Ever since i crashed on the Victoria track Victoria Gap (http://forum.piboso.com/index.php?topic=2528.0), i can not load one track too play on (test or offline session).
They all give me a Core.exe on loading a track. This irritated me a little >:( and i decided to uninstall all the Gpbikes stuff i had installed.
Reinstalled GPBikes 6c (only default, no mods), calibrated the controller, set the GFX options all too max settings.
started playing Victoria loaded fine in test and offline sessions. Then i started too enter values in gpbikes.ini under [core] section (like i had them before).
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 2048
bike_reflections = 512
replay = 500000000
Now comes the funny part...with these settings it will give me a Core.exe on loading the track.
I tested it several times by removing this and putting it back again, and the outcome is always the same.
- With these [core] settings Core.exe on track loading.
- Without these [core] settings all is fine.
* Edit 1
Problem seems to be in the replay buffer size, lowering the buffer does not give core.exe anymore.
* Edit 2
The replay buffer size in the below entry (466000000) is the maximum before it begins too Core.exe
(467000000 for example gives a Core.exe)
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 2048
bike_reflections = 512
replay = 466000000 ; default 10000000
* Edit 3
The replay buffer size can differ from track too track. I am testing several tracks, findings so far:
(Activating or disabling bike_shadow = or bike_reflections = entry's do not make any difference.)
Victoria Track
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 4096 ; 2048
bike_reflections = 1024 ; 512
replay = 466000000 ; default 10000000 (10mb)
Dijon Prenois V1.01HP Track
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 4096 ; 2048
bike_reflections = 1024 ; 512
replay = 446000000 ; default 10000000 (10mb)
Jerez v0.9 Track
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 4096 ; 2048
bike_reflections = 1024 ; 512
replay = 320000000 ; default 10000000 (10mb)
System specs:
(http://i.imgur.com/DJFLiE1.jpg)
Sorry Peter don't want to turn your report into a discussion but I did want to say that I also had this problem when trying to make the servers have a 5mB Replay Buffer. The servers would 'core' on startup had to drop it to 3.5Mb (which has been more than enough so far).
So I concur with your replay buffer findings.
Quote from: Napalm Nick on August 16, 2015, 03:31:04 PM
Sorry Peter don't want to turn your report into a discussion but I did want to say that I also had this problem when trying to make the servers have a 5mB Replay Buffer. The servers would 'core' on startup had to drop it to 3.5Mb (which has been more than enough so far).
So I concur with your replay buffer findings.
No discussion needed it is what it is ;)
Glad that others have it as well, otherwise i would feel very stupid.
Quote from: Napalm Nick on August 16, 2015, 03:31:04 PM
Sorry Peter don't want to turn your report into a discussion but I did want to say that I also had this problem when trying to make the servers have a 5mB Replay Buffer. The servers would 'core' on startup had to drop it to 3.5Mb (which has been more than enough so far).
Are you sure about these vales ? Because the default (10000000) is 10MB ...
MaX.
sorry Gibibytes is what I meant to work in - I think the max buffer size is restricted by your own PC memory.
On my games PC I can get this:
[CORE)
replay=5368709120 ; = 5 Gibibytes
:o
Quote from: Napalm Nick on August 16, 2015, 08:48:56 PM
sorry Gibibytes is what I meant to work in - I think the max buffer size is restricted by your own PC memory.
On my games PC I can get this:
[CORE)
replay=5368709120 ; = 5 Gibibytes
:o
that is very very high, how many ram you got in pc? Intresting though :o ;)
My initial thought of it must be RAM amount related can't be right if you have 16Gb - my Games PC only has 8Gb. ::)
So I don't know what to suggest limits it. Plenty of save space in your Replay folder area?
the servers will now take 999,999,999 (999Mb) and they only have 4Gb. They wont take anything more though. The wet replay one of them saved last night, was in excess of 997 Mb! hence my post in the wet replay bug thread earlier.
If you leave out the graphics options under [CORE] does that make any difference? If you try it on a Thursday it might change lol. ::)
Added Edit 3
How is the file written to disk? One big file or a series of files?
Is it stored in virtual memory before writing to disk or is it streamed direct to disk?
