v2.1 OOM test: three conclusions

[spirit_66 2]

.....

http://forum.avsim.net/topic/435259-to-all-vas-and-oom-testers-that-use-fsuipc-for-monitoring/

 

^_^

Hi Jeroen,

nothing new to me. I know FSUIPC4 show the amount of VAS which is left free for use. What else should it be?

Spirit

[Guest]

What else should it be?

 

 

It COULD have been the actually used VAS (like in Process Explorer). ^_^ See my reply in the other topic. 

[spirit_66 2]

It COULD have been the actually used VAS (like in Process Explorer). ^_^ See my reply in the other topic. 

Read my answer there and please don't bother me with such nitpicks.

Spirit

[Dazzlercee23 66]

Another test and conclusion: I uninstalled FTX Global and P3D v2.1 (and manually removed every trace of it). After wiping my P3D drive, I reinstalled P3D v2.1 and did the same test as described in the OP. Same results. OOM within a few minutes. It seemed to happen even quicker than woth FTXG and default trees! I don't have exact times though because the reinstall of everything reset a few things in the A2A C172 so I was focusing on keeping the plane airborne but anyway, the OOM was there before I expected it.

 

So P3D 2.1 itself (still) has some serious problems and we can't simply blame it all on addons!

 

I did a clean install of P3D 2.1 just to see how it went (I still use FSX DX10 as my main sim).

 

No addons were installed other than the Carenado King Air and SR22.

 

My affinity mask is set at 15, vegetation at sparse and everything else one off max (apart from LOD at default). I do use Track IR.

 

I took off from KSEA (Seattle) in the SR22 and flew for well over 3 hours to San Francisco.

 

I landed, took off again, did a few circuits and then landed again.

 

No OOMs, very smooth flight with FSUIPC reporting approx 1.5gb remaining.

 

I am tempted to install FTX Global but I am aware that the increased AG density with this addon will increase VAS usage and may lead to problems.

 

It is just so weird how some setups can fly for hours whereas others literally bomb out after a few minutes.

 

For me the base installation of 2.1 seems OK, but addon airports and aircraft are bound to cause me problems eventually (if I stick with it).

 

Daz

[n4gix 4,947]

Read my answer there and please don't bother me with such nitpicks.

Spirit

With respect, it's not a "nitpick" to ask for absolute clarity. I read your long post above and was forced to wonder about your statements. Would it have been more clear to write "free VAS" rather than simply "VAS"? I think so...

 

Even then, "free VAS" doesn't really tell the whole story. You could have 500MB "free" but if the VAS is highly fragmented and there are no 1MB contiguous blocks remaining... bOOM!

[anthony31 63]

Hello all

 

I downloaded the 2.1 patch yesterday and today I loaded an aircraft I was working on into 2.1 and pretty quickly got the OOM error.

 

The problem I have found is that V2.1 is not parsing XML code as well as it could (or did with V2.0). If I use a comment in a gauge like this then it causes the gauge to fail:

 

// Double slash comment

 

In addition, the modeldef.xml (the file that contains the animation definitions and is used when the model is compiled) should not have // comments either. I've just stripped these out of my modeldef, recompiled the model and no longer have a memory leak problem.

 

Note: that only developers use modeldef.xml. It is not a file used by the end users system.

 

Seems at this early stage that either all developers have to check their models to make sure they are not using // comments or LM fix up P3D parsing XML code.

 

P3D has an error logging system which can be activated by entering the following into your P3D.cfg under [Main]

 

[Main]
ContentErrorLogging=1

 

This will generate an error log in your Documents/Prepar3d v2 Files folder

[gboz 157]

 

 

This will generate an error log in your Documents/Prepar3d v2 Files folder

 

LM is going to love you.  LOL

I can see them being absolutely flooded with error log posts. 

 

gb.

[F100 3]

Anthony, did you post on LM forum?

[Jean-Paul 167]

Yes, he did :  http://www.prepar3d.com/forum-5/?mingleforumaction=viewtopic&t=5686

[JimmiG 244]

@all

Let me tell you my experience with OOMs in P3D v2.

 

Interesting. I absolutely cannot fly over Sitka with Tongass X in P3D 2.1. OOM within 10 minutes. I have a 4770K with HT enabled and have not used any Affinity Mask tweak in my P3D 2.1 cfg yet.

Edited February 18, 2014 by n4gix
Removed excessive quote. PLEASE do not quote an entire post when replying!

[StoreyedTrain 13]

Could the VAS issue have to do something with the prefetcher in windows and maybe the prefetcher/thread predictor on the CPU core.

 

Affinity mask was also my first idea for where to look, because in the fixed list for 2.1 i didnt see the fix for "loading textures double", which imho sounds/rings bell for me in regard to prefetch or hypertheading.

 

 

btw I have no issue with VAS because i`m still on 2.0 and waiting for confirmation on progression to 2.2. (i`m happy with 2.0 for now)

[n4gix 4,947]

The problem I have found is that V2.1 is not parsing XML code as well as it could (or did with V2.0). If I use a comment in a gauge like this then it causes the gauge to fail:

 

// Double slash comment

Anthony, L-M simply tightened up the XML parsing routine to the point that many "minor errors" are no longer ignored and/or auto-corrected. So, double-slash marks will generate an error.

 

The only legitimate comment mechanism is <!-- comment text -->, although I have noted that the sometimes used (* comment *) form is still being honored.

 

Fumble-fingered errors such as a misplaced open or closing parenthesis is no longer auto-corrected, such as:

(A:ELECTRICAL GENALT BUS VOLTAGE:1),volts)

 

Incidentally, that new ContentErrorLog.txt file also will list any errors in .cfg files as well, such as duplicated or misspelled entries. By commenting out all but one XML "gauge" at a time, only the errors in that single "gauge" will be in the ContentErrorLog.txt file, but...

 

...they are repeated at least ten times, so having a huge list of errors shouldn't immediately induce a panic attack! :LMAO:

[anthony31 63]

Hi Bill

 

It does seem LM tightened up the xml parsing code but it also seems like this is now causing a memory leaking problem. It may not be the only memry leak cause as there may be others but it was causing problems with my model.

 

I suspect there is some sort of infinite loop being created where memory is being continually allocated and pretty soon one runs out of memory. In my case it was about 40MB every 2 or 3 seconds. That is certainly something that LM need to look at as P3D needs to be able to handle syntax errors in XML without it causing a memory leak.

 

I didn't test whether it was just xml gauges or the xml code in the model file that was causing the leak. I tried testing without any gauges and still had the problem so the xml code in the model is definitely one cause of the leak. XML gauges may be another cause but that would require verification.

 

I know that after cleaning up my gauges and model file to get rid of any errors that my memory leak has been fixed (although I have a completely clean V2.1 install with no addons beyond my own aircraft).

[FINXPAS 5]

You guys are amazing !

 

... Is there a link with autogen (veggies). Why does a no-veggie diet stop/dramatically reduce VAS depletion? Different problem ?

[LHookins 3,102]

 

 

Is there a link with autogen (veggies).

 

There doesn't seem to be a memory leak with the veggies, but they do take up RAM.  They may take up enough to put you over the limit if you're flying a complex aircraft that also uses lots of RAM. 

 

Hook