[Closed] Creating a Mesh is faster than Instancing One?

It seems like if I create a copy of the actual mesh and assign it to an empty Editable_Mesh, its faster than Instancing??

(
	start = timestamp()
	
	local objs = #()
	local geo = selection[1]
	if (classof geo != Editable_Poly) do convertToPoly geo
	local mesh = selection[1].verts as bitarray
	local pGetVert = polyop.getVert
	local theMesh = snapshotAsMesh $Teapot01
	local rndClr = (random [0,0,0] [255,255,255])
	local wireColor = color rndClr.X rndClr.Y rndClr.Z 255
	DisableSceneRedraw()
	
	with undo off 
	(
		for vert in mesh do 
		(
			a = pGetVert geo vert node:geo
			local t = Editable_mesh()
			t.mesh = theMesh --assign TriMesh value to the EMesh
			--t = teapot()
			--t= Instance $Teapot01
			--t.radius = 1.0
			t.pivot = [t.center.X,t.center.Y,t.min.Z]
			t.pos = a
			t.wirecolor = wireColor
			freeze t
			append objs t
		)
		
		select objs
	)
	
	delete theMesh --free up memory
	EnableSceneRedraw()
	end = timeStamp()
	gc()
	format "Processing took % seconds
" ((end - start) / 1000.0)
)

Yeah, creating 961 Teapots in Max 2010 64bit, using Vert Positions, it took

1.873 seconds using Instance $Teapot01
vs.
0.903 seconds using the Empty Mesh and copying over the data.

Maybe this is because of all the other properties that come with instancing like Materials and such? Although, I prob. have to stick with the slower method, because I need to have Instance/Reference working in my other script that this is for…:twisted:

I forgot, assigning a specific name in Max 2010/2011 makes it quite a bit faster.

(
	start = timestamp()
	DisableSceneRedraw()
	
	local objs = #()
	local geo = selection[1]
	if (classof geo != Editable_Poly) do convertToPoly geo
	local mesh = selection[1].verts as bitarray
	local pGetVert = polyop.getVert
	local theTeapot = $Teapot01
	local isnt = Instance
	
	with undo off 
	(
		for vert in mesh do 
		(
			a = pGetVert geo vert node:geo
			t = isnt theTeapot
			t.pos = a
			t.name = "wee"
			t.isFrozen = true
			append objs t
		)
		
		select objs
	)
	
	EnableSceneRedraw()
	end = timeStamp()
	gc()
	format "Processing took % seconds
" ((end - start) / 1000.0)
)

This is very odd…

So I used the standard Shift + Drag/Instance Method, and created 600 Teapots, and it was basically instant.

Then right after that I did the same but using my code again on the same mesh as before, but now is is executing the code roughly 3x faster… I’m creating almost 1000 objects in 0.355 seconds vs. around 1 or 1.2 before.

Edit: I found out why. I didn’t realize I did it. But if you’re instancing a non-convering Standard Mesh, its at least 2x faster to instance, if not more.

I noticed that it’s faster to instance primitive objects, like teapot, than editable meshes. Or was it the other way round?

It was alot faster to instance primitives than converted meshes.

Maybe I need to do some sort of Proxy Mode for what I’m doing using Primitives.

Tested again, and it still holds true, Primitives are 3-4x faster than instancing a converted Mesh.

Well yes, exactly. But in regards to the built in ones, if you use a standard teapot, and don’t convert it, its much faster to Instanced that if you do convert it.

I have a Preview Mode in my Object Placer tool, which makes the objects grey/frozen anyways in preview. So if the user wanted faster previewing, and just wanted to get rough placement/randomization, they could use the primitive mesh I suppose.