PDA

View Full Version : More dynamics blues …



DrStrik9
02-05-2016, 02:15 PM
Ground = Static
Object(s) to be Broken = Parts, Start Sleeping
Ball to break object(s) = Rigid, Start Active (two keyframes, start and somewhere out toward the breakable object(s)) .

I modeled the objects, set up the scene quickly, and set the cursor at a low frame (80) to make sure things would work ok. They calculated pretty quickly and seemed ok, so I saved the scene, and set the last frame at 480. It took 14 hours to calculate the sim. Is Layout dynamics multi-threaded? If it is, I’d be surprised.

Having set the cursor at 0 and unsuccessfully re-saving the scene before, I decided not to risk it this time and attempted to export MDD. (I’m afraid to change the scene name, knowing dynacache has already been saved under the previous scene name). Am I right about this?

I’m wondering WHY it’s called “Multi” MDD Baker, since you have to export each object one at a time —??? also, you MUST choose the MDD output directory EACH TIME, or bad things happen.

After exporting MDD for each object individually, I was UNABLE to successfully load them back into the scene. Well, they “load” but do NOTHING after turning dynamics off and associating the MDD files with scene objects. In fact, I’m getting weird multiple views of at least one object, as I did before with a corrupt scene, although this scene is thankfully not corrupted (as far as I can tell).

I apparently need some guidance on using MDD “Multi” Baker and “Multi” Loader. This is anything but intuitive. Is there a tutorial somewhere on exporting and importing MDD?

Also, why is there an “Abort” button when simulating dynamics? You can hit that button a thousand times during sim calc, and it never does ANYTHING. All you get is a 1-2-second pause, then calculation continues unabated.

I was able (thankfully) to save the scene (which took 25 minutes), but so far, MDD is NO HELP at all. After a few tries with MDD, I turned dynamics back on and re-saved the scene. During this, Layout crashed, and now the scene is corrupted.

When it comes to LW dynamics, I’m asking myself, “how much pain do you want?” :-(

MonroePoteet
02-06-2016, 02:37 AM
It sounds like your Parts object is very complex. What's the polygon count on that object?

mTp

jboudreau
02-06-2016, 08:01 AM
Are you able to provide us the scene for testing? I've used bullet dynamics on complex scenes quite a bit and never ran into the problems you are having. Everything worked great mdd's, MDD multi loader, dynamics saving etc.

DrStrik9
02-06-2016, 09:50 AM
The parts object has 209,808 polys. I only went this far because Lino said Layout should be able to handle very complex scenes. I've wasted quite a few days trying to make this work, with similar results each time (super-long sim calc times, MDD failure, corrupted scenes).

I'd share the scene, but it is corrupted (again).

This has historically been my curse in life: to be the unwilling beta tester who encounters the killer bugs (or whatever this is). Right now, attempting to open this scene crashes Layout 2015.3.

I just need a very long break from LW dynamics.

Surrealist.
02-06-2016, 09:52 AM
Well screen caps of the set up and in particular the panels with settings at least.

MonroePoteet
02-06-2016, 01:18 PM
How much physical memory do you have on your system? My (very basic) experiments show that a Bullet simulation on a 240K polygon object tops out the working set at about 6GB, and page faults are up around 11K-14K towards the end of the simulation.

On my 8GB quad core 2.3Ghz processor, the simulation takes about 15 seconds per frame, 21:30 for the last 94 frames of the simulation. The 240K polygon object was just 1m cubes piled up in X and Y with Array. Requires patience, but seems workable (do other stuff while calculating if I wasn't timing it). Perhaps the individual parts of your object are many points & polygons apiece (my cubes only had 8 points and 6 polygons apiece).

This was a very simple simulation, simply letting gravity collapse the wall of 1m cubes. I'm running V11.6.3.

Yes, it appears to be multi-threaded. I see it using up to six of the 8 available CPU's during simulation.

When I want to interrupt a complex Bullet simulation, I don't use the Abort button, just hold down the ESC key until it notices.

Good luck! If you have less than 8GB of memory, I'd guess you're "thrashing" and need to either simplify the scene, get more memory, or try a Metalink'd simulation (i.e. simulate on a low-poly object, Metalink deformation node to a higher-poly object).

mTp

Snosrap
02-06-2016, 04:38 PM
I've got a related question to DrStrik9's concerning MDD's. Like DrStrik9 I had a heck of a time getting MultiReader to import my MDD's - I ended up importing them one at a time. But my question is, once imported do MDD's really slow do layout playback and is there a way to bake out an MDD on the timeline? And will that make layout more responsive?

jboudreau
02-06-2016, 09:35 PM
The parts object has 209,808 polys. I only went this far because Lino said Layout should be able to handle very complex scenes. I've wasted quite a few days trying to make this work, with similar results each time (super-long sim calc times, MDD failure, corrupted scenes).

When you are calculation your dynamics on the parts object under shape type are you using mesh? If so that is why your calculations are taking so long to calculate. If you use convex pieces it will be a lot faster

Also I'm doing a test scene now, I'll let you know what I find out

Hope this helps
Thanks,
Jason

- - - Updated - - -


The parts object has 209,808 polys. I only went this far because Lino said Layout should be able to handle very complex scenes. I've wasted quite a few days trying to make this work, with similar results each time (super-long sim calc times, MDD failure, corrupted scenes).