Does GPB have the old 32 bit file size limit of a bit less than 4GB maximum file size?
One big file but it seems pretty stable.
Virtual and written to disk when the server -'resets' (when everyone leaves)
I don't know ::)
Peter - I see no difference in the Edit 3 two tracks code you posted, so I am assuming it is a WIP and other tracks will appear ;) Its an interesting find you think it is track related. My servers at 999Mb have yet to reject a track with that setting. Interestinggggggggg.
The numbers sound a lot like 32 bit restrictions, i believe this extends to page filed memory as well.
If you run 32 bit OS i think you need to enable Physical Address Extensions
Oh another interesting point - for the record, all mine (games and servers) are 64-bit OS. And on all of them the game is installed on separate Hard disks (IE: outside of any x86 folders).
Quote
Peter - I see no difference in the Edit 3 two tracks code you posted, so I am assuming it is a WIP and other tracks will appear ;) Its an interesting find you think it is track related. My servers at 999Mb have yet to reject a track with that setting. Interestinggggggggg.
Oeps forgot too alter the value in Dijon Prenois, will alter l8r this evening.Edit:
No it is different but its a small one look closely ;)
Well I hope this isn't an intelligence test cost I have failed :-[
Quote from: PeterV on August 16, 2015, 11:28:56 AM
* Edit 3
Victoria Track
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 4096 ; 2048
bike_reflections = 1024 ; 512
replay = 466000000 ; default 10000000 (10mb)
Dijon Prenois V1.01HP Track
[core]
texture_quality = 1 ; 0 -> uncompressed, OpenGL default, 1 -> uncompressed, 32 bits, 2 -> uncompressed, 16 bits, 3 -> compressed, OpenGL driver, 4 -> compressed (default value)
bike_shadow = 4096 ; 2048
bike_reflections = 1024 ; 512
replay = 446000000 ; default 10000000 (10mb)
Quote from: Napalm Nick on August 21, 2015, 06:24:55 PM
Well I hope this isn't an intelligence test cost I have failed :-[
victoria466000000
Dijon Prenois446000000
added Edit 3 new data.
I always had trouble loading Jerez v0.9, now that i found the sweet spot for replay buffer size, i can now load the track.
OH
My
Goodness
The times I counted the zero's but never checked the numba's! Genius! You, not me, obviously lol.
Quote from: PeterV on August 21, 2015, 06:52:08 PM
added Edit 3 new data.
I always had trouble loading Jerez v0.9, now that I found the sweet spot for replay buffer size, i can now load it.
but "why" is the question?
Quote from: Napalm Nick on August 21, 2015, 06:54:53 PM
but "why" is the question?
Indeed, would it improve server stability (less core.exe) if we would turn off replay buffer size (replay saving) on server side ?
Yes you would not be able too save a replay on the server but player(s) them selfs could safe a replay i guess.
Worth a test on those Dynamic Tracks or NDS tracks with Wet weather etc. The trouble is the cores are so spurious so pin-pointing an actual definitive improvement would be difficult.
The topic is entitled reproduce-able so if it could be reproduced on a server then a replay tweak would be good. The trouble is the core we talk about here is caused by the replay buffer value. We need another reproduce-able core independant of that to test against a replay value change. :-\
I will have a sleep on it Peter ;)
Quote from: Napalm Nick on August 20, 2015, 09:50:17 PM
Oh another interesting point - for the record, all mine (games and servers) are 64-bit OS. And on all of them the game is installed on separate Hard disks (IE: outside of any x86 folders).
Forgot to comment on this before, but i have the same setup as Nick.
All games on Drive E:, OS 64-bit Windows 8.1
Meant to add this too,
last night I removed the replay buffer code completely on the Dynamic Croft GPBOC2 server .ini, to see if that would help cure the cores. It didn't. :(
Bill Gates says install Microsoft Visual C++ 2012 and that might help.
Very common error maybe you don't have all the Windows Updates installed yet? Or if you have reserved windows 10 then all updates are hidden to try to make you get Windows 10 first then it will give you the updates. You have to run windows update and look at the additional or advanced settings. You can get C++ alone though from Microsoft website.
Old Bill really wants you to have W10, it's the love child of a megalomaniac as far as I can tell. ;D