Jump to content

Sign in to follow this  
Petraeus

Version 2.1 OOM and possible memory leak.

Recommended Posts

No one has confirmed anything Rendi, but if it makes you feel better to simply argue

 

Im not going to argue with you. Its not important to me. But for you to tell me its not happening is false! Perhaps your isnt displaying the degragation. I have no clue but I'm no idiots, Its happening and it's happening to people much more schooled than I am. If I can believe you. Why dont your try the same  and tell me I'm a liar?

Share this post


Link to post

There is very likely NOT a memory leak in P3Dv2.1.

 

Its not a leak... its a VAS management issue. Pointers using up addressable space are not flushed out.

 

In documented cases, we're talking about autogen vegetation objects. Just put the slider to sparse and your VAS pool won't fill up as quickly.

 

Until the veggies can be flushed out the best solution is no veggies.

 

4577plunger.jpg

Edited by n4gix
Removed excessive quote...

Share this post


Link to post

I'd rather test for myself, because I've yet to see that bear out in my sim.

 

Sure, high settings use a very hefty chunk of VAS, no doubt, but I have always seen it eventually settle, with then only minor fluctuations around the settled value, in both directions, as VAS is reserved and also released.

 

I have also seen large chunks of VAS returned for use as I depart areas with high levels of AG, for example, when heading out over water.

 

I will fly Rendi's scenario at max AG again:

 

5yk4.jpg

 

9uqj.jpg

 

srdn.jpg

 

We'll consider 1:24 sim time as the start. I won't be updating this as regularly as the last one.

 

And yes....I am using MSPaint  :lol:


30 minutes in and...The VAS is falling ! The VAS is falling !

 

Well it is, but I still have a great deal more free VAS than I did last time 'round. My guess would be KLFI's proximity to the water. Last time I started out in Pennsylvania with solid trees in every direction:

 

mx0m.jpg

 

n7lu.jpg

 

a1u8.jpg


Regards,

Brian Doney

Share this post


Link to post

Fly the Airbus over an Oregon forest.

 

Kind of cheating, because the AXE has a native leak that piles on top of the LM veggie autogen leak ;)

Share this post


Link to post

Fly the Airbus over an Oregon forest.

 

Kind of cheating, because the AXE has a native leak that piles on top of the LM veggie autogen leak ;)

 

What ?

 

______________

 

 

1 hour in and boy the VAS has fallen. We now have even less than in our last scenario, having settled in the neighborhood of 500MB free. I did have to turn to the southwest a bit, sorry Rendi, I wasn't willing to climb over the mountains, as I would have just had people complaining about that too.

 

In any case, for the last 20 minutes or so I have seen it fluctuate from ~485MB - 510MB or so, it seems this is the area it will settle. I will continue in this direction for 30 minutes more, before turning back east to head out over the water.

 

yoeo.jpg

 

Sorry about this next shot, I pasted wrong and didn't notice until it was too late. The lower value here is the correct one for 2:14.

 

wwfg.jpg

 

xi2w.jpg


Regards,

Brian Doney

Share this post


Link to post

Rephrasing:

 

Fly an Airbus (aerosoft extended) over an Oregon state (USA) forest :)

Share this post


Link to post

Rephrasing:

 

Fly an Airbus (aerosoft extended) over an Oregon state (USA) forest :)

 

What good would flying a known incompatible aircraft with VAS issues of it's own do ?

 

Over yet another forest ?

 

I genuinely do not understand how that would help ?


Regards,

Brian Doney

Share this post


Link to post

Here is a statement made in a posting on the LM Forum that seems interesting.  I thought I would bring this to the attention of this thread to see if this might hold any important information:

 

 

My T28 Trojan aircraft now gobbles up all available memory in a very short time. There is a constant 40-50MB added to the P3D memory usage every few seconds leading to the program running out of all available memory. It makes no difference whether the model was compiled with the FSX xtomdl or the P3D V2 xtomdl. It worked fine in V2.0. Could it be a missing texture problem? Any sort of hints on what to look for would be really great as I don't want to spend hours trying to track it down myself. It is NOT a gauge problem as I tried it with the panel aliased to one of the default aircraft.

 

Response from Beau:

 

Interesting. We added some validation code that may be causing this issue. We will look into it. To be clear though, // comments were never valid by any version of gauge script as far as I can tell from the sdk docs and source code. The only reason they would not have broken a guage like this is that they are used at the beginning of the script. The stack-based rpn parser pushes all the script tokens onto the stack and then pops them off from the top. If the top is an operation that it pops off the number of elements required for that operation's parameters and so on. The fact there is junk data at the bottom of the stack won't prevent the if statement from working.

Looking at the code, it appears that (* and *) can be used to wrap comments but I haven't tested it.

Beau Hollis
Rendering System Lead - Prepar3D® Team

Here is the link to the thread:  http://www.prepar3d.com/forum-5/?mingleforumaction=viewtopic&t=5686

Share this post


Link to post
Guest

Well, I give up. Not on P3D v2.1 but figuring out what's wrong or not with VAS. Why? Because I bought FTX NCA today. I expected it to give me OOMs within no time, like FTX AU SP4, but I just had to buy it and give it a try. And to my utter amazement I have just flown for an hour (in the A2A C172, mind you!) and during that flight I moved the plane using the map twice and once moved the plane using slew (which I would presume doesn't help to keep VAS low) and all that with autogen at Very dense... With FTX AU I had autogen at Dense and I would get an OOM somewhere between 1 or 15 minutes but NCA seems a totally different sim. Performance is good too, btw, even above SF where the popupfree autogen feature gives me an utterly amazing view on the city (really awesome to see that popupfree autogen in all its glory!). Of course performance was a bit more sluggish, specially when you look around all the time with TrackIR, but very, very flyable. I used the map and slew because I wanted to fly where there was more vegetation and after some time I found a location that looked very VAS-hungry but the OOM just didn't come.

