[speedtest] iso, slopes and not-tilebased tiles

[nGFX]

hi folks,

some of you might have seen my iso-engine while reading the tales frome the codefront blog

i dumped the development of the engine for some weeks, but i had some time this weekend and started to optimize the code. although the result is somewhat dis-satisfying (rewrite 50 lines of code to gain 2ms) i think in the sum of it, it will be worth the efford.

the other option to make the thing faster would be to either drop the second screen (the option i favor) or use another way to playback the "opponent" (currently the playback uses the same physics as the player, cause this results in way less data ... but this way a second calculation cycle is needed)

to the demo
instructions:
press the record button, move the player with the cursor keys, use ctrl to jump
at any time (or after the record-buffer is full) press stop, then "rew" and afterwards "play" - the sprite on the second screen should now move the way you did before.
NOW move your (left screen) sprite again and have a look at the fps ...

post them along with some machine stats (os/browser/mhz/ram)

thanks in advance, <olli/>

ps: as for the not-tilebased tiles ... well read the first posts on crystalis for a description ...

[leight]

Dude, that's awesome You could make a Price of Persia: The Sands of Time replica - just put it in reverse

The fps for the recorded one:
- lowest fps: 17
- highest fps: 22
- average fps: 20

Machine stats:
- Broadband
- Pentium 4 1.5ghz 128mb RAM
- 6 MSN windows upen
- 1 word
- 1 outlook express
- 1 iTunes
- 2 Firefox

I love it!!!

[MiSSenLinX]

truely nice engine,

OMG WAS I IN THOSE MSN WINDOWS

[BlinkOk]

22-23
XP|2.8Ghz|2GB
nice work ollie

[adit_ya_sharma]

Wow, nice engine ngfx, and I mean absoultly amzing. You even got replay....cool man!

Win Me
750Mhz pentium 3processor.
128 mb ram
Avg. 20 Fps

[nGFX]

thanks guys, keep 'em coming ...

leight: they paid me an horrible amount of money not to ...

adit_ya_sharma: oop is your friend here (ok in this case also your foe) i was quite lazy when i coded the replay function ...
i just record the keystrokes (so there's a max of 3 values to store) and pass them to the "parseKeystroke" method - viola, done.

it would have been by far faster to record screen and player postions, but this would have required another method for playback ...

maybe this ends in a game ...

<olli/>

[leight]

i do suggest making this into a game. With the different slopes you could create a variety of games. Including iso platformer's with the POP:SOT function. Or you could have a partial level preview like on POP:SOT

Even if you don't end up using that replay/fast forward function. you could always make a war game, with terrain now you have slopes!

Many many options with a fantastic engine!

Ok ok, I do love that game!

[thehumanchimp]

750K broadband
Pentium 4
5 Firefox (it was run in Firefox which i find is usually slower than IE)
1 Word
1 Outlook Express

Highest: 18 (stationary)
Lowest: 8 (going up the "hills")
Average: 10-12 (generally moving about)

[St. Nick]

IE6
Pentium III 1 Ghz
Broadband "Lite" that really only works like dial up

Averaged about 21 fps for me, moving with replay on brought it down to 12 fps.

Cool stuff man

[yellowman]

xp - ie6 - amd1400 - 512mb
runs ~20 something fps.

Looks smooth so far. Would be cool to see how it performs on a larger map, the current one seems a little limited for speedtesting. Or maybe it´s just me who can´t get any flow... I just keep hitting those edges

Great work!
/klas

[metanet]

hey,
i don't suppose you want to share how you did that?

specifically, the world graphics look pre-rendered.. are they?
and if so, how did you generate the data used for collision?

anyway.. cool stuff.

thanks,
raigan

[dogtown08]

To through my super slow (at flash) ibook g3 os 10.3 fps:

Low: 8fps
High: 16 fps

Average ~12-14 fps

Very nice. That may seem slow, but for something like this on my computer, it is quite fast

[strille]

Looks great!

I got ~14 fps during playback and moving around (AMD 1400, Win XP, IE 6, some programs open).

About recording keystrokes, the way I do it in Sylvaniah is to have each key correspond to 1 bit in a number. That way you only need to store an integer for the current key combination (an integer 0-255 if you have 8 keys to contol the game). It's not stored each frame, but only when the number changes. When it changes, you also store the number of frames the previous number lasted.

Again, great work, looking forward to the complete game!

[T1ger]

Originally posted by strille
Looks great!

I got ~14 fps during playback and moving around (AMD 1400, Win XP, IE 6, some programs open).

About recording keystrokes, the way I do it in Sylvaniah is to have each key correspond to 1 bit in a number. That way you only need to store an integer for the current key combination (an integer 0-255 if you have 8 keys to contol the game). It's not stored each frame, but only when the number changes. When it changes, you also store the number of frames the previous number lasted.

Again, great work, looking forward to the complete game!