When you are calculation your dynamics on the parts object under shape type are you using mesh? If so that is why your calculations are taking so long to calculate. If you use convex pieces it will be a lot faster

Also I'm doing a test scene now, I'll let you know what I find out

Hope this helps
Thanks,
Jason

Surrealist.
02-07-2016, 12:15 AM
The parts object has 209,808 polys. I only went this far because Lino said Layout should be able to handle very complex scenes. I've wasted quite a few days trying to make this work, with similar results each time (super-long sim calc times, MDD failure, corrupted scenes).

I'd share the scene, but it is corrupted (again).

This has historically been my curse in life: to be the unwilling beta tester who encounters the killer bugs (or whatever this is). Right now, attempting to open this scene crashes Layout 2015.3.

I just need a very long break from LW dynamics.

I think there is absolutely truth in finding out where things hit the limits in the software when dealing with large scenes. And this goes for rendering as well.

In your case, since you are dealing with such a large scene, one solution might be to run the same scenario - exactly - with a smaller scene. If it works or does not work it will be much easier to find out where the problem is if there is one on your end. Then once you have a working scene, you know that set up works.

Then repeat it exactly or add to it with the same working scenario as the smaller scene. If it falls apart, get that scene to support some way, even if the scene crashes on load, 'cause they might be able to look at the scene in text format and get it to open and debug it. In fact they might even be able to locate the bug in text format without even opening it.

If you are feeling brave, open your .lws in a text editor (save the original) and read it. Then start chopping out parts of the scene methodically and narrow it down to the line or lines that is causing the scene to crash. I have done this with Maya ASCII files and I am sure it works basically the same way with LightWave. You just have to determine the parts of the scene that are a must have for it to open. And you can probably read that or even research it online, Everything else, chop away.

This is the tut I found for Maya scenes. The basic method is what I am talking about:

http://polycount.com/discussion/66262/tutorial-holy-crap-i-lost-all-my-work

DrStrik9
02-07-2016, 08:24 AM
Thanks, everyone, for your comments.

MonroePoteet, this computer has 32 GB RAM. As I said in my original post (a different thread), it sure would be nice to know where that "scene-complexity" boundary is, apparently paired with how much RAM is installed. Glad to hear Bullet sims are multi-threaded; I should have taken a look at the Activity Monitor on Mac OS. I also tried the Escape key to interrupt sim calc, but apparently didn't hold it down long enough. The "Abort" button should be removed, and its function paired to a repeating direct line to the escape key with error trapping until success is achieved, if this is how the sim is actually stopped, since the Abort button at this time is pointless.

jbondreau, the parts object was set to mesh. I had no clue that convex pieces would be faster to calculate. If this is so, shouldn't parts bodies be convex by DEFAULT? -- but the parts were 6-sided elongated CUBES, not convex parts. (Fracture is fraught with errors anyway, so I don't use it anymore.) There would be no point in uploading this scene, since it is corrupted, and crashes Layout when it is opened.

Surrealist, I DID a simple version of this scene, and it worked fine, including MDD, with no problems. Lino insisted that Bullet will do very large sims, so I tried this, over a four-day period, with tragic results, ending in MDD failure, weird double-objects, and Layout-bombing corrupted scenes, TWICE in a row.

--

I've seen HUGE bullet sims on YouTube, created in Blender. I can't stand Blender's interface, but I'm thinking future complex bullet sims might need to be created with a tool that actually works. But then again, who knows ... that guy may have 128 GB of RAM, and days/weeks to wait for his sims to calculate ...

jboudreau
02-07-2016, 08:37 AM
Dynamics is definitely broken in 2015.3 especially when it comes to convex pieces. Collision is not working well with convex pieces in lightwave 2015, 2015.1, 2015.2, 2015.3 the pieces keep going through the floor no matter what I change also the visible representation of convex pieces is completely different too from 11.6.3 to 2015.3

I opened the same scene in layout 11.6.3 and it works perfectly, there are no pieces that go through the floor. It works as it should and the visible display is correct as well. In 2015.3 the display looks like a bunch of dots what's changed and why am I getting these dots instead of how it use to look in 11.6.3?

Here are some screen grabs showing you the difference

132250

132251

I'm going to keep on checking what else might be broken in 2015.3 regarding dynamics.

DrStrik9 - Have you tried using lightwave 11.6.3 dynamics instead of 2015.3, it may fix your problem

Thanks,
Jason

jboudreau
02-07-2016, 08:48 AM
Thanks, everyone, for your comments.

jbondreau, the parts object was set to mesh. I had no clue that convex pieces would be faster to calculate. If this is so, shouldn't parts bodies be convex by DEFAULT? -- but the parts were 6-sided elongated CUBES, not convex parts. (Fracture is fraught with errors anyway, so I don't use it anymore.) There would be no point in uploading this scene, since it is corrupted, and crashes Layout when it is opened.

