Hello fellow Racketeers, during the spring and summer I worked on a side project to test what can be done in pure Racket when it comes to raster graphics rendering. You can imagine it like demo or game programming before the advent of GPUs.
It turned out that quite a lot can be achieved and it also turned out that Racket performs really well when it comes to some hard number crunching. And what else is computer graphics than just a number crunching? As I consider this experiment basically completed, I release the source code[1] to the public under the usual dual Apache/MIT license. A short (25s) video in action can be seen on the project "website"[2]. For those wondering about the tut25 naming - the project started as an attempt to rewrite all Denthor's VGA Trainers published between 1994 and 1996. There were 21 tutorials, I added four. I am publishing only the final result - the 25th "tutorial". If I have some spare time in the future, I'll write a series of articles about the journey from drawing the first pixel in the window to full 3D triangle rasterizer with alpha blending, perspective-correct texturing, z-buffering and Goraud shading. And for those wondering "why?" - there are actually more answers: 1) Why not? 2) Projects like this stress-test Racket implementation and therefore help its development (more on that below). 3) I want more people to start using futures for improving performance without sacrificing code readability. 3b) Also I don't want to hear from the people in the core team "Oh, you are probably the only one seriously using futures" ;-) To inspire more people to start seemingly crazy projects like this, here is a list of things that are now better in current Racket version as a side-effect of this project (and yes, Matthew wrote all the patches): * futures versus GC in BC - no more crashes[3] * futures and the write barrier in BC - no more freezes[4] * fsemaphores - no longer can go negative (actually the bug was elsewhere)[5] * safe and unsafe fl->fx unification[6] * internal stack overflow fixed[7] * box CAS operation troubles with futures[8] The code comes however with a big red warning: some parts are as ugly as they can get. On the other hand, there are some nice language features that I consider worth investigating further. Mainly: * using syntax macros to create procedure templates and then with another layer of syntax macros instantiate those for specific task. See for example the define/provide-draw-proc-variants and template-clear in rcge/drawing.rkt. Yes, they are unhygienic - but just very slightly... * context nesting using parameters (although I personally dislike the need to wrap everything in so many parameterize layers) - see any of rcge/*-run.rkt files for example. * using syntax macros to provide hints to the compiler. The best example of this is the matrix multiplication in rcge/algebra.rkt - define-mat4*mat4!/let. Yes, definitely not the best practice in general, but combined with unboxed flonums on CS, it improved the performance by an order of magnitude compared to simple loop. * define-futurized (rcge/futures.rkt) - in an ideal world, a generalized version of a macro like this should be able to parallelize computation-heavy algorithms with little to no effort for the programmer. Things NOT to do: blit-to-cairo-data*-u16vector-hack - rcge/blit.rkt is not for the weak. But I like the fact, that it is possible to interact between different layers of the program, compiler and JIT. Thank you for reading this far and if you have any questions, I'll be more than happy to answer those. Cheers, Dominik [1] https://joe.cz/rcge/ [2] https://gitlab.com/racketeer/rcge-tut25 [3] https://github.com/racket/racket/issues/3145 [4] https://github.com/racket/racket/commit/fae20b4b8904bfcfba9808f56744950b049b6afd [5] https://github.com/racket/racket/commit/1117392cb5524a2ce150384d405fb2a81ef47daa [6] https://github.com/racket/racket/commit/42cb80bc70e1fb97dabd1e0a2dff5d93257b1f40 [7] https://github.com/racket/racket/commit/265c9eaa57661a045feed85e1db45029aa9de7e2 [8] https://github.com/racket/racket/commit/0af11de62b2eaa1f4f0977794a4dc9dcacabd945 -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/a46338ba-51e1-e3f4-dba2-99d319b0b503%40trustica.cz.
