pouët.net

entity foo is
generic( d : integer );
port( ck, rs_n : in std_logic; x, y, z : in integer; v : out integer );
end entity;
architecture bar of foo is
type z_t is array(0 to d) of std_logic_vector(r-1 downto 0);
type y_t is array(0 to d) of z_t;
type x_t is array(0 to d) of y_y;
signal box : x_t ;
begin
process(ck, rs)
begin
if rs='1' then
for _z in 0 to d-1 loop
for _y in 0 to d-1 loop
for _x in 0 to d-1 loop
box(z)(y)(x) <= x*y*z;
end loop;
end loop;
end loop;
elsif rising_edge(ck) then
v <= box(z)(y)(x);
end if;
end process;

@torus: http://hilbert-space.de/?p=22 looks familiar :T

SIMD kicks ass for image filters. I wish the HTC Hero had a Neon-enabled ARM.. doing scalar processing of RGB565 data is bullshit :P

From "pytarl" a text adventure/rougelike dungeon crawler game I currently develop. If anyone cares, the dev site is over here. You can download a win32 version from here. :)

shmup-basics :)

private function gameLoop(e:Event):void
{
updateEnemies();
updatePlayerShip();
updatePlayerShots();
updateGameDisplay();
}

@mic - well. That hilbert-space thing is my blog :-)

OnSoL: Notice that the result of dx*presin increase by presin for every x-iteration; forward differencing it would be faster. The same goes for dx*precos. Basically, everything before ">>14)&63" can be turned into one add per pixel (or, two. one for _x and one for _y). Then notice how you can get rid of the dx-tracking - in fact, all you have left is the x-counter, x/y, x/y-deltas, and screen pointers. Next, turn your x-counter to count downwards, as that allows the compiler to have the sub-instruction update the zero-flag, giving the compare for free. Then you should start unrolling the inner loop :)

If you really want to be hard-core, you can combine the x and y counters (and deltas) into the same register, by using 16 bit per each. But beware of precision. Doing something like tracking high-precision x and y for each 16th pixel, and using low-precision between usually pays off.

And, as a final tweak you could use mode 4 instead, giving you double buffering. In that case, you need to write two horizontal pixels at the time, due to the lack of 8-bit writes to VRAM. This fits nicely with the unrolling, though :)

Congrats, now you have one of the fastest braindead rotozoomers for the GBA -- just using the BG transformation matrix is just soooo much faster :)

As a final nit-pick: your use of leading underscores for the _x and _y variables is not entirely kosher C++. Symbols starting with underscore is reserved for the C++ implementation.

@tomaes: when de monster asks you to guess his trow, it's really vage.
i don't know how to format my reply. " 1 2 3" "123" "1,2,3" is all possible, but a wrongly formatted guess just counts as a wrong reply. (even entering "?" made me loose a guess.)

also, let some space between the messages, it's hard to know where you need to start reading for the last new message.

but i can see a cool game comming up :)

whynot2000: You just enter them one by one. F.e., an easy example: If the throw is 4, you know it can only be 1+1+2. So you enter your first guess "1" <enter> etc. Depending on the answers (f.e. a guess you were not sure of, because multiple options would be possible), you can change your strategy.

I'll upload a win32 version of the latest 0.6.2b to sourceforge later today. Bitbucket always has the latest version. :)

(on second thought, that routine was kinda broken anyway. I am at 0.6.5b now and with the help of a really anal playtester, his thing is approaching actual playability/extended-coffee-break-fun rather quickly... ;)

or just consider enemies, playership, playshots and so as entities and just do this : (with actor being of course a class and playership a sub-class of actor)

if worlddestroyedbylargehadroncollider = 1 then print "The world has been destroyed by the large hadron collider" else print "The world has not been destroyed by the large hadron collider" endif

artrotate#[threeDs[i]] = ATan2(ourY#,ourX#)

slightly OT yet code-related:

emacs ibuffer mode is the best thing since sliced bread! :)

here's my .emacs ini file , compatible with emacs 23.1 / win32 (place in c:/.emacs)

random-LOC:

wurstkäse

bra andpantsoff

andpantsoff ret

...can't wait to fire up the ASM again this weekend