pages, maybe that's a good answer?
So, I tried it out. I, of course, couldn't get the intended design into one of Google's templates, but that might have been a good thing. So, working on the Google site for now.
Also, Google doesn't support uploading or using of javascript. So, I am writing a simple translator application in Hi to emit HTML snippets from the lexer. The compile, of course, failed somewhere.
Hmpf, I am going to rewrite each and every function till it's so painstakingly obvious and fast until I drop dead somewhere.
(It turns out I should keep better track of which compiler I am using. It parses, but doesn't support, series of statements like "E0;E1" yet, I use `unnamed' lets like "_=E0;E1" for that usually. I actually think I am going to get rid of series statements for the moment, it is convenient but clutters the bootstrap. Exception throwing works.)
Monday, August 9, 2010
Log 080910
Nice date. Not doing a lot, need a push. I printed the lot and some code just looks wrong and rushed. Some refactoring and some rewriting is necessary. But, can't say I am really up to it at the moment.
Stats on the source. Two hundred and fifty pages source code, fifty pages system library and utilities, seventy five pages AST and parsing, fifty five pages semantical analysis, sixty pages conversion to code, and about ten pages handling modules, options and the main routine.
Stats on the source. Two hundred and fifty pages source code, fifty pages system library and utilities, seventy five pages AST and parsing, fifty five pages semantical analysis, sixty pages conversion to code, and about ten pages handling modules, options and the main routine.
Saturday, August 7, 2010
Log 080710
Definitely full conformance first, which means I want the whole bootstrap to be accepted cleanly with type checking turned on. On some modules, I am forced to turn the type checker off because of: 1) A bug in kind-checking multiple instantiations during typing (no idea, the sources look correct). 2) Small let-definitions not being accepted by the compiler because when generalized they don't type check because of a lack of type information on the values worked upon. The latter is not a bug, but just means I am probably better off not generalizing let definitions.
Friday, August 6, 2010
Log 080610
I took a few days off after the bootstrap. Not sure where I am heading from here. Perfomance is an issue, so is full conformance to spec. It makes the most sense to just finish most of it (add support for literal longs, floats, doubles, solve a logical error in exception throwing) so I can get rid of printing most of the diagnostic information.
The logical error in exception throwing is simple. Throwing constants work, but for some reason calculations in the body of the thrown exception are not run fully to completion, meaning I probably garbled something there. (Ticked that off I think, looks like I forget to supply the exception handler inside throw statements.)
I tried to optimize ackermann by inlining some functions which doubled its speed. So, my assumption was right, most time is spend setting up thunks (of which it sets up too many at the moment) and making libffi calls (which aren't memoized a the moment).
Since the compiler spends most of it time printing stuff at the moment, where every char is a libffi call, guess it's correct exception support followed by memoization first.
The logical error in exception throwing is simple. Throwing constants work, but for some reason calculations in the body of the thrown exception are not run fully to completion, meaning I probably garbled something there. (Ticked that off I think, looks like I forget to supply the exception handler inside throw statements.)
I tried to optimize ackermann by inlining some functions which doubled its speed. So, my assumption was right, most time is spend setting up thunks (of which it sets up too many at the moment) and making libffi calls (which aren't memoized a the moment).
Since the compiler spends most of it time printing stuff at the moment, where every char is a libffi call, guess it's correct exception support followed by memoization first.
Subscribe to:
Posts (Atom)