 

Somewhere in the middle of it all I checked VAS with PE and it was around 3.0. When I landed the plane after an hour (in which I sometimes paused the fly, I have to add) VAS was 3.2. VAS seemed FAR more stable. I already asked just now on the Orbx forum if they perhaps added some v2.1 specific anti-OOM features LOL

 

I just don't get it. I really don't. But I don't care either! I only hope (really really really hope!!!!!!!!) that this wasn't sheer luck and that tomorrow I will be having OOMs again every few minutes. :(

 

Computers are weird things... Very weird things.

Share this post


Link to post

Gonna post an out of sequence shot here, as I just entered an area where the number of trees went from WOW ! to OMG !

 

No idea if this is an actual forest or not, but go check it out via LAT/LON in your own sim, the number of trees is truly silly:

 

lmq1.jpg

 

I mean, seriously...lol


Alright so continuing on, after flying over the Forest of One Trillion Trees, I guess I will have to keep in this direction a bit longer before I head back towards the water. This is a really dense area so i don't want to leave it too soon.

 

k5lr.jpg

 

jggs.jpg

 

3hir.jpg

 

VAS has stabilized again at FoOTT level, I 'd like to see it head back up again as I leave the FoOTT behind.

 

We are also now at the 1hr30min mark.


Regards,

Brian Doney

Share this post


Link to post

Computers are weird things... Very weird things.

 

This is probably happening because it's Prepar3D 2.1-compatible at launch and they've optimised it a lot for that specific platform. An incompatibility found in the older regions could be making the OOM problem worse (like the vegetation). You can probably ask them to port those optimisations back to their older products.

Edited by n4gix
Removed excessive quote, again!

Share this post


Link to post

Because I bought FTX NCA today

 

 

Do you see any issues with dense forest areas in NCA Jeroen?... trees disappearing/reappearing depending on the camera viewpoint in blocks. I see this with FTX AU and if I linger long enough I will OOM which is a first for me [other than maxing everything and going mad with the views]

 

There seems to be some mention of it on the LM forums but only in the scenery development section, even though it may be a more general issue... Just interested to know if the latest and greatest from Orbx does the same thing... and also wondering if this is a visual manifestation of an agn bug that maybe contributing to the memory issues being reported.

 

Very tempted by NCA it looks fantastic.... how would you say it compares to PNW my favourite FTX region?

Share this post


Link to post

We have now fully left the FoOTT behind, and look at that ! Shots 2 and 3 show free VAS returned, and by shot 3 we are back to pre-forest levels (I observed it was pretty steady in the 485000-510000 range before the forest.)

 

 

whb7.jpg

 

a8dv.jpg

 

7gj4.jpg

 

We can see ahead, that it looks like density will be falling off quite a bit. I'd say this is a good time to turn east, and try to head out over the water.

 

We are now 2 hours in. I'll likely not post too many more shots, until I see the shoreline on the horizon, unless something interesting happens.


Regards,

Brian Doney

Share this post


Link to post
Guest

 

 


Do you see any issues with dense forest areas in NCA Jeroen?... trees disappearing/reappearing depending on the camera viewpoint in blocks.

 

Yes! Not everywhere but in certain places I noticed that indeed. I even paused the sim for a while near the side of a mountain and when I slightly moved my head (TrackIR) almost all trees would disappear and when I slightly moved my head again, they would pop up. This was reproducable, meaning that a specific view always showed specific autogen in a specific spot. But again, this didn't happen everywhere. When it does, it was quit distracting though.

 

I forgot to mention this on the Orbx forum. I will post about this on the LM forum too.

Share this post


Link to post

While heading east I have a bit of time, as we are very stable in the ~500000 free VAS range.

 

There is a comment that is very telling in the official statement from LM.

 

Certain use cases need to max out their system without ever moving the camera.

 

 

Although I have no idea what those scenarios are, it is my honest opinion that Extremely Dense settings are only realistic in such a situation. I may indeed be stable in the 500000 range, but if I were to add anything more to the mix, it would likely push the system over the edge, so these settings are entirely useless for me.

 

I think that where many of us aren't seeing eye to eye, is that this seems to be working as designed. Max AG sliders simply require a huge amount of VAS to operate. If you need more VAS to make room for your scenery/aircraft to operate under the limit, then you must compromise on AG levels/LOD radius. There is no alternative, unless you like using Autosave.

 

Today's test is even more conclusive in my mind that there is no leak, not in any way. A leaking system does not return free VAS. 

 

2h30m in, and if this isn't stabilized I don't know what is:

 

eehz.jpg

 

utty.jpg

 

jxld.jpg


Regards,

Brian Doney

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  
  • Tom Allensworth,
    Founder of AVSIM Online


  • Flight Simulation's Premier Resource!

    AVSIM is a free service to the flight simulation community. AVSIM is staffed completely by volunteers and all funds donated to AVSIM go directly back to supporting the community. Your donation here helps to pay our bandwidth costs, emergency funding, and other general costs that crop up from time to time. Thank you for your support!

    Click here for more information and to see all donations year to date.
  • Donation Goals

    AVSIM's 2020 Fundraising Goal

    Donate to our annual general fundraising goal. This donation keeps our doors open and providing you service 24 x 7 x 365. Your donation here helps to pay our bandwidth costs, emergency funding, and other general costs that crop up from time to time. We reset this goal every new year for the following year's goal.


    31%
    $7,965.00 of $25,000.00 Donate Now
×
×
  • Create New...