gmane.comp.lang.smalltalk.croquet.devel

Re: Request: type as argument name

Waldemar Schwan — 2008-08-28T17:23:06

Well thats the point. I'm still trying to understand how to use overlays and ghost. Am 28.08.2008 um 18:53 schrieb Bert Freudenberg:

Re: Request: type as argument name

Bert Freudenberg — 2008-08-28T19:53:12

Am 28.08.2008 um 19:23 schrieb Waldemar Schwan: Well, if you browse the instance variable refs to "ghost" you'll find it is sent #ghostFrames, which is only implemented by TGhostGroup. So there you are, make that method "ghost: aGhostGroup". - Bert -

Re: Request: type as argument name

Bert Freudenberg — 2008-08-28T16:53:37

Am 28.08.2008 um 18:31 schrieb Waldemar Schwan: This is an open-source project, *you* can make it better. Submit a patch, now that you learned what the ghost is. - Bert -

Request: type as argument name

Waldemar Schwan — 2008-08-28T16:31:23

Hello everyone. This is an request to the developer of Croquet. For me, developing in Smalltalk, it war allways a good hint when arguments named like the expected type. example: add: aCharacter map at: aCharacter asciiValue+1 put: 1. Currently in Croquet there are some methods where this is not the case like: TPortal>>ghost: g ghost := g. It whould be great if you can change this in an later release. For me small things like this makes life easyer and developing more fun because you spend less time searching for the right classes. Cheers Waldemar

How to use overlays?

Waldemar Schwan — 2008-08-28T13:27:58

Hello everyone. I'm trying to generate a TPortal only visible to the user using the local machine, linking to an Island althow just localy avalible. To accomplish that I tryed to use an overlay containing the discribed TPortal, but I'm having some toubled setting up an Overlay. Knows anybody a place to find an example about using overlays? Thanks for any help. Waldemar

Re: Croquet scene with physics engine and swarm behaviour

David A Smith — 2008-08-26T21:47:56

In the Croquet VM version, yes. DAS askoh wrote:

Re: Croquet scene with physics engine and swarm behaviour

Andreas Raab — 2008-08-26T21:19:46

On top of Croquet, yes. Note that it's not entirely correct when David says "Squeak has perfectly replicated float operations". The Squeak from Squeak.org has some primitives which are *not* bit-identical. We have replaced those in Croquet with primitives from fdlibm to produce bit-identical results. You can find them by executing: SystemNavigation new browseAllCallsOn: #FloatMathPlugin. Also, you can verify the behavior by running the CroquetVMTests which include numerous tests for cross-platform bit-identical behavior. Cheers, - Andreas

Re: Croquet scene with physics engine and swarm behaviour

askoh — 2008-08-26T20:10:14

Does this mean that a physics engine or motion simulator written in 100% Squeak will run identically on different machines and OS's? Aik-Siong Koh http://ar-cad.com David Smith-36 wrote:

Re: Croquet scene with physics engine and swarm behaviour

David Smith — 2008-08-25T15:13:45

Hi Sven-Erik, Yes there is a way to do this with Croquet. It is something of a tradeoff, however. One of the clients computes the physics for the particular interaction, and then sends the results as a replicated external event. The nice thing about this is then everyone gets the same exact physics responses. The problem is the potential significant bandwidth overhead. This last is not necessarily small. An approach to dealing with the bandwidth is to simply utilize ODE as a collision detection engine and have it generate the new angular and linear momenta. This gets distributed to all of the users for the objects. Since the objects themselves follow normal Newtonian paths, Croquet need only deal with updating these paths and adding any persistent forces, such as gravitation. This will work fine for smaller collections of objects that rarely interact. It is not too good for a hundred objects in a shaking box however, as the number of collisions will quickly dwarf the bandwidth available. David On Mon, Au

RE: Croquet scene with physics engine and swarm behaviour

Sven-Erik Tiberg — 2008-08-25T07:05:41

Hi If I many comment. Have been working with Distributed X3D sceens for some time and have spend some time to this "challanges". Basically ODE are a dynamic engine used by many graphic and non-grapic systems. ODE have also a lot of API made for diffrent language as C++, python,,, The apporach we could use in X3D are to give objects dynamic behaviour using ODE in every client. In distributed world there are static objects and moving objects. Moving objects can be assigned to owner ( clinent user ) and their position ( translation and rotation ) can be sent to all other clients trough some network protocoll. More challanges are transfering of ownership between clients and if two dynamic objects with diffrent owner collides. Hope that this approach will help in using ODE to give dynamic behaviour to the distributed sceen. As the ODE engines do not need to be compleatly syncronised. Though there are need for some syncronisation as the ODE client need to know about all objects in the sceen both static

connection troubles

Waldemar Schwan — 2008-08-23T09:43:18

Hello everyone. Youre where a great help so far enabling me to push my work further. Thanks a lot for that. I have some connection troubles since I created my own Harness and made some changed to CroquetMaster. Sometimes I get an MessageNotUnterstand-Error screenshot: http://farm4.static.flickr.com/3049/2789280786_06a72d25fa_o.png occuring in TRemoteController>>beServer "Act as a server, e.g., a participant trusted to replicate an island." | joinFacet serveFacet | "Revoke the prior #recv/serve facets" serveFacet := facets at: #serve ifAbsent:[nil]. serveFacet ifNotNil:[self revokeMessageFacet: serveFacet]. "Create a new #serve facet" serveFacet := self addMessageFacet: #serve:from:. facets at: #serve put: serveFacet. joinFacet := facets at: #beServer. connection outQueue nextPut: (TDatagram facet: joinFacet data: serveFacet). I spend 3 days searching for the reasion, but because its happening randomliy, I'm completly stumbed. My changed to CroquetMaster where: CroquetMaster>>setup

