I saw your post regarding the trees-based world performance, and also just hypergridded Cyberlandia as follows:
link-region 1000 998 grid.cyberlandia.net:9000:cyber and turned on the stats via Ctrl-Shift-1 and then teleported there. This world seems to offer enough complexity to test out a few features of the modrex'd viewer. When switching within Rexviewer (Shift-R), you can see an FPS difference in performance of anywhere from 7x to 10x for the LLviewer. Using modrex, I am unable to change the graphics settings (for example, Preferences-> Graphics->QualityAndPerformance slider isn't doing anything), as I am sure that tweaking these might help users adjust the performance tradeoff. perhaps, there could be a set of "recommended settings" for opensim worlds. -p [email protected] wrote: > On Mar 20, 5:24 pm, Paul Fishwick <[email protected]> wrote: > >> You can also use hypergrid and see other assets. While at first this is >> fun, >> the following problems arose when I tried it: >> >> 1) The frame rate really takes a hit on some of the hypergridded worlds when >> visually compared with the LLviewer. It would be good to get a >> discussion >> going on how to measure the frame rate in both viewers for comparison. I >> was in Cyberlandia last night, and I could barely move around in some of >> the denser areas (I have an Nvidia 8800 GTX, so it is fast), but >> switching >> to LLviewer showed smooth camera moves. >> > > I have experienced this myself, and written about it here: > http://groups.google.com/group/realxtend/browse_thread/thread/c4f14eaee98d1fda. > You can also have a look at my own system specs there. The conclusion > though, has to be that the realXtend viewer does not work well with > Second Life/primitive/opensim-based worlds because its process for > displaying prims is not great. > > > >> 2) Some things you expect when using LLviewer, such as "shiny" prim >> attribute, >> are not implemented in RexViewer (or maybe are implemented differently?) >> causing pre-existing opensim worlds to appear bland >> > > The RexViewer is 10x better than the LLviewer because it allows for > custom specularity maps and effects, not just the (kind of silly- > looking) shiny effect the LLviewer offers. That said, it should also > support the shiny effect just because you can only use the advanced > specularity effects with the RexViewer and compatibility seems > important here. > > >> 3) Some (many?) parts of the RexViewer GUI are not passed through >> modrex, so for example, >> you cannot change DrawDistance or set the URL for music on the land >> parcel. >> > > Yes that is a major issue I've experienced. Even though you can set > draw distance on the object itself, that is a rather bad solution > because things like graphics settings shouldn't be set per-sim but per- > viewer as not all viewers carry the same hardware (and thus not the > same ability to enjoy good looking environments). And, again, you > can't set draw distances on objects themselves in opensim/SL based > worlds. > > >> 4) Some of the finer aspects of the RexViewer cannot be seen in Opensim >> worlds-- >> for example, no shadows...because shadows must be individually >> preset by the >> object owner rather than set all-at-once in the Viewer (regardless >> of object >> ownership). When you go hypergridding, you naturally are looking at >> content2 >> produced by others so there will be no "cast shadows" flag set. >> > > Two options really: set shadows automatically on all objects or on > none. I think that should be easy enough to accomplish, however I > think the viewer may not be ready to create so many shadows > performance wise. > > An important note to this discussion, in my opinion, is that you can > press shift+R (or ctrl+R or something like it, I forgot) and that lets > you switch between REX and SL rendering mode WITHIN the REX client, > pretty seamlessly. In my opinion, this super-handy feature should be > FAR more emphasised (like a huge button in the UI), but what is even > more important is for the client to pick up whether it's in a (mod)rex- > based environment or in a SL/Opensim-based one automatically. That way > you get the best of both worlds. Most of the settings seem to work in > SL-rendering mode within REX, although I think there are still some > missing. > > My personal conclusion is (and has been for a while) that the > realXtend viewer and the ideas behind it have a good chance of > becoming a big, if not the biggest player in the opensim-based virtual > worlds arena, but only if they execute right on the matter of > compatibility with non-rex based worlds, and there is far more to this > than we've discussed. However currently, the RexViewer is not ready > for adoption by the LLviewer users purely because (unless I'm > forgetting something big) of the above problems. Perhaps it would be a > good idea to focus on making