Played around with outlines in my engine today 
I suggest you guys try out the browser Vivaldi, itās pretty neat. Iām enjoying it so far.
Today I started implementing an ext2 driver for my OS.
Not java or (directly) gaming related, but Iām feeling good about it so Iāll post:
Some may remember the tiny little stack-based language I wrote a while back: https://github.com/BurntPizza/ghost
Iām finally making good progress on a more serious language Iāve been kicking around in my head for a while: https://github.com/BurntPizza/aleph
Itās a lisp, with a (hopefully, as itās not implemented yet) static type system.
Today I got my first macro implemented: the let form:
// "let a = 4, b = 6, and c = 8 in a + b + c"
(let (a 4
b 6
c 8)
(+ a b c))
// will return 18
Itās implemented (currently) by expanding into an anonymous function which is immediately called:
(anonymous functions are currently the only way to assign labels in a new scope, just like lambda calculus)
((fn (a b c)
(+ a b c))
4 6 8)
// will return 18
This in turn compiles to the following program:
---- interpreter::test::let_form stdout ----
SymbolTable:
| id | ident | kind |
|----|-------|----------|
| 12 | def | special |
| 13 | fn | special |
| 14 | do | special |
| 15 | + | special |
| 16 | a | var |
| 17 | b | var |
| 18 | c | var |
| 21 | fn$0 | fn |
Constant table:
[0] 4
[1] 6
[2] 8
Bytecode:
0: ldc 4
3: ldc 6
6: ldc 8
9: call fn$0
14: exit
fn$0:
15: save %2 // I don't have an optimization pass for
17: save %1 // pointless stack-shuffling like this yet
19: save %0 // or inlining for that matter
21: load %0
23: load %1
25: load %2
27: add 3
29: ret
I can prove it works because the following test passed: 
Now just to figure out how to do namespaces wellā¦
This is not Java, but Iām thinking of posting this as an update to my workshop instructorā¦ Any recommendations on how to improve on the next update?
QfY2QZFhWN0
If you want to play with designing a language, Racket people have some freely available books. Racketās solution is: https://docs.racket-lang.org/reference/Namespaces.html
Indeed. Iām also aware of (a possible, I forget if theyāre using it yet) racketās macro expansion solution, which I will be taking a thorough look at: https://www.cs.utah.edu/plt/publications/popl16-f.pdf
Aleph is definitely inspired by Clojure, but one thing I want to make sure I donāt wind up with is anything resembling Clojureās terrible namespaces: mutable (!) collections of Vars at runtime. Given the languageās advocacy of REPL-driven development and immutability, the fact that it is very easy to accidentally mutate your namespace into an erroneous state is ridiculous. The namespaces almost feel like they came from a poor fork of the language, even though theyāre core component.
Started a new project called as GwtOpenType which is a binding of the awesome opentype.js project. With this, Iām pretty sure I can enable the TTF support for my GWT backend of SilenceEngine.
My target is to release a snapshot in two days.
Ok, that didnāt work.
gotta love qemuās half broken BIOS memory mapā¦ my physical memory manager is working now, on to virtual memory manager and creation of malloc and free! yay finally real mem management with
userspace->kernel interaction.
edit:
As a proof of concept I tried to overwrite the actual assembly code section (where the kernel starts in memory, or at . = 0x00100000) and it actually worked hahaā¦ wherever I put this code, as soon as it executes, nothing else happens because I literally overwrote the actual code in memory. In reality, the processor is trying to execute ā0x80ā over and over and over again. I love coding.
uint32_t* memtest = &kernstart;
for(int i = 0; i < (&kernend - &kernstart); i++)
{
memtest[i] = 0x80;
}
edit: Some more cool stuff, if I instead overwrite with ā31C0ā which is
xor eax, eax
it doesnāt stop incrementing āiā, and overwrites into video memory in that short time it has, which results in some goofy stuff. At least I think this is whatās happening. AFAIK EDX is used for loops in standard compilers, but EAX obviously has something to do with it since itās overwriting vmem. Memory protection is important!
I re-created an old lighting engine of mine, this time optimizing the recursion with a stack. It was so much faster that I accidentally had it running every frame without any noticeable performance decline. Of course, the real (human) performance decline was when I originally used a queue and wondered why there were so many objects waitingā¦
Iāve tried my hand at isometric projection (and a little bit of pseudo-shading) in Java2D:
Iām as of yet unsure how Iād store textures to render on top.
Cubes make the depth hard to āseeā (the alignment of top with bottom of adj). Try changing the height.
@Roquen: I do agree, I also think that itās got to do with the fact that Iām really only drawing a grid until I make some textures to draw on top. Iām gonna do a bit of experimenting.
@Coldstream
[quote=āRoquen,post:4192,topic:49634ā]
I reckon you should try halving the height. Iāve already seen that in games (canāt think of any right now unfortunately) and it usually works pretty damn good. :point:
Sorting the isometric sprites the ācorrectā way is a hell, if you go for perfection youāll end up with a O(n^2) algorithm in the worst case or O(n * log n) algorithm eating up your mem in the best case.
How about height = (0.25~0.3)*width?
That picture is for 2 cubes (equal size in each dimension) isometric on left, dimetric on right. Different scales for āupā are possible for dimetric. Choose one you like to model a unit cube and see how it plays out for other objects.
It could be kinda cool to have a generalized axonometric playground. (Not me!)
I really donāt follow, esp n^2. I did a isometric engine way back and had no such problems. Also n log n is fast enough even on old hardware for pretty big lists. If its mostly sorted, many sorts tend to O(n).
Anything that has a consistent ordering should always be sortable in n log n worst case. (broken quick sorts not withstanding)