Hi ferdinand,
Thank you for your time please do not prioritize this conversation, I feel bad for hijacking your time.

Heavy Scene from above
disable all drawing operations (Anim Without View, No Ui Update) looks fine i guess

Total rendering time:  648.58 ms
Single-threaded rendering time:  566.39 ms
Single-threaded finalizer time:  0.00 ms
Multithreaded rendering time:  82.19 ms
Multithreaded rendering 12.673 %
Kernel execution time: 82.19 ms
Worker idle time: 0.00 ms
Worker sleep time: 0.00 ms
Threading efficiency 100.000 %
Expected scaling would be about 3.611 x

I am expressing myself not precise enough.

its not about the raw FPS I am worried about, 11 FPS on our vehicles is fine I guess. what keeps me up at night is these recurring stutters when moving the camera through the scene (no animation just looking at stuff)
and things like the time to see a simple cube in a empty scene - it always feels laggy or you could say delayed
or how the Live selection circle lags behind the mouse pointer (empty scene)

oversimplified example:
Heavy scene FPS graph ( 0, 0, 0, 11, 11, 11, 11, 0, 0, 11, 11, 11, 11, 0, 0, 11, 11 )
Cube creation FPS graph ( 0, 0, 0, 60, 60, 60, 60, ... )

Plugin related this is more or less what started this hunt .... :
I can interact with my Draw line Plugin (that is not the state of the plugin its evolved now but to give a exaple) on my:

• laptop with ~60 FPS (Quadro 2000)
• on my private PC with R20 with ~200+ FPS (RTX 4080)
• on my workstation with ~8 FPS (RTX 3090)
  yes crawlingly slow

the plugin performance depends on Viewport size and point number but the difference repeatable between these "test" machines.

(Cinebench 2024.1 - GPU Test 3090 ~ 17000 points 25Minutes continuous) Feels fine

cheers

Thanks Ferdinand,

the Summary only pops on "Anim Test"

Empty Scene:
my old laptop has as a live readout (~ 5ms, at 60 FPS )
my Workstation has as a live readout (~ 33ms 52 FPS )

compared to yours, mine seems off - while this will not help me immediate it keeps me sane ( I was starting to doubt myself if I really have a lag or problem ...)

thank you

Live readout 17ms, 59 FPS
Workstation: RTX 3090, i9-10980XE 
empty scene 
 ============================ General =================================

Performance cores:  36
Virtual cores:  18
Efficiency cores:  0
Threads used for the queue:  36

Total rendering time:  1744.82 ms
Single-threaded rendering time:  1711.42 ms
Single-threaded finalizer time:  0.00 ms
Multithreaded rendering time:  33.40 ms
Multithreaded rendering 1.914 %
Kernel execution time: 31.02 ms
Worker idle time: 0.00 ms
Worker sleep time: 0.00 ms
Threading efficiency 92.887 %
Expected scaling would be about 1.365 x

Total rendering time equals single-threaded rendering time plus finalizer time plus multithreaded rendering time.
Threading efficiency is the percentage of multithreaded rendering time during which kernel code is executed
The expected scaling is calculated based on the number of physical performance, efficiency and virtual cores.
This how the scaling should be given the percentage of single-threaded code and the efficiency of the threading.
If your actual scaling is far lower than the expected scaling your code does not scale properly: It might 
contain atomics, mutexes or any other form of synchronisation or is memory-bound and/or out of CPU caches.
No lost events during profiling.

============================ By Scope =================================

 total 1.07 ms
ViewScheduler total 1411.58 ms Threaded 33.40 ms avg 31.02 ms min 31.02 ms max 31.02 ms [latency avg 0.00 us min 0.00 us max 0.00 us]

============================ Detailed in Order of Execution =================================

86.7 us: 	 total 1.07 ms self 1.07 ms
1.2 ms: 	ViewScheduler total 62.02 ms self 61.56 ms
2.7 ms: 		JobGroup total 0.46 ms self 0.03 ms avg 0.43 ms min 0.43 ms max 0.43 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
... truncated

Live readout 97ms, 11 FPS
Workstation: RTX 3090, i9-10980XE 
vehicle with 10.000 objects 119mio polys  ~ 10% displayed
============================ General =================================

Performance cores:  36
Virtual cores:  18
Efficiency cores:  0
Threads used for the queue:  36

Total rendering time:  7433.32 ms
Single-threaded rendering time:  5642.04 ms
Single-threaded finalizer time:  0.00 ms
Multithreaded rendering time:  1791.29 ms
Multithreaded rendering 24.098 %
Kernel execution time: 1788.86 ms
Worker idle time: 0.00 ms
Worker sleep time: 0.00 ms
Threading efficiency 99.864 %
Expected scaling would be about 5.957 x

Total rendering time equals single-threaded rendering time plus finalizer time plus multithreaded rendering time.
Threading efficiency is the percentage of multithreaded rendering time during which kernel code is executed
The expected scaling is calculated based on the number of physical performance, efficiency and virtual cores.
This how the scaling should be given the percentage of single-threaded code and the efficiency of the threading.
If your actual scaling is far lower than the expected scaling your code does not scale properly: It might 
contain atomics, mutexes or any other form of synchronisation or is memory-bound and/or out of CPU caches.
No lost events during profiling.