Hi DrStrik9 - Oh yeah mesh takes forever to calculate compared to convex pieces. (I agree convex shapes should be the default) The best thing to do when setting up your dynamics is turn of dynamics, Set all your objects the way you need them, Rigid, Parts, Static etc. Change your shape for your parts object to Convex Pieces, Once this is done they you can turn on dyamics. Dyamics will start decomposing the convex pieces this doesn't take very long. Now run your simulation you will see a huge difference in calculation speed. I had a object with over 300,000 polygons calculate in about 1min, With it set to mesh it was taking forever just to get it to even start calcuating so I had to stop it using the task manager because it was just not responding.

The issue you will have now is it looks like convex pieces has changed or is broken in 2015.3 it works great in 11.6.3 so maybe you could give that a try. I've made very complex simulations with 11.6.3 and it never had a problem. I had this scene where a cup of coffee came sliding through 1000's of coffee beans, This scene was very comples and had millions of polygons, The calcuations were very fast, 5 - 10min or something like that I can't remember exactly but definitely not hours. I also used convex pieces for that simulation. If I had used mesh it would of taken forever and to be honest the job wouldn't of gotten done because I don't think it would of been able to handle such a complex scene.

If you had 6 sided elongated cubes, then you should be using box, mesh is for very complex shapes. To be honest I never use mesh on complex objects I just use convex pieces which saves hours when it comes to simulating complex scenes.

UPDATE:

Just did a test with the same scene

Lightwave Scene

Parts object 21,000 polygons
Shape set to mesh (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 1 min

Parts object 21,000 polygons
Shape set to convex pieces (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 12 seconds (Thats almost 6x faster)

That's a huge difference in my opnion and this is just with 21,000 polygons

Just did a test,

Hope this helps
Thanks,
Jason

DrStrik9
02-07-2016, 09:36 AM
Thanks, Jason. I'll give it one more try.

... I'll try the convex pieces, and let you know what happens.

One last time ...

DrStrik9
02-07-2016, 10:34 AM
Hi DrStrik9 - Oh yeah mesh takes forever to calculate compared to convex pieces. (I agree convex shapes should be the default) The best thing to do when setting up your dynamics is turn of dynamics, Set all your objects the way you need them, Rigid, Parts, Static etc. Change your shape for your parts object to Convex Pieces, Once this is done they you can turn on dyamics. Dyamics will start decomposing the convex pieces this doesn't take very long. Now run your simulation you will see a huge difference in calculation speed. I had a object with over 300,000 polygons calculate in about 1min, With it set to mesh it was taking forever just to get it to even start calcuating so I had to stop it using the task manager because it was just not responding.

The issue you will have now is it looks like convex pieces has changed or is broken in 2015.3 it works great in 11.6.3 so maybe you could give that a try. I've made very complex simulations with 11.6.3 and it never had a problem. I had this scene where a cup of coffee came sliding through 1000's of coffee beans, This scene was very comples and had millions of polygons, The calcuations were very fast, 5 - 10min or something like that I can't remember exactly but definitely not hours. I also used convex pieces for that simulation. If I had used mesh it would of taken forever and to be honest the job wouldn't of gotten done because I don't think it would of been able to handle such a complex scene.

If you had 6 sided elongated cubes, then you should be using box, mesh is for very complex shapes. To be honest I never use mesh on complex objects I just use convex pieces which saves hours when it comes to simulating complex scenes.

UPDATE:

Just did a test with the same scene

Lightwave Scene

Parts object 21,000 polygons
Shape set to mesh (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 1 min

Parts object 21,000 polygons
Shape set to convex pieces (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 12 seconds (Thats almost 6x faster)

That's a huge difference in my opnion and this is just with 21,000 polygons

Just did a test,

Hope this helps
Thanks,
Jason

I really appreciate your responses and help. I'm still using 2015.3 ... just to see if I can narrow down some repeatable Bullet bug. -- I am unable to set up dynamics when it's off, so I had to leave it on. The trick seems to be leaving the cursor at fm 0 for this. It's interesting how when changing the parts objects from the default mesh to convex pieces, Layout spends several minutes "decomposing" each parts object ... I don't know what it's doing, but hopefully it isn't decomposing my objects to crap again :D ... and that this will dramatically reduce calc time.

... Well, you're right, in 2015.3, convex pieces is definitely broken. And generally, 2015.3 bullet with complex objects, even with mesh objects, is also broken.

This time, the calculation went a lot more quickly than before (13 minutes to fm 200 of 480). But you're right, a large percentage of the parts are now falling through the floor (something that didn't happen with mesh parts -- prior to scene corruption, that is).

Have you reported this bug?

--

If this survives the long calc and scene saving, I'll try "box" on the parts, to see if that works any better ... do you know if "box" doesn't fall through the floor in 2015.3? And do you have much success changing the rigid object's second keyframe and recalculating the sim? -- This usually caused mayhem in my earlier failed tests.

jboudreau
02-07-2016, 11:26 AM
I advise you to use 11.6.3 if possible. Dynamics is seriously messed up in 2015.3. I just did that same calculation with mesh that I mentioned above.