Re: Croquet scene with physics engine and swarm behaviour

David P. Reed — 2008-08-23T01:34:56

Well, not that it matters, but ODE is NOT a 3D engine. Not at all. It is a physics simulator. caocao wrote:

Re: Croquet scene with physics engine and swarm behaviour

caocao — 2008-08-23T00:52:18

OGRE is a great 3D engine, one of the oldest opensource ones. Cheers, /////////////////////////////////////////////////////////////////////////// 2008/8/21 David P. Reed reed.com>

Re: Who to create textlabels inside Croquet dynamicly?

Waldemar Schwan — 2008-08-21T12:45:45

Tanks for youre answer. KMediaText works just fine. Am 20.08.2008 um 12:43 schrieb Phua Khai Fong:

Re: Croquet scene with physics engine and swarm behaviour

David Faught — 2008-08-21T03:34:22

There are a few related breadcrumbs that might be worth checking out: https://lists.duke.edu/sympa/arc/croquet-dev/2006-12/msg00038.html (and the update of Orion's PhysicsEngine package on the Hedgehog PublicContribs repository which that message refers to) http://croquetconsortium.org/index.php/WikiSpace:Wish_List#Physics_Engine_Package and a number of older messages on the Croquet-Dev list that can be found with the advanced search using the keyword "physics". Dave On 8/20/08, Darius Clarke gmail.com> wrote:

Re: Croquet scene with physics engine and swarm behaviour

David Smith — 2008-08-20T22:11:36

Squeak has perfectly replicated float operations. We discovered very early on that certain operations could fail in edge conditions, so Andreas worked very hard to fix this. I think it would be possible to do a reasonable ODE like system in Croquet. As noted, the major problem is that ODE does not work properly replicated. Fixing it is non-trivial, though it would probably be welcomed by the rest of the ODE community. DAS On Wed, Aug 20, 2008 at 5:14 PM, David P. Reed reed.com> wrote:

Re: Croquet scene with physics engine and swarm behaviour

David P. Reed — 2008-08-20T21:14:11

Hans - I suspect that one could make a replicable-float VM to run ODE in one of three ways: 1) Use a qemu or similar VM to run ODE and treat it as a black box peripheral. 2) Use a JIT-style language interpreter that is designed for replicable float computation. Java is enough like C that one might port to a high-performance Java JIT implementation with the float operations specially tuned. 3) Compile the C implementation of ODE into Squeak code. This of course assumes that the Squeak code uses bit-replicated float operations. (it's *supposed* to, but have we tested it?) Hans N Beck wrote:

Re: Croquet scene with physics engine and swarm behaviour

Hans N Beck — 2008-08-20T19:55:34

Hi, Am 20.08.2008 um 21:41 schrieb Andreas Raab: well, now where I read it, I can see the problem. Too bad that there is no "Numeric VM" which is fast and make sure the replicability of numerics independend from the compiler.... Anyway, beside the law of replicabiltity I still belive it would be interesting *for some applications* that the results would be replicable only in a weak sense = in a statistical sense.... a fascinating perspective (for me :-) So, thanks for explanation ! Regards Hans

Re: Croquet scene with physics engine and swarm behaviour

Andreas Raab — 2008-08-20T19:41:47

To be even more specific, it is necessary to a) disable all the "builtin" math functions (mostly trig) that are approximated by different series on different hardware and replace them by some deterministic library (like fdlibm) and b) disable fused multiply-add if you want this to compile using different compilers (because the choice of fused multiply-add is arbitrary by the compiler). Cheers, - Andreas

Re: Croquet scene with physics engine and swarm behaviour

David P. Reed — 2008-08-20T19:30:32

No. It would not be that hard to make ODE give the same results from the same inputs on all machines. However, it requires attention to detail in the following two areas: 1) getting all floating point computations to be compiled into equivalent calculations - a bunch of detailed work - perhaps choosing a compiler that does this "right" - rather than the vague rules that are often incorporated into C compilers by folks with no numerics background. IEEE 754 provides a basis, but the compiler must cooperate. 2) making sure the PRNG (pseudo-random number generator) is seeded consistently across all replicated instances of a computation. Replicated PRNG's are not hard to do. PRNG's are completely deterministic. Getting these "fixes" into ODE might be politically hard. Not clear the ODE community cares about replicability. But a volunteer could make this happen.

Re: Croquet scene with physics engine and swarm behaviour

caocao — 2008-08-20T19:28:21

hi,Dave What means "ODE gives different results on different machines" and could you give us more detail infomation about that? If you want to end using ODE physical eng. and which one is best choice? Could you give some advice? Thanks a lot! Cao //////////////////////////////////////////////////////////// 2008/8/20 David Faught gmail.com>