============================ By Scope =================================

 total 6.35 ms
ViewScheduler total 7001.25 ms Threaded 1791.29 ms avg 1788.86 ms min 1788.86 ms max 1788.86 ms [latency avg 0.00 us min 0.00 us max 0.00 us]

============================ Detailed in Order of Execution =================================

63.5 us: 	 total 6.35 ms self 6.35 ms
6.4 ms: 	ViewScheduler total 99.50 ms self 76.23 ms
10.6 ms: 		JobGroup total 0.08 ms self 0.00 ms avg 0.08 ms min 0.08 ms max 0.08 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
10.7 ms: 		JobGroup total 0.17 ms self 0.00 ms avg 0.17 ms min 0.17 ms max 0.17 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
11.0 ms: 		JobGroup total 0.04 ms self 0.00 ms avg 0.04 ms min 0.04 ms max 0.04 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
11.0 ms: 		JobGroup total 0.35 ms self 0.00 ms avg 0.35 ms min 0.35 ms max 0.35 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
11.4 ms: 		JobGroup total 0.13 ms self 0.00 ms avg 0.13 ms min 0.13 ms max 0.13 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
11.6 ms: 		JobGroup total 0.37 ms self 0.00 ms avg 0.37 ms min 0.37 ms max 0.37 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
12.0 ms: 		JobGroup total 0.14 ms self 0.00 ms avg 0.14 ms min 0.14 ms max 0.14 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
48.1 ms: 		JobGroup total 21.98 ms self 0.03 ms avg 21.95 ms min 21.95 ms max 21.95 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
112.9 ms: 	ViewScheduler total 138.31 ms self 116.85 ms
truncated

heavy scene camera orbit animation
============================ General =================================

Performance cores:  36
Virtual cores:  18
Efficiency cores:  0
Threads used for the queue:  36

Total rendering time:  9071.78 ms
Single-threaded rendering time:  7352.54 ms
Single-threaded finalizer time:  0.00 ms
Multithreaded rendering time:  1719.24 ms
Multithreaded rendering 18.952 %
Kernel execution time: 1698.41 ms
Worker idle time: 0.49 ms
Worker sleep time: 0.00 ms
Threading efficiency 98.788 %
Expected scaling would be about 4.854 x

Total rendering time equals single-threaded rendering time plus finalizer time plus multithreaded rendering time.
Threading efficiency is the percentage of multithreaded rendering time during which kernel code is executed
The expected scaling is calculated based on the number of physical performance, efficiency and virtual cores.
This how the scaling should be given the percentage of single-threaded code and the efficiency of the threading.
If your actual scaling is far lower than the expected scaling your code does not scale properly: It might 
contain atomics, mutexes or any other form of synchronisation or is memory-bound and/or out of CPU caches.
No lost events during profiling.

============================ By Scope =================================

 total 5.63 ms
ViewScheduler total 8711.70 ms Threaded 1719.24 ms avg 1698.41 ms min 1693.29 ms max 1703.53 ms [latency avg 11857.61 us min 11857.61 us max 11857.61 us] [sleep 0.00 us idle 493.53 us]

============================ Detailed in Order of Execution =================================

61.5 us: 	 total 5.63 ms self 5.63 ms
5.7 ms: 	ViewScheduler total 129.34 ms self 108.51 ms
9.5 ms: 		JobGroup total 0.07 ms self 0.00 ms avg 0.07 ms min 0.07 ms max 0.07 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
9.6 ms: 		JobGroup total 0.19 ms self 0.00 ms avg 0.19 ms min 0.19 ms max 0.19 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
9.8 ms: 		JobGroup total 0.03 ms self 0.00 ms avg 0.03 ms min 0.03 ms max 0.03 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
9.9 ms: 		JobGroup total 0.34 ms self 0.00 ms avg 0.34 ms min 0.34 ms max 0.34 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
10.3 ms: 		JobGroup total 0.10 ms self 0.00 ms avg 0.10 ms min 0.10 ms max 0.10 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
10.4 ms: 		JobGroup total 0.08 ms self 0.00 ms avg 0.08 ms min 0.08 ms max 0.08 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
10.6 ms: 		JobGroup total 0.07 ms self 0.00 ms avg 0.07 ms min 0.07 ms max 0.07 ms [latency avg 0.00 us min 0.00 us max 0.00 us]
42.9 ms: 		JobGroup total 19.49 ms self 0.03 ms avg 19.47 ms min 19.47 ms max 19.47 ms [latency avg 0.00 us min 0.00 us max 0.00 us]```

Thank you ferdinand.
I did not know about the "new" FPS helper, -> there is a checkbox "Save Profile Summary" -> where is that stored and how can I review it ?

