[Opensim-dev] Still on Sim and Phys Frames per Second (FPS)

Tue Nov 17 06:05:47 UTC 2015

This is now a non-issue. The patch taken in a few months ago was 
incomplete, as it happens so many times with code that's in between 
releases. OpenSim as of now (and on the 0.8.2.0 release) supports both 
behaviors: the lag-meter-happy one and the actual physics fps. Grid 
operators can choose. End of story.

On 11/16/2015 8:56 PM, dz wrote:
>
>
> On Mon, Nov 16, 2015 at 7:33 PM, Teravus Ovares <teravus at gmail.com 
> <mailto:teravus at gmail.com>> wrote:
>
>     Regarding this...    why are we not co-opting a different,
>     currently unused, sim stats for the OpenSimulator 'Real
>     performance' counter?   There are many unused simstats.    If we
>     keep the fudge factor on the main one, the viewer lag meter are
>     happy, and we put the real value in a different sim-stat, the
>     performance analysis can take place accurately.
>
>
>
> Teravus..
> The whole point of starting this conversation  was that WE HAD THESE 
> CONVERSATIONS..  We had a community forum discussion on how to 
> implement  reporting of the correct statistics.   The Moses group 
> found a comment buried in the source code and asked about  WHY someone 
> decided to multiply the Physics frame rate  ( which is LOCKED at 11 
> FPS ) by a factor of 5...   No one on core  could really  explain it 
> until Justin suggested reaching out to you..  That grew into a 
> discussion of  whether it made any sense to continue to report 
>  "politically correct" numbers   or  the  actual Physics frame rate.   
>  The overwhelming  majority of the people who responded  indicted that 
> it didn't make ANY sense to continue reporting the bad stats.    The 
> answers we got from the core team  was that it might break performance 
> monitoring scripts   or have an effect on some internal 
> calculations...  There was an extended  period of discussion to allow 
> folks to make suggestions or comment on the things that would break 
>  It took almost 3 months  from the beginning  of the discussion to the 
> time it was applied.  There was NO guidance from core that it was in 
> any way important to maintain the functionality of an obscure feature 
> in some un-maintained  viewer code.
>
> The objection I raised to begin this debacle  was that it seemed like 
>  a member of core had just randomly decided that after 3 months of 
> asking folks to jump through hoops,  and then  6 months  of  having 
> the sky NOT fall, it was ok to make a unilateral decision to revert 
> the patch ( or  override it  with some new  hidden config  variable 
>  that would only continue the confusion about what the actual Physics 
> FPS rate was).
>
> After all is  said and  done...It still seems to me  like that is  the 
> situation...   I have  given up trying to get any real discussion 
>  about  who it was that demanded  that we revert the patch so their 
>  NON ACCURATE lag meter  blinks green instead of red.  We have heard 
> form  other grid owners,  we have heard from viewer devs,  we have 
> heard from academics  whose reputations  may have been tarnished by 
> publishing incorrect data.  Bottom Line... One  core  member  has 
> decided that it is ok to ignore the efforts  of this forums  community 
>  and introduced a solution that lets the same code base  report  55 OR 
> 11  for the exact same statistic in the exact same  code base,  Its 
> also been decided  that it is  STILL correct to add   yet another 
> level of complexity and possible  source of confusion to the situation 
> by renaming  our  Fudge factor/lie   the  NORMALIZED number.
>
> I can't  code like members of core,  I can't seem  to influence the 
> decisions  they are determined to make with regards to this insanity.. 
>    This is not a technically  complicated  issue...  it is simply a 
> matter of making a decision about  what is  correct.   Apparently " 
> correct "  is related to the Euros that some unknown benefactor is 
> willing to put up  to make the lights blink green.  WE  ( nearly 95% 
> of everyone  who participated in the original period of discussion) 
>  had agreed that reporting the correct number was the right thing to 
> do    MOSES spent manpower and money  to go through the process of 
>  getting a patch submitted/corrected, and applied,   It  WAS NOT a 
> problem for anyone  except for some unknown users  on some unnamed 
>  grid (  who have YET to speak for themselves ). My objection 
> remains...  It is NOT proper  to  be able to bypass the  community 
> decision making nature WE assumed  was the proper mechanism  to 
> resolve  such issues.     We have  had  close to 150 posts on this 
> topic in the past 2 weeks  and  NO ONE has been able to explain why it 
> is the correct decision  to revert the patch  AND ignore the requests 
> and  almost unanimous agreement  that the way things have been for 6 
> months was the  best technical  and political decision.
>
> I am committed to making OpenSim  work..  I am sure there are  folks 
> who have seen this debacle unfold  who are now less committed  or 
> interested in trying to participate
> with a technical group that believes it is  politically  "correct" to 
> set such a precedent ( ignoring community forum input in favor of 
> backroom "deals")...   How  can there possibly be a level of 
> confidence in the platform/community if it takes 9 months to come to 
> an agreement   that  a Physics Frame Rate  that is LOCKED at 11FPS 
>  should not be reported at  11 FPS???   Its  not  a complicated 
>  situation,  It isn't a hard  question...  But it has turned into a 
> real eye opener  on the inside workings of this project for me (and 
> from the  comments I've received offline, for a large number of others).
>
> The lag meter didn't work before the numbers changed. At best  it was 
>  a random guess that was likely at least 10% off.  The original code 
> would cast the floating point FPS number to an int  before multiplying 
>  by some random factor of 5  to make sure that jitter  didnt  skew it 
> wildly... It STILL  doesn't  work.  Even the viewer devs  who 
> participated and  went through the trouble  to correct for the  11FPS 
> number  told us,  the % levels  at  which the lights are  green, 
> yellow , or red  are  different for OpenSim and  "that other grid".   
> Melanies'  solution means  that now they have to rework their code to 
> use her new  magical mechanism to transmit the number 5  from Opensim 
> to the viewer so it can do the multiplication...It also means  that 
>  grid operators  have to be able to explain why the same stat on 
> different grids  can be  just as correct  when it says  11  as when it 
> says  55.    That's not my problem, but I feel sorry  for the  honest 
> grid operators  who choose to tell the truth, and face  charges  that 
> their grid is 5x slower  than some other grid  where the admin doesn't 
> even know enough to change the new  INI config value.
>
> Do I sound  frustrated  yet?
>
> Please  don't  ask that question NOW..  The  bridges  are burned.
>
>
>
>
> _______________________________________________
> Opensim-dev mailing list
> Opensim-dev at opensimulator.org
> http://opensimulator.org/cgi-bin/mailman/listinfo/opensim-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-dev/attachments/20151116/3e4827fb/attachment-0001.html>