Lightwave Scene

http://www.animatrixproductions.com/Dynamics.zip

Lightwave 11.6.3

Parts object 21,000 polygons
Shape set to mesh (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 1 min

Parts object 21,000 polygons
Shape set to convex pieces (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 12 seconds (That's almost 6x faster)

Lightwave 2015.3

Parts object 21,000 polygons
Shape set to mesh (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 7 min ( That's 7x slower than 11.6.3)

Parts object 21,000 polygons
Shape set to convex pieces (The parts object falls onto the ground plane and shatters into a bunch of pieces)
Simulation Time: 14 seconds (Thats almost 6x faster)

Really 1 minute to simulate in 11.6.3 and 7 min to simulate the same scene in 2015.3. Something seriously got messed up in lightwave 2015 when it comes to bullet dynamics

Thanks
Jason

jboudreau
02-07-2016, 12:31 PM
I really appreciate your responses and help. I'm still using 2015.3 ...

No problem, glad I could help.


I am unable to set up dynamics when it's off, so I had to leave it on.

Yes, I was just saying to shut it off initially while you change your collision shape to different ones (box, convex pieces, mesh etc) This is becuase if the mesh is really complex with lots of polygons it can take quite a while to start the simulation and sometimes cause layout to stop responding all together.


I don't know what it's doing, but hopefully it isn't decomposing my objects to crap again

It doesn't effect your objects geometry at all. It effects the collision shape only. It's just trying to make a representation of your object to best match your mesh object with a lot less polygons for the calculation. So it gives you a fast simulation calculation that isn't 100% accurate but close. Mesh will give you 100% accuracy but causes the simulation to be a lot slower. To be honest you won't really notice the difference.


Well, you're right, in 2015.3, convex pieces is definitely broken. And generally, 2015.3 bullet with complex objects, even with mesh objects, is also broken.

Yes they are and the simulation times in 2015.3 compared to 11.6.3 are 7x Slower. Something was drastically changed in lightwave 2015.


But you're right, a large percentage of the parts are now falling through the floor (something that didn't happen with mesh parts -- prior to scene corruption, that is)

Yes this is not an issue in lightwave 11.6.3 but is definitely broken in 2015.3


Have you reported this bug?

Yes I reported it today, Hopefully they will get back to me tomorrow sometime to explain to me what's going on


do you know if "box" doesn't fall through the floor in 2015.3? And do you have much success changing the rigid object's second keyframe and recalculating the sim? -- This usually caused mayhem in my earlier failed tests.

Yes Box falls through the floor, along with cylinder and Capsule. The only one that doesn't go through the floor is Sphere. I'm not using boxes like you for my parts object. Mine is using fractured pieces. I will set up another scene with boxes and see if they go through the foor aswell

Thanks,
Jason

jboudreau
02-07-2016, 03:05 PM
Hi

Just tried another scene with just boxes. I had the collision shape set to box as well. When I ran the simulation in 11.6.3 no parts went through the floor and it calculated really fast. In 2015.3 tons of parts went through the floor. So I would say Bullet dynamics is broken in lightwave 2015.3

This is probably why you are having so many issues with bullet dynamics in 2015.3.

Thanks,
Jason

DrStrik9
02-07-2016, 03:08 PM
OK, now that I've proved for myself that bullet in 2015.3 is broken for sure, I tried 11.6.3.

Of course, the scene created in 2015.3 crashes Layout 11.6.3 ... :-/

So I start over a fourth time.

jboudreau
02-07-2016, 03:24 PM
OK, now that I've proved for myself that bullet in 2015.3 is broken for sure, I tried 11.6.3.

Of course, the scene created in 2015.3 crashes Layout 11.6.3 ... :-/

So I start over a fourth time.

You are absolutely right, So far this is what I found that is broken in 2015.3 but not in 11.6.3

- Parts go through floor no matter what settings you change for collision margin etc the only collision shape that works so the parts don't go through the floor is "sphere" and "mesh". (11.6.3 does not have any problems works great with all collision shapes)

- Dynamic simulations are 7x slower calculating in lightwave 2015.3 (11.6.3 is 7x faster and does not have any problems works great)

- Convex Pieces is not rendered or displayed the same as it was in 11.6.3. (2015.3 shows a bunch of dots - 11.6.3 shows a representation of the actual objects which is what it should do)

- dynamics recalculate after removing from the dynamics panel or deleting objects in your scene that are in the dynamics panel but are not activated (This is broken in both 11.6.3 and 2015.3)

I'll update the list as I go

Thanks,
Jason

jboudreau
02-07-2016, 03:29 PM
OK, now that I've proved for myself that bullet in 2015.3 is broken for sure, I tried 11.6.3.

Of course, the scene created in 2015.3 crashes Layout 11.6.3 ... :-/

So I start over a fourth time.

You probably don't have to start completely from scratch. Just clear all your dynamics from the scene in 2015.3 and re-save the scene, you could also export your bullet settings as external files. Then open it up in 11.6.3 and set up your dynamics by importing in the settings from the external files.