What do all the options do exactly, is there a short explanation somewhere ? I could not find anything in the C4D Help.

thanks

Dear community,

It would be really nice if there was a SDK way to measure viewport performance. ( I guess profiling wont suffice) - Something like a plugin to measure viewport performance or something that logs frames, frames consistency, from the C4D internals.
C++ plugins are out of my expertise, and I suspect an "external" plugin wont have the access to all the measurements needed.

For plugin developers this could be useful I guess.

Where is the question ? I don't know this is probably a desperate cry for help.

Reason: I am having lag and freezes (empty scene) since S26 onwards (vague guess) and I can not get rid of them, even after several support sessions with Maxon and internal IT.

Thank you for your condolence.
mogh

Hey ferdinand,

I mixed different questions - I'll keep to the theoretical part to keep the Post relevant to the original question.

I appreciate the informative post as always.

Parametric concept of a spline is clear to me -> that's why i asked about "C4D can not get more precise than a 0° degree line object ?" which was formulated to lose as 0° is mathematical problematic? and I know 0.1 would be better. sorry about that.
The tip about spline 'Adaptive' setting was a good hint, I never use it in work so I forgot about it.

So to summarize regarding my question.

The Line object always samples what the user has set in the spline settings and not by a unknown internal setting (for example to give good viewport performance, or nice equal sections)
And its a better fit for my application, hence there is no "linear transport" shenanigans going on.
And if I want to display a high resolution of my calculation, I make an internal copy of that spline and edit the spline settings, instead of sampling a low division spline with spline helper.

Normalization / Jagged lines:
I use Menger-Krümmung. -> will open a new topic.
Your code might come in handy when I'll extend the plugin to polygon meshes.
...
...
...

Remark

as humans cannot see that level of detail.

I have to disagree on this topic. As a tell tale we import CAD data here into C4D to evaluate the work of the engineers.

Not long ago I said to our surface engineer. "take a look at the end of that surface something feels off, it looks like the surface is lifting" -> The engineer reportet back "you just spoted a 0.001mm surface deviation of an unconstrained spline ending I did not set."

We could argue that the seen result was in fact not 0.001mm but perhaps to surface setup more than 1mm, (or by some rounding error C4D normal tag shading where continuous and then stagnant for 1 pixel) but my point is settings in precision are important and carry through till the final product. Even though a setting of 0.1° for a spline might might be unsuitable in most cases (rendering, performance, ...) it is not for evaluation.

So even C4D and all that rounding / conversion / floating point precision might look like unrealiable, unecessary or even esotheric it caries through to a certain degree and its more reliable than you think.

(Yes we evaluate in the original CAD Program, but the tale is that I saw something even after importing it into C4D)

As reference to why I need this:

Blender Plugin: Surface Diagnostics - Josef Ludvík Böhm

Thank You

Thank you ferdinand,

I will implement my own "tagent" / "colinear" ending then, and not use this flag.

@ferdinand said in Usage of SplineHelp.InitSplineWith() regarding flags -> returned data:

As a warning: All the spline helpers implement parallel transport, which can make their output for curvature tasks undesirable (as it messes with what is the mathematical tangent of a spline in favour of what humans would expect. When you are at a loss about what I am talking about, search the forum for parallel transport and my username, the topic has come up multiple times in the past). It might be better to get the LineObject of a spline, i.e., its current discrete form, and do all the math based on its vertices.

Aha ... thank you - this explains why I get so different from spline-helpers tangent and cross tangent. Could also corelate to my "noisy" sampling with splinehelper (see screen below)
My question other topic but same context would be how to get a more high resolution Line-Object. We talked about this in an early stage about this, but my code wasn't mature enough to be sure. -> Is the short answer C4D can not get more precise than a 0° degree line object ?

Screenshot line object Bezier 0° degree (for enough resolution). Nice plot but resolution enough? Probably not for edge cases like flatish dips.

Splinehelper 200x samples not moving average smoothed. The noise starts to show at about double the resolution of a line object ...

