GrabImage() question

Monkey Forums/Monkey Programming/GrabImage() question

Shinkiro1

(Posted 2011) [#1]

Local atlas:Image = LoadImage ("atlas.png")

Global image1:Image = atlas.GrabImage (0,0,32,32)
Global Image2:Image = atlas.GrabImage (32,32,32,32)

Considering the images I am grabbing are always global, do I have to keep a direct reference to the variable 'atlas' too?

Rixarn

(Posted 2011) [#2]

Hi Shinkiro,

Only if you need to keep grabbing parts of the atlas in other moments of your application. If you only do the grabbing once, perhaphs there is no need.

Still, I would encourage you not to use Globals unless you really don't have other choice. Once they start to grow (the amount of globals you have) you'll have a hard time mantaining your code. You could instead isolate your atlas (and the grabbed images) to a Singleton in your app and keep things organized:

Class ImageResources

	'---- singleton code --------
	Private
	
	Field instance:ImageResources
	
	Method New()
		If instance Then Error("Cannot create new instance of singleton type")
		instance = self
		instance.Init()
	End
	
	Method Init:Void()
		atlas = LoadImage ("atlas.png")
		image1 = atlas.GrabImage(0,0,32,32)
		Image2 = atlas.GrabImage(32,32,32,32)		
	End
	
	Public
	
	Function GetInstance:ImageResources()
		if Not instance
			instance = New ImageResources()
			instance.Init()
		End If
		Return instance
	End
	'--------------------------
	
	'your code
	Field atlas:Image 
	
	Field image1:Image 
	
	Field Image2:Image 
	
End

Call ImageResources.GetInstance().yourImage anywere in your app to access the resources in a controlled way. Just my two cents.

Shinkiro1

(Posted 2011) [#3]

Thanks for your answer,

yes globals would be a really bad idea (I just wanted to illustrate my problem, in fact they are fields).

By now my system is really cool. I can now do this:

UploadImageAtlas ("atlas1.png")
playerSprite.SetImage ("atlas1.png/player.png")

When the whole system is more mature maybe I should publish it.
(oh my ... not another framework ^^)

So is GrabImage as efficient as DrawImageRect()?

Rixarn

(Posted 2011) [#4]

So is GrabImage as efficient as DrawImageRect()?

Well, I'm not sure how DrawImageRect works. GrabImage will take the portion of the image and copy it to the new image. DrawImageRect will probably do some clipping. Since grab image is expected to be done only once, what you really are comparing is DrawImage vs DrawImageRect...

Can I assume that DrawImage is slightly faster than DrawImageRect? Probably yes.

AdamRedwoods

(Posted 2011) [#5]

I don't think there's much speed difference, it's just easier to treat the grabbed image as an image. Here's my implementation of an atlas loader:
http://monkeycoder.co.nz/Community/posts.php?topic=1216

There are quite a few out there, but there's always room for more.

Samah

(Posted 2011) [#6]

A grabbed image and its source image share the same surface texture in mojo, so there's not really any *copying* going on. It simply creates a lightweight object that points to the existing surface and stores some offsets.

Edit (since I was posting from the bus):
GrabImage will be as fast as creating a new object. DrawImage and DrawImageRect should be pretty much identical in terms of performance. DrawImageRect just uses different UV coordinates.

Shinkiro1

(Posted 2011) [#7]

Ok, did a small test to see if GrabImage really improves performance that much.

I did all the tests for html on a 21,5' iMac 2011.
The test was to load 10 different images and render them all over the canvas so it's completely filled, somtetimes overlapping. Nearly every image had dimensions like:120x120.

Without GrabImage:
Firefox: ~27fps
Safari: ~27fps
Chrome: ~18fps

With GrabImage (all 10 images in one big picture as atlas):
Firefox: 60-80fps
Safari: ~107fps
Chrome: ~115fps

WHOOO!!!
I was shocked when i saw Chrome's fps counter skyrocket. This is just amazing, it should be typed bold in the docs!

So, always use GrabImage, as it may increase your performance from 2x up to 6x (see Chrome).

Samah

(Posted 2011) [#8]

Using GrabImage is essentially the same as DrawImageRect, except it stores the offsets for you. Using an atlas is usually common practice since it doesn't have to do any image context switching. And yeah, HTML5 support in Firefox is still abysmal. :-)

anawiki

(Posted 2011) [#9]

Shinkiro1 - was your test with Atlas and without atlas? Or with atlas but drawing via drawimagerect vs grabimage? I believe that it was the first case, so no surprise here.

Shinkiro1

(Posted 2011) [#10]

@anawiki
Yes, it was atlas vs without atlas.

Fred	(Posted 2012) [#11]

And what is the best with an Atlas: DrawImageRect or DrawImage from a GrabImage ?

Samah

(Posted 2012) [#12]

And what is the best with an Atlas: DrawImageRect or DrawImage from a GrabImage ?

They're essentially the same thing, so personal preference. GrabImage is probably more convenient since you can forget about calculating offsets, etc.

fsoft

(Posted 2012) [#13]

could you please post the test source code, so we can do our timings and test?