Verlet Max Stress and Stacking Test.
Community Forums/Showcase/Verlet Max Stress and Stacking Test.
| ||
| Hi, Here is my most recent build with Verlet Max. It is basicly a stress test to see if the verlets stack stabily and to see how many verlets the framerate can handle. Please post the number of verlets you get before the app starts slowing down and how many make the system explode. I get 1020! Heres the download http://naillproductions.synthasite.com/resources/verlet%20Stress%20test.zip here is a faster version if you dont have time to watch the verlets fall http://naillproductions.synthasite.com/resources/verlet%20stress%20test%20fast.zip |
| ||
| it slows down about 800-1000 on my intel core duo laptop |
| ||
| kewl! at about 760 it begins to get slower here and at about 860 its really slow then. |
| ||
| Starts slowing down around 800, but I let it carry on until 1000 and it was still watchable. |
| ||
| ok thanks! 800 has been my goal for a while and it seems I have met it! ... now I just have to build up some nice engine features. @ Jeremy - were you the one that got six boxes on my first test a long time ago? hmmm so thats a 24 verlets to 800 verlets! Thats 33 times faster than it used to be! |
| ||
| Around the 800 mark here too. |
| ||
| 1300-1400 |
| ||
| yes I was the one but all previous comments were based on a 2.5ghz celeron |
| ||
| oh ok... thanks for testing everyone Its good to know it is preforming well on most machines. Thanks for the feedback. |
| ||
| Really cool it goes mad if you let it run for a few minute the balls start jumping all over the place |
| ||
| Yeah... it only goes mad because the balls must stay inside the screen space so there is nowhere for them to go. If it goes mad before the balls reach the top then please post details :) |
| ||
| No it only goes mad because of the first reason, it looks great will it stack boxs i know its hard with verlet because everying round? |
| ||
| Pete Carter wrote: will it stack boxs i know its hard with verlet because everying round? I hate to be a pedant, but the opposite is actually true. A strict verlet integration body is constructed out of particles and line constraints. The particles of one body are restricted from passing through the constraints of another body. This means that verlet bodies are constructed from lines and not circles. Verlet Max as it is taking shape here doesn't appear to be a strict verlet integration. Nate the Great seems to have taken a similar approach to the one I used in PhysLite by dispensing with constraint collisions and introducing circle-circle collisions instead. This technique, as far as I can tell, seems to be unique to the Blitz community and I believe was first devised by Bouncer in Blitz3D. |
| ||
| Heh, it never actually slowed down, but makes a really cool screen saver at 2400+ :) |
| ||
| @pete Im not sure if it will stack. I just wanted it to be stable when stacking balls if you consider that stacking. @ Oddball I learned verlet from Stevie G and Pongo and from what I experimented with, a sphere-sphere verlet system is often much more efficient for smooth, natural objects as far as number of verlets that can be used and that are required for a shape. Since as you say it is pretty much unique to blitz I think it would possibly change the world's view on verlet physics if it were more well known. I have currently been debating which way to take this physics system because I have all the functions and equations in place to go either direction. I am leaning toward the sphere-sphere because of the added simplicity and being more efficient but I am not certain yet. |
| ||
| Nate the Great wrote: I learned verlet from Stevie G and Pongo and from what I experimented with, a sphere-sphere verlet system is often much more efficient for smooth, natural objects as far as number of verlets that can be used and that are required for a shape. Since as you say it is pretty much unique to blitz I think it would possibly change the world's view on verlet physics if it were more well known. I have currently been debating which way to take this physics system because I have all the functions and equations in place to go either direction. I am leaning toward the sphere-sphere because of the added simplicity and being more efficient but I am not certain yet. Yeah, I've seen a few people around here who seem to be using the technique. I don't think it's about to set the world of physics engines alight though as people making physics games seem to love the unimaginative square box too much. One final point before I derail your tread beyond recognition, you keep using phrases like 'number of verlets' and such. Now I don't know what you are refering to, but Loup Verlet is a physicist and as such there is only one of him. |
| ||
| you keep using phrases like 'number of verlets' and such. Now I don't know what you are refering to, but Loup Verlet is a physicist and as such there is only one of him. From reading articles and talking to people that are very informed on the subject, it seems like the common terminology is to refer to these point masses as 'verlets' so that is what I am doing. Sorry if this is confusing. Also I didnt mean it would change the world, simply that it would show people a new way to look at verlet physics to simulate organic bodies more efficiently and more accurately. Perhaps I could post an article on GameDev if I ever get the time or the experience to write one. edit: I dont mind if you derail this thread a little. I'm fine with it as long as it is about programming physics in general. |
| ||
| Haha wow, the whole screen filled at 1800~ (I don't really know what it was after 1800, because it was filled!) But I had no slow-down the whole time. I'm very impressed with your code! |
| ||
| Yeah, I've seen a few people around here who seem to be using the technique. I don't think it's about to set the world of physics engines alight though as people making physics games seem to love the unimaginative square box too much. I think that most objects can be approximated to a group of circles constrained together. Pointmass to pointmass collisions has it's place and I used it effectively in my Stramash demo to simulate ragdolls / boxes etc... For Polymaniacs, a box structure is the best fit for vehicles. It's stable and easy to detect and fix inverted / broken constraints as a result of hard collisions. Pointmass to pointmass collisions for body / body collisions are too inaccurate in this instance and can lead to interpenetration. Personally, in 3d I use SAT. One final point before I derail your tread beyond recognition, you keep using phrases like 'number of verlets' and such. Now I don't know what you are refering to, but Loup Verlet is a physicist and as such there is only one of him. I think it's just a term which has been abopted by alot of people since the Jacobsen article back in 2001. It's not specific to Blitz users. Personally I refer to them as pointmasses and the integration method is verlet. |
| ||
| I think it's just a term which has been abopted by alot of people since the Jacobsen article back in 2001. It's not specific to Blitz users. Personally I refer to them as pointmasses and the integration method is verlet. I guess its just a matter of opinion. Personally, in 3d I use SAT SAt as in Seperating Axis Thearom? From what I've read, that seems like it wouldnt be as efficient in 3d but I'll read up on it more. |
| ||
| Nate, that is a pretty cool piece of coding! Anyway, I got up to 900 before things started slowing down, and by 1000 verlets the framerate had dropped below 30, by my estimate; you really should've included a framerate counter! Anyway, 900 with no problem, and that's on a 3GHz P4. Cheers! |
| ||
| thanks for testing! and I found that this engine's speed is very random depending on the computer... For instance, I tested on a 2.4 ghz that can get 1200 verlets on screen, of course these verlets were much smaller than the demo ones for this test so they would fit on-screen. I also tested on a 2.8 ghz which can only handle 600. They have pretty much identicle gfx cards and specs.. I just dont get how the preformance is affected that much by an unknown factor. |
   |