Should I open another Topic?
thank you again

Dear community,

Returned Data of Flag c4d.SPLINEHELPFLAGS_CONTINUECURVE unclear.

I've read that the SplineHelp() has flag called c4d.SPLINEHELPFLAGS_CONTINUECURVE.
While the written explanation is clear to me :
Continue the curvature of an existing spline at the end points.

Its unclear to me how to utilize the data or even how the data I get is different from a "normal" helper.

Reason of interest: I am coding a "Curvature" displaying Tag Plugin atm.

thank you for your time.
cheers mogh

C4D Python SDK - Splinehelper

Hm, interesting, It works now,

either you did something or the solution is to post a comment to trigger it. Anyway Thanks.
Cheers mogh

I still have the same problem. "No push" no "unread toppics". Clearing my Browser Chache does nothing.

any idea ?
cheers mogh

I use this to Thank You Ferdinand.

This is how I imagine an SDK example should look like .... a simplest version , an altered simple version and a complex touching the boundaries of the subject.
I am thankful and praise your work .... !

cheers mogh

Ok Understood,

Ok I guess that explains my "nice" comb from the cache and my jagged flip/flop . sometimes zero comb from the "full" matrix normal ...

I also have different tangents from manually "calculating" them then v3, v2. But I will make a "sanity" check this weekend (feed manual matrixes into my code instead of spline "point" matrixes), before I post about this.

Thanks Ferdinand
I'll divide and conquer away.

@ferdinand
Thank you for your answer, exactly what i need, I am trying to incorporate the additional data but get weird results.

Is there a switch up /typo in the above code I get m.v3 <-> m.v1 ? meaning v1 is the normal ?

Is there a page in the SDK besides the "Matrix" one that explains the additional values and their features ?

Thank you for your Time.
mogh

Understood Thanks.

Ok I guess I have been vague what my issue is.

from the above helper I only get the points the spline is set to (eg. adaptive 5°) but I perhaps want more points in between. That of course only makes sense in certain scenarios but even though if the spline is a straight line there must be a solution to get points in between.

Example different adaptive degree values and the resulting points displayed as lines on the spline - Spline is a copy so no altering of the gizmos.

sorry to be so basic
mogh

How to receive the high definition version / Interpolation of a spline even when the user has set it to "low" detail?
How high definition can I get ? is there more than the user Interface of C4D offers ?

E.g. Interpolation adaptive, angle 0.0

Or am I using c4d.utils.SplineHelp() wrong and there is a another helper that gives me requested amount of points from a spline?

helper = c4d.utils.SplineHelp()
if not helper.InitSplineWith(spline, c4d.SPLINEHELPFLAGS_RETAINLINEOBJECT):
    raise RuntimeError("Could not initialize spline helper.")
line = helper.GetLineObject()
points = line.GetAllPoints()

cheers
mogh

How to add a GUI to a TagData Plugin without *.res files ?
... I am on the fence with these

Hope this is basic an clear enough for no further explanation.

cheers
mogh

not bad for 5 hours of dabbling --- thanks to you @ferdinand

Gotcha's:
The Curvature was 90dgree rotated - LLM to the rescue ...

# c4d.plugins.RegisterTagPlugin( ....
# That one nearly got me ;-)
# c4d.TAG_IMPLEMENTS_DRAW_FUNCTION # R22 

cheers

Thanks @ferdinand for your considered reply,

At this stage because I just want to "make stuff" I will start as a tool and perhaps learn / have an idea to transfer it to something more useful.

The obvious simple minded question - why is a tag Plugin not enough can a tag not draw into viewport ? Or only not via Python?

Thanks
mogh

Dear Developers,

I am starting to build a new Plugin, which should in the end, draw Curvature Combs / plots onto the selected edges / splines, preferable real time on editing the splines/models.

My question is which Plugintype would be best suited ?

Thinking out loud:
I was thinking a Tag Plugin but It could be to restricted as per the Tag plugin has some programing limits as I understand. Command or Tool plugin could work but I fear it will be clunky from a user experience.
At best a Polygon model with Hypernurbs applied should display several combs at the same time using the Hypernurbs Subdivision for more "accuracy" than a pure Polygon edge.

The plugin probably needs a Link Box to "receive" a selection. I am thinking of Ferdinands Dynamyc Link Example would be cool to have more than on comb active but only one "Tag / Plugin / ...." per object ...

Closest thing C4D has is measure and construct Tool.
As a starting point I would start to display Combs on Splines ...

So to Pick the hive mind - > What Plugin Type is best suited for stuff like this?

Moderators: feel free to move this Topic if aplicable.

Cheers
mogh