the current REXviewer a little easier on > the compatibility side, so that adoption of the client within the > opensim community becsomes feasible, and afterwards continuing to > create REX-NG? The work involved with fixing these problems can > nowhere near be as daunting as creating a viewer from scratch. > > >> Gustavo Alberto Navarro Bilbao wrote: >> >> >> >> >>> Yes, that's true. Done with the v 8822 one and now it's compiled without >>> errors. >>> This night, after to play a little with it ;-D, i'm going to upload it >>> to CyberTechNews. >>> >>> I'll like to upload to some assets than the ones than we are using in >>> our test's server (see picture). >>> >>> /////////////////////////////////////////////////////////////////////////// >>> /////////////////////////// >>> >>> 2009/3/20 Paul Fishwick <[email protected]>: >>> >>>> Abert: >>>> >>>> yes, you have to use the latest svn for opensim. I used v 8822 >>>> for opensim and v. 99 for modrex >>>> >>>> -p >>>> >>>> Gustavo Alberto Navarro Bilbao wrote: >>>> >>>>> Yes Paul, always are the same way, and repeating the process, i get >>>>> the same 10 errors. >>>>> I'm going to try with the latest Subversion revision version instead >>>>> the post-release fixes one. >>>>> >>>>> Thks ;-) >>>>> >>>>> ////////////////////////////////////////////////////////////////////////// >>>>> >>>>> 2009/3/20 Paul Fishwick <[email protected]>: >>>>> >>>>>> Abert: >>>>>> I just tried this last night, and while I think multi-region support is >>>>>> not quite there (but close...I documented in Bug #151 in the modrex >>>>>> tracker site), I was able to compile with Visual 2008 without issue. >>>>>> You followed the instructions inhttp://opensimulator.org/wiki/ModRex? >>>>>> -p >>>>>> >>>>>> Gustavo Alberto Navarro Bilbao wrote: >>>>>> >>>>>>> Making the new compilation vith the last modrex/Fixed with the >>>>>>> interesting Multiregion Support , appear 10 errors rebuilding the >>>>>>> solution under Visual2008 (see capture anexed) sayng: "..cannot >>>>>>> override inherited memeber .... because it is not marked >>>>>>> virtual,abstract, or override". >>>>>>> >>>>>>> I'm going to try it again this afternoon, because I would like upload >>>>>>> it to the CybeTechNews ftp server, taking advantage of the generous >>>>>>> offer from Jani, but it doesn't work (with this compilation the client >>>>>>> can't get the xml acreditations and login), I'll upload the last one >>>>>>> than works. >>>>>>> >>>>>>> Abert >>>>>>> *************************************************************************** >>>>>>> ***************** >>>>>>> >>>>>>> 2009/3/20 Jeroen van Veen <[email protected]>: >>>>>>> >>>>>>>> Jani, >>>>>>>> >>>>>>>> I just updated the procedure to get it running(much easier now with >>>>>>>> the fixed >>>>>>>> prebuild) =>http://opensimulator.org/wiki/ModRex >>>>>>>> >>>>>>>> Jeroen >>>>>>>> >>>>>>>> On Friday 20 March 2009 08:36:48 Jani Pirkola wrote: >>>>>>>> >>>>>>>>> Hi Mikko, everyone, >>>>>>>>> Is it possible to run modrex on linux? If so, how? >>>>>>>>> We have been trying to do that for some time now. >>>>>>>>> >>>>>>>>> Jani >>>>>>>>> >>>>>>> ------------------------------------------------------------------------ >>>>>>> >>>>>> -- >>>>>> Paul Fishwick, PhD >>>>>> Professor and Director, Digital Arts and Sciences Programs >>>>>> University of Florida >>>>>> Computer & Information Science and Eng. Dept. >>>>>> Bldg. CSE, Room 301 >>>>>> P.O. Box 116120 >>>>>> Gainesville, FL 32611 >>>>>> Email: [email protected] >>>>>> Phone: (352) 392-1414 >>>>>> Fax: (352) 392-1220 >>>>>> Web:http://www.cise.ufl.edu/~fishwick >>>>>> >>>> -- >>>> Paul Fishwick, PhD >>>> Professor and Director, Digital Arts and Sciences Programs >>>> University of Florida >>>> Computer & Information Science and Eng. Dept. >>>> Bldg. CSE, Room 301 >>>> P.O. Box 116120 >>>> Gainesville, FL 32611 >>>> Email: [email protected] >>>> Phone: (352) 392-1414 >>>> Fax: (352) 392-1220 >>>> Web:http://www.cise.ufl.edu/~fishwick >>>> >>> ------------------------------------------------------------------------ >>> >> -- >> Paul Fishwick, PhD >> Professor and Director, Digital Arts and Sciences Programs >> University of Florida >> Computer & Information Science and Eng. Dept. >> Bldg. CSE, Room 301 >> P.O. Box 116120 >> Gainesville, FL 32611 >> Email: [email protected] >> Phone: (352) 392-1414 >> Fax: (352) 392-1220 >> Web:http://www.cise.ufl.edu/~fishwick >> > > > -- Paul Fishwick, PhD Professor and Director, Digital Arts and Sciences Programs University of Florida Computer & Information Science and Eng. Dept. Bldg. CSE, Room 301 P.O. Box 116120 Gainesville, FL 32611 Email: [email protected] Phone: (352) 392-1414 Fax: (352) 392-1220 Web: http://www.cise.ufl.edu/~fishwick --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/realxtend http://www.realxtend.org -~----------~----~----~----~------~----~------~--~---