That's a really smart method for recording.. But how do you determine which buttons were pressed? subtract each value, starting from the highest, and check if the value is greater or equals zero? i think that would do the trick... But how are you doing it?

The game is great. well, "game" :P

But I can't see why a second screen is necessary? Why dont you just view the recordings on the same screen?

Stats:
Opera (really slow with flash) (latest v.)
flash mx04, photoshop, notepad, winamp, ftp-server, web-server, aim and msn-messenger running in the background.
avg. fps while playing and playing: 22

[token 3]

nGFX Nice work - I played this a while ago after seeing your blog, 22-23 fps XP|2.8Ghz|2GB, sorry for posting a whole load of code (below), but it may be useful to others

strille Is your compression anything like this?:

var ticks :Number = 0
var kU		:Number = Key.UP
var kD		:Number = Key.DOWN
var kL		:Number = Key.LEFT
var kR		:Number = Key.RIGHT
var lCombo	:String = undefined
var recording	:Array  = new Array
var lTicks	:Number = 0
var dTicks	:Number

onEnterFrame = function () :Void {
	var combo = new String
	// push either 0 or 1 into the combo string, and handle in-game events 
	combo += Number (Key.isDown ( kU ) )
	combo += Number (Key.isDown ( kD ) )
	combo += Number (Key.isDown ( kL ) )
	combo += Number (Key.isDown ( kR ) )
	// compress the combo
	combo  = parseInt( combo, 2 ).toString( 36 );
	// is the new combo different from the previous?
	// if so store the tick difference, last ticks, and the combo
	if (combo != lCombo) {
		dTicks	= ticks - lTicks
		lCombo	= combo
		lTicks	= ticks		
		recording.push ( String (dTicks + combo) )
	}		
	// increase the ticks / frame count
	ticks++
}

// just for testing
var kListener		= new Object ()
kListener.onKeyDown	= function ()  {	
	if ( Key.getCode () == Key.SPACE ) {
		trace ( 'full recording:\n' + recording.join ( ',' ) )
		// initial ticks is 0, unless specified
		var ticks 	= 0
		var i		= -1
		// loop throught the recording
		while (++i < recording.length) {
			// the length of the recording code
			var len	= recording[ i ].length - 1
			// tick difference is the all the recording code led the last character			
			var tickDif	= Number ( recording[ i ].substr( 0, len ) )
			// key code is the last character
			var code	= recording[ i ].charAt ( len )		
			// 'undo' the bit compression
			var code 	= parseInt( code , 36 ).toString( 2 )
			// add '0's to the start to pad the code if necessary
			while ( code.length < 4 ) code = '0' + code
			// add the stored tick difference
			ticks      += tickDif
			// test trace
			trace ( 'ticks: ' + ticks + ' > combo: ' + code + ' compressed version: ' + recording[ i ] )	
		}
	}	
}

Key.addListener( kListener )

Mike

[nGFX]

wow, this means a long answer ... ok ...

jtnw:
non-tilebased may be the wrong word, there are tiles in it, the difference is, that the tiles don't have the same size. so "irregular tiles" would be better.

yellowman:
in the final "game" you would fall of the edges ...
to the larger maps: np, the version that i have here and work with uses a VERY large map (about 500x600 "regular tiles") but with the irregular tiles there are a maximum of 30 tiles visible at one time. so if one tile disapears from the screen it is marked and reused when needed. (no attaching / removing)

metanet:
nope, sorry, not atm ...

strille:
yep, a way more practical way, but i was soooo lazy when i wrote the replay code ...
atm i store just 3 letters :"xyj", where x is "u" or "d", y is "l" or "r" and j " " or "j". cause you can't use up and down key at the same time

T1ger:
never said it's agame, didn't i?
oh, the second screen, as i started the "game", i wanted it to be multiplayer, either by playing against a recorded opponent or "one on one" using a socket connection. for speed reasons i think i have to drop the multiplayer idea 'till mm releases v10 of flash. i also thought about using the same screen for the playback, but this way you don't have the "feeling" of playing against someone else.

token 3:
nice piece of code ...

<olli/>

[strille]

token 3:

Yes, it's similar to yours. You can use the bits in an integer as switches though to tell if a key is pressed or not. Let's take a 4 bit integer, like 9. It looks like this in binary:

1001

Each bit represents a key being pressed or not. Let's say that they represent the keys up, down, left and right in that order. So, in this case (9), the up key and the right key is pressed.

All 4 keys pressed would give a number of 15, no keys a value of 0.

Start with a value of 0. If up key is pressed you add 2^3 = 8, if down is pressed you add 2^2 = 4, if left is pressed you add 2^1 = 2 and if right key is pressed 2^0 = 1.

To find out if up key is pressed or not, you can type:

code:

---

if (combo & 8) {

  // up key pressed

}

---

[io3]

(os/browser/mhz/ram)
w2k/ie6/p3 1GHz/400 mb and a few other apps open.

fps: varied from 13fps to 21fps but most commonly 15-17fps.