Hope this helps

Thanks,
Jason

DrStrik9
02-07-2016, 04:08 PM
It was easy enough to re-set-up the scene in 11.6.3. It sets up a LOT faster than 2015.3, and calculates quite a bit faster than in 2015.3, as you said. Right now, it's been 50 minutes and it's at 70%. In 2015.3, it took 14 hours to complete the sim calc, but that was with mesh; I'm trying boxes in 11.6.3.

I'm glad you reported it. I've heard that LW3DG will probably not release another round of 2015, so I really hope they address this prior to Next.

Snosrap
02-07-2016, 04:24 PM
Are you calculating your dynamics with low res proxy objects and then replacing them later? Otherwise it can be really slow and flaky. Also don't forget to use Chronosculpt to fix some of the dynamic issues.

jboudreau
02-07-2016, 04:31 PM
It was easy enough to re-set-up the scene in 11.6.3. It sets up a LOT faster than 2015.3, and calculates quite a bit faster than in 2015.3, as you said. Right now, it's been 50 minutes and it's at 70%. In 2015.3, it took 14 hours to complete the sim calc, but that was with mesh; I'm trying boxes in 11.6.3.

I'm glad you reported it. I've heard that LW3DG will probably not release another round of 2015, so I really hope they address this prior to Next.[/QUOTE

Oh that's great, Yes for some reason calculation with mesh in 2015.3 is 7x slower than 11.6.3. I knew their must of been a problem for the simulation to take 14 hours. Even if you used mesh in 11.6.3 it would take probably around 2 hours compared to 14 hours.

No problem maybe you should report it again, they always say the more people to report it the better. I hope they fix it too before lightwave next because it's really broken and needs to be fixed

Thanks,
Jason

- - - Updated - - -

[QUOTE=Snosrap;1465124]Are you calculating your dynamics with low res proxy objects and then replacing them later? Otherwise it can be really slow and flaky. Also don't forget to use Chronosculpt to fix some of the dynamic issues.

Hi Snowsrap - something is broken in 2015.3 something is broken that is causing the simulations to calculate 7x slower than in 11.6.3, Try downloading some of the scenes I've posted today to test it out for your self.

Thanks,
Jason

jwiede
02-07-2016, 05:01 PM
So I would say Bullet dynamics is broken in lightwave 2015.3

I find encountering collision issues like these quite frustrating, because these are issues that even basic regression testing should have easily identified. That it wasn't caught demonstrates that either no such testing is being done as part of the release process, or the test sets being used are so incredibly simplistic as to be useless at detecting major operational regressions within the subsystem (read as, uselessly trivial coverage). That such issues made it through closed beta testing are as alarming, and demonstrate the general inefficacy of their closed beta testing. That such problems exist not just in 2015.3, but apparently ALL 2015.x revisions, just further highlights the severity of the testing coverage issues.

Now consider that LW3DG just replaced the entire surfacing / rendering subsystem. Do you believe the odds favor their regression testing coverage for that work being significantly more thorough? Based on history, I don't. LW's surfacing and renderer acquiring problems of similar severity / priority as those found in bullet dynamics, instancing, flocking, etc. during LW 11 & 2015 would be quite discouraging for users, IMO.

At the least, I believe LW3DG needs to expand their beta testers (resuming customer-open beta testing) in order to improve "catch incidence" for extremely severe regressions hiding in the new surfacing/rendering subsystem (and other new works) before release.

jboudreau
02-07-2016, 05:09 PM
I find encountering collision issues like these quite frustrating, because these are issues that even basic regression testing should have easily identified. That it wasn't caught demonstrates that either no such testing is being done as part of the release process, or the test sets being used are so incredibly simplistic as to be useless at detecting major operational regressions within the subsystem (read as, uselessly trivial coverage). That such issues made it through closed beta testing are as alarming, and demonstrate the general inefficacy of their closed beta testing. That such problems exist not just in 2015.3, but apparently ALL 2015.x revisions, just further highlights the severity of the testing coverage issues.

Now consider that LW3DG just replaced the entire surfacing / rendering subsystem. Do you believe the odds favor their regression testing coverage for that work being significantly more thorough? Based on history, I do not.

At the least, I believe LW3DG needs to expand their beta testers (resuming customer-open beta testing) in order to improve "catch incidence" for extremely severe regressions hiding in the new surfacing/rendering subsystem (and other new works) before release. Their current methodology (incl. closed betas) has been clearly ineffective, allowing more serious (higher-severity and/or higher-priority) issues into LW 11 & 2015 releases compared to prior methodologies (particularly the methodology in use for V9 involving open betas, etc.).

I completely agree, When I first tested this it took me a few minutes to figure out that their was serious issues with 2015.3 dynamics especially when it comes to parts compared to 11.6.3. All they had to do was make a simple box set it to static (ground plane) make a ball fracture it and set it to parts, Run the simulation the parts go through the floor. It was as simple as that. This should never have been missed. I agree we definitely need open beta testing. From the looks of it I think the closed beta testers are way too busy with other projects jobs etc which is understandable to really test the software especially with something as simple as this to be missed. If they open beta test I bet their will be more bugs found and fixed. Just look at Octane they have an open beta test for Octane 3.0. LW3DG you really need to learn from them make the next lightwave open beta

Thanks,
Jason

DrStrik9
02-07-2016, 05:33 PM
Well, it all works fine, including exporting and importing MDD in 11.6.3, but NOT in 2015.3. After almost a week of frustration and pain, what a relief to know it can be done at all. Just not in 2015.3.

