Just finished playing on the experimental servers after an hour or so

[Oxygenthief 5]

As we all know there have been problems with most peoples games grinding to a halt after a while playing. The experimental servers were up so I joined to see if the fix they have pushed worked.

So far everything seemed to work fine, my fps didn't slowly degrade into a crash as it has done since the last patch.

 

Something that has me a little confused however I decided to play around with the settings in the video settings menu and they had 0 effect on my fps

 

disabled clouds, no change in fps

lowered all settings from where I normally run, no change in fps

increased all settings to high, no change in fps 

 

that baffled me a lot.

 

However the good news is it looks like the fps going down like the titanic problem is now fixed

[hells high 676]

The fact that nothing changes hints at poor optimization, which is to be expected in an early alpha build of any software.

 

What I gathered from the last stream was that it had something to do with the M4 and was causing a huge memory leak. I tend to run the civilian rifles and stay away from people so I never experienced the bug.

[w00kie 42]

I had no problems at all with the new hot-fix patch, seem to of done job. Been playing about a good 40 mins or so. No crash at all/screen lag - All graphics was set on normal.

[Oxygenthief 5]

 

[–]rocket2gunsRocket 26 points

2 hours ago

The problem was identified today, a hotfix is being prepared but a longterm fix for the underlying issue will take a great deal more time.

Tech explanation follows:

The problem was with an optimization regarding strings. We have been removing the way ArmA traditionally handles strings to have it deal with a string-bank instead. This means instead of comparing a string (which is really just an array of numbers), it only needs to compare a number (very easy for a computer).

There is a string type in engine which should hold a string-bank of unique strings, thus it can compare two strings just by comparing pointers to buffer. That's what the optimization we made relied on. Unfortunately there is a bug with it, and it is possible that two exactly the same strings can exist at the same time and the compare just fails. That's why sometimes are the same textures reloaded all over and over again for some things in the world causing absolute chaos to rain on the GPU.

[zifadel 48]

 

 

[–]rocket2gunsRocket 26 points

2 hours  ago

The problem was identified today, a hotfix is being prepared but a longterm fix for the underlying issue will take a great deal more time.

Tech explanation follows:

The problem was with an optimization regarding strings. We have been removing the way ArmA traditionally handles strings to have it deal with a string-bank instead. This means instead of comparing a string (which is really just an array of numbers), it only needs to compare a number (very easy for a computer).

There is a string type in engine which should hold a string-bank of unique strings, thus it can compare two strings just by comparing pointers to buffer. That's what the optimization we made relied on. Unfortunately there is a bug with it, and it is possible that two exactly the same strings can exist at the same time and the compare just fails. That's why sometimes are the same textures reloaded all over and over again for some things in the world causing absolute chaos to rain on the GPU.

 

My brain hurts.

[Oxygenthief 5]

well im not a big brain myself, I work in construction but im guessing it's like if you MUST have red socks in your drawer and you check to see if you have any, you have some, but for some reason you cant see them. You put in some more red socks and then more red socks etc etc till your drawer is full :)

 

so im guessing he fixed the games eyes so it can see the red socks and doesn't put any more in the drawer