[Closed] MXS 2017 'For Loop' performance

Hah, yes I got your joke, but thought you might have some maxscript ninja manouvres to workaround the shortcomings of my code

i don’t remember where but i saw the using of MXSCallstackCaptureEnabled as a Sticky Context

i don’t have max 2017 and can’t try if but in case of the context is sticky you can set it ON and OFF as:

set MXSCallstackCaptureEnabled  on
set MXSCallstackCaptureEnabled  off

if it works the version universal code could be:

global CallstackCapture
(
	struct CallstackCaptureStruct
	(
	private
		_canset = (maxVersion())[1] / 1000 > 18,
	public
		fn canSet = _canset,
		fn enable = if _canset do execute "set MXSCallstackCaptureEnabled on",
		fn disable = if _canset do execute "set MXSCallstackCaptureEnabled off"
	)
	CallstackCapture = CallstackCaptureStruct()
	ok
)
-- CallstackCapture.canSet()
-- CallstackCapture.enable()
-- CallstackCapture.disable()

Hah, your ninja skills know no bounds. That is a nice workaround indeed!
Thanks again Denis.

does it work?

It doesn’t crash in 2016, that’s for sure. I haven’t run any benchmarks though.

Confirmed 4x faster using this code on 1000 iterations of failed try()catch()

Without MXSCallstackCaptureEnabled
time:162 heap:1272208L

and with :
time:44 heap:1149124L

But yes, a pretty rare use case. Apologies for hijacking this thread!

actually it’s interesting to know the difference between enabled vs disabled using of the CallstackCapture. how difference the time and memory usage?

Here you go :

3dsmax 2016
CallstackCapture enabled : time:28ms heap:1144928L
CallstackCapture disabled : time:30ms heap:1144872L

3dsmax 2017
CallstackCapture enabled : time:173ms heap:1288826L
CallstackCapture disabled : time:42ms heap:1164850L

code used : (in case I’m messing the test up)

global CallstackCapture
(
	gc()
	struct CallstackCaptureStruct
	(
	private
		_canset = (maxVersion())[1] / 1000 > 18,
	public
		fn canSet = _canset,
		fn enable = if _canset do execute "set MXSCallstackCaptureEnabled on",
		fn disable = if _canset do execute "set MXSCallstackCaptureEnabled off"
	)
	CallstackCapture = CallstackCaptureStruct()
	ok

	t = timestamp()
	h = heapfree
	count = 1000
	for k = 1 to count do (try(execute "test 123")catch())
	format "3dsmax 20%
CallstackCapture enabled : time:%ms heap:%
" (((maxVersion())[1] / 1000)-2) (timestamp() - t) (h - heapfree)     
	gc()

	CallstackCapture.disable()
	t = timestamp()
	h = heapfree
	count = 1000
	for k = 1 to count do (try(execute "test 123")catch())
	format "CallstackCapture disabled : time:%ms heap:%
" (((maxVersion())[1] / 1000)-2) (timestamp() - t) (h - heapfree)     
	CallstackCapture.enable()
)

it’s very interesting to see when this ‘CallstackCapture’ feature works well VS any previous version (2016 for example).

as i can see from the test results above this feature makes both performance and memory usage worse.

Decided to try this with the new service pack, didn’t really expect it would get much better… but didn’t really expect it would get that much worse:

Max 2016: time:696 ram:248L
Max 2017 SP1: time:8734 ram:264L
Max 2017 SP2: time:14630 ram:272L

Same increase with few of the other code snippets here, is it just me… or what the hell is going on?