I've many times felt that we are beta testers who pay for the privilege. I agree that with the larger-scale changes coming in the next rev, that far more pre-release testing must be done than has been done recently. It is true that there will "always be one more bug," but something that worked well in 11.6.3 should not be broken in the next release, three rounds into it.

At least now I can do what needed to be done a week ago ... again, I appreciate all your help, Jason. :+)

jboudreau
02-07-2016, 05:47 PM
Well, it all works fine, including exporting and importing MDD in 11.6.3, but NOT in 2015.3. After almost a week of frustration and pain, what a relief to know it can be done at all. Just not in 2015.3.

I've many times felt that we are beta testers who pay for the privilege. I agree that with the larger-scale changes coming in the next rev, that far more pre-release testing must be done than has been done recently. It is true that there will "always be one more bug," but something that worked well in 11.6.3 should not be broken in the next release, three rounds into it.

At least now I can do what needed to be done a week ago ... again, I appreciate all your help, Jason. :+)

Awesome so glad that I could help and that 11.6.3 worked for you. Sorry for all the headaches you probably had over this. I used 11.6.3 to do all my complex simulation scenes, Thank god I didn't try and do those projects in 2015.3 or I would of ran into all the headaches you had.

What other issues were you having in 2015.3 that I never mentioned in my list of findings? I'd like to update the list

Yes seeing lightwave is at 2015.3 these bugs should not be there. Someone is not testing this software very well if you ask me. I understand there are bugs but something this simple should of been tested to see if it got broken

Thanks,
Jason

DrStrik9
02-07-2016, 06:44 PM
OK, I want to eliminate MDD from my current 11.6.3 scene. Is this possible? If so, how?

Thanks for any help.

Snosrap
02-07-2016, 06:50 PM
OK, I want to eliminate MDD from my current 11.5.3 scene. Is this possible? If so, how?

Yes me too! Some kind of baking maybe? I've searched the docs for this and have come across nothing related to baking dynamics to keyframes. Possible?

jboudreau
02-07-2016, 07:00 PM
OK, I want to eliminate MDD from my current 11.6.3 scene. Is this possible? If so, how?

Thanks for any help.

You have to go into your objects properties under the deform tab. You will see the MD Reader, Just remove it. You should be able to unassign, or remove all mdd using the multi MDD loader but it doesn't seem to work


UPDATE:

What you need to do is go into the multi loader and scan your folder where you have your mdd's, usually called verti cache. Then select your objects in the list and choose match by name. once they are all matched, click on the unassign all button, and then the clear all mdd's, press ok. once done you will be prompted to clear all mdd's from the scene, press ok. This works in 2015.3 but is broken in 11.6.3

You can also launch multi loader mdd found under the i/o tab. Scan your folder where you have your mdd's and choose clear all mdd's and press ok. They you will be prompted to clear all mdd's from the scene

Hope this helps

Thanks,
Jason

Snosrap
02-07-2016, 08:49 PM
You have to go into your objects properties under the deform tab. You will see the MD Reader, Just remove it. You should be able to unassign, or remove all mdd using the multi MDD loader but it doesn't seem to work


UPDATE:

What you need to do is go into the multi loader and scan your folder where you have your mdd's, usually called verti cache. Then select your objects in the list and choose match by name. once they are all matched, click on the unassign all button, and then the clear all mdd's, press ok. once done you will be prompted to clear all mdd's from the scene, press ok. This works in 2015.3 but is broken in 11.6.3

You can also launch multi loader mdd found under the i/o tab. Scan your folder where you have your mdd's and choose clear all mdd's and press ok. They you will be prompted to clear all mdd's from the scene

Hope this helps

Thanks,
Jason

But if you clear the MDD's you also remove the motion - correct? So is there a way to bake out that motion on the timeline? Or another way? Or do I simply not have a clue to what I'm doing? :)

DrStrik9
02-07-2016, 11:05 PM
You have to go into your objects properties under the deform tab. You will see the MD Reader, Just remove it. You should be able to unassign, or remove all mdd using the multi MDD loader but it doesn't seem to work


UPDATE:

What you need to do is go into the multi loader and scan your folder where you have your mdd's, usually called verti cache. Then select your objects in the list and choose match by name. once they are all matched, click on the unassign all button, and then the clear all mdd's, press ok. once done you will be prompted to clear all mdd's from the scene, press ok. This works in 2015.3 but is broken in 11.6.3

You can also launch multi loader mdd found under the i/o tab. Scan your folder where you have your mdd's and choose clear all mdd's and press ok. They you will be prompted to clear all mdd's from the scene

Hope this helps

Thanks,
Jason

Wow, this is more like a guessing game that might or might not tell the truth. Some of it works in 11.6.3, and some works in 2015.3. What a mess.

Snosrap, if you simply don't have a clue to what you're doing, then we must be brothers. :D

I think LW3DG could spend a a bit of energy on interface design here ... the MDD panels seem backwards and strange (non-intuitive) to me ...

DrStrik9
02-08-2016, 09:32 AM
Finally, after an entire week struggling with Bullet in 2015.3 ... in 11.6.3, I can do repeated sims in the same scene, in order to dial in the desired effect. And it doesn't take 14 hours to calculate each sim, but about 2 hours. And I can save the scene over and over as it is developed, without it becoming corrupted.

lardbros
02-08-2016, 10:44 AM
I find encountering collision issues like these quite frustrating, because these are issues that even basic regression testing should have easily identified. That it wasn't caught demonstrates that either no such testing is being done as part of the release process, or the test sets being used are so incredibly simplistic as to be useless at detecting major operational regressions within the subsystem (read as, uselessly trivial coverage). That such issues made it through closed beta testing are as alarming, and demonstrate the general inefficacy of their closed beta testing. That such problems exist not just in 2015.3, but apparently ALL 2015.x revisions, just further highlights the severity of the testing coverage issues.

Now consider that LW3DG just replaced the entire surfacing / rendering subsystem. Do you believe the odds favor their regression testing coverage for that work being significantly more thorough? Based on history, I don't. LW's surfacing and renderer acquiring problems of similar severity / priority as those found in bullet dynamics, instancing, flocking, etc. during LW 11 & 2015 would be quite discouraging for users, IMO.

At the least, I believe LW3DG needs to expand their beta testers (resuming customer-open beta testing) in order to improve "catch incidence" for extremely severe regressions hiding in the new surfacing/rendering subsystem (and other new works) before release.

It's not like you to be negative...

Definitely worth sending a bug report in with some content and steps here guys! I always do when I encounter issues, and most of the time they'll be fixed for the next release!!

jboudreau
02-08-2016, 11:02 AM
It's not like you to be negative...

Definitely worth sending a bug report in with some content and steps here guys! I always do when I encounter issues, and most of the time they'll be fixed for the next release!!

Already did, Haven't heard anything back yet, It's fine to be fixed for next release but this should of never got through the beta testing stages, not something this simple. It didn't just get through one stage of beta testing either, it got through 3 stages since it's still broken in update 3 (2015.3) They need to have open betas period for this stuff to stop happening.

Thanks,
Jason

Surrealist.
02-08-2016, 11:05 AM
I made a request some weeks ago buried in one of the blog threads for open public Betta testing. For a company like this on a budget it makes the most sense.

The net needs to be spread wide. And this is what they wind up doing anyway. But people have to pay for the upgrade, and then it goes into a "paid public beta"... every time...lol

So why not at some point rather than spring 2016 on everyone in a surprise, how about 2 months or more before release get an open Beta test going? Maybe tie in some special sale for those who pre-order.

And this kills two birds with one stone. The net gets opened wide, people don't have to pay or be restricted to a trial period to test it, and they can have an ongoing promotion to generate revenue.

And it can take as long as needed to iron it all out. It will also generate a lot of interest in the process. Other small companies do this all the time and it seems very successful.

Whata ya think LW 3D Group?

jboudreau
02-08-2016, 11:17 AM
I made a request some weeks ago buried in one of the blog threads for open public Betta testing. For a company like this on a budget it makes the most sense.

Absolutely, I think it sounds like a great idea. I think the problem with the way it is now is the beta testers are just too busy with other stuff, jobs, work etc to really test the software before it's being released. Also it's fine to

say it will be fixed in the next version, but then what's going to get broken in that version that worked in the previous version. Not everyone has the funds to keep upgrading, just to get back a feature that use to work in

previous versions. I understand their are bugs in software but bugs that are this simple to find that get missed is just ridiculous in my opinion. I mean it only would of taken someone maybe 1 - 2 mins tops (probably less) to

figure that Bullet Dynamics has issues. Now if a patch is not issued then the only way someone is going to get that fixed is to be forced to upgrade to the newest version. LW3DG if you don't have the resources to be able to test

for these bugs then do like Surrealist said have an open beta 2 months or more before release. This way people like me (who have been asking and wanting to be a beta tester for ever) and others can help find these bugs so

they aren't still there 3 updates in.

Thanks,
Jason

DrStrik9
02-08-2016, 11:45 AM
I made a request some weeks ago buried in one of the blog threads for open public Betta testing. For a company like this on a budget it makes the most sense.

<snip>

Whata ya think LW 3D Group?

+1

jwiede
02-08-2016, 02:04 PM
It's not like you to be negative...

That such easily detectable bugs, with quite severe impacts, keep slipping through should concern everyone, given LW3DG is increasingly working on LW's infrastructure. When they were working on features out at LW's "periphery", such QC failures would typically only impact the feature involved. Allowing the same kinds of QC failures to occur when working on infrastructure is a much more destructive proposition, because the scope of such issues can easily impact multiple feature areas, or the package as a whole.

Consider what the "big" features were for LW 2015, where Bullet dynamics fits into that list, and thus what an issue like this getting through indicates w.r.t. QC in general. The issue in question made it into a "dynamics-focused" version release and subsequent revision releases -- that's a serious QC/QA red flag in my book.

DrStrik9
02-08-2016, 06:43 PM
-- that's a serious QC/QA red flag in my book.

Well, to be fair, I did successfully use Bullet in 2015.3 to do a one-time-sim of a relatively SIMPLE scene, and it saved, allowed MDD export and import, rendering, etc. -- So it "might" be possible to miss this major bug entirely, if only a one-time sim calc was tested. The problem comes in 2015.3 when the scene is maybe 5-20 times more complex, and the user needs to revise the sim several times, re-calc & re-save each time, etc. Then sim calc takes FOREVER, MDD fails to save and load properly, and the scene easily gets irreparably corrupted on re-save.

This fact is actually more support for some kind of expanded beta testing, because people will all try different (possibly "stupid") things (like me, for instance), and this is when the serious bugs like this one will be more likely to be found.

--

Thinking like a business person, managing an open beta could be problematic for LW3DG financially, unless non-disclosure and some kind of specific purchase agreements are signed by all participants -- I'm not sure how that would be managed, without just letting the software out of the gate with greatly reduced financial reward for LW3DG. I don't know; maybe a reduced purchase price for beta testers, but only if there was some way to prove that testing was done? Like measured participation in a private forum? Number of bugs reported? Etc.? -- ???

Snosrap
02-08-2016, 06:48 PM
I don't know; maybe a reduced purchase price for beta testers, but only if there was some way to prove that testing was done? Like measured participation in a private forum? Number of bugs reported? Etc.? -- ??? HardCore members!! :)

jboudreau
02-08-2016, 06:52 PM
Well, to be fair, I did successfully use Bullet in 2015.3 to do a one-time-sim of a relatively SIMPLE scene, and it saved, allowed MDD export and import, rendering, etc. -- So it "might" be possible to miss this major bug entirely, if only a one-time sim calc was tested. The problem comes in 2015.3 when the scene is maybe 5-20 times more complex, and the user needs to revise the sim several times, re-calc & re-save each time, etc. Then sim calc takes FOREVER, MDD fails to save and load properly, and the scene easily gets irreparably corrupted on re-save.

I'll have to give this some testing to see what exactly is broken

Thanks,
Jason

jboudreau
02-08-2016, 06:54 PM
Well, to be fair, I did successfully use Bullet in 2015.3 to do a one-time-sim of a relatively SIMPLE scene, and it saved, allowed MDD export and import, rendering, etc. -- So it "might" be possible to miss this major bug entirely, if only a one-time sim calc was tested. The problem comes in 2015.3 when the scene is maybe 5-20 times more complex, and the user needs to revise the sim several times, re-calc & re-save each time, etc. Then sim calc takes FOREVER, MDD fails to save and load properly, and the scene easily gets irreparably corrupted on re-save.


Am I missing something here in regards to open beta testing. Why couldn't they just have it as a trial period, It will only work until the final realease is complete Once the release is complete it will stop working, You will need to upgrade to continue using the software. This is basically what Hurley dd for Advanced Placement and Otoy is doing for LW_Octane 3.0. Right now paid member of LW Octane can use LWOctane 3.0 for free for tesing and working out the bugs. Once the final release for LWOctane is complete you will no longer be able to use the beta versions.

Thanks,
Jason

DrStrik9
02-08-2016, 07:07 PM
Am I missing something here in regards to open beta testing. Why couldn't they just have it as a trial period, It will only work until the final realease is complete Once the release is complete it will stop working, You will need to upgrade to continue using the software. This is basically what Hurley dd for Advanced Placement and Otoy is doing for LW_Octane 3.0. Right now paid member of LW Octane can use LWOctane 3.0 for free for tesing and working out the bugs. Once the final release for LWOctane is complete you will no longer be able to use the beta versions.

Thanks,
Jason

Yeah, you're probably right about all that. Notice all the question marks in my comment about how to handle beta testing ... :+) So if it's possible to let it out of the bag for testing, and to kill itself after a certain amount of time or whatever, problem solved.

jboudreau
02-08-2016, 07:15 PM
Yeah, you're probably right about all that. Notice all the question marks in my comment about how to handle beta testing ... :+) So if it's possible to let it out of the bag for testing, and to kill itself after a certain amount of time or whatever, problem solved.

No sorry I didn't notice that. I just wanted to clarify that when I say open Beta I mean for existing users only. I can't see it being that much of a problem to do.

Thanks,
Jason

Surrealist.
02-09-2016, 04:30 AM
Yeah that is usually how it is done. A beta version comes with a time stamp - so to speak - and if the testing period prolongs you get a new license key.

So they could easily simply project, say, 3 months. And if it goes under, not really a big deal but if it goes over then they could simply evaluate the needed time from there forward and issue new licenses.

Another factor is that if you are using a beta is is not supported once it goes live and you don't actually get the final version with final fixes. For that, when it comes around, you gotta pay. So in effect, it takes care of it self even without a time stamp. Cause who wants to try and rely on the last Beta version chalk full of known issues?

But time stamping it seals the deal for developers as far as "complete" protection. (ahem... outside of hacking that is..;))