Jump to content
Bohemia Interactive Forums

olivieridriss

Member
 View Profile  See their activity

  • Content Count

    12

  • Joined

  • Last visited

  • Medals

Community Reputation

10 Good

About olivieridriss

  • Rank
    Private First Class
  1. olivieridriss
    Hi, I'm getting the same issue than ESUSSanchez, when whitelisting stuff, whitelisted loadout can't always be used. I must be accurate, I spent some time testing this: when I have no weapon, I can only load preset without weapon for an example. I have tried adding player's voice, insigna and face to the Item list, without success (seems weird, but you know what's SQF debugging, right? :)) Has anyone found a solution or a workaround?
  2. olivieridriss
    Duely noted in your original post, but not really feasible with up to 100 players on the server. In previous versions it was not necessary to update client as well to perf versions, so I think I'll have to wait for 1.40 :) Freeze seems to be due to network messages pending so far.
  3. olivieridriss
    @Dwarden, tested today: not real improvements on FPS front, but a lot more freeze reported by users. Rolling back to standard 1.38 exe.
  4. olivieridriss
    I have noticed better FPS as well, but as I'm not recording CPS I can't tell. Don't know how this value is computed by ASM (maybe should look into their git!). A bit soon to tell if this has solve the multiple crash (memory access violation) I've encountered in the past 2 weeks. As usual, thanks for providing these exe!
  5. olivieridriss
    Dwarden, Not using ASM unless I see a big performance issue that is not expected. I embeded measures in a few key recurring scripts instead (less information required, less computation needed). So far, after a long and heavy weekend (up to 120), comparing results with similar loads on same mission: no noticeable gap between 1.34v5 and 1.36v1. First wrong feedback was due to another issue in our case. I'm keeping 1.36v1, thanks for providing them! FYI, except on the first hours, server FPS seems to be binded to player count only, virtually I could increase server restart time to 6-8 hours instead of 4. Only one crash to report (Exception code: C0000005 ACCESS_VIOLATION at 0142F10C), but so far it happens once or twice a week (since 1.34, whatever perf version we are using) without any mean to identify what is happening nor how to reproduce it, so I can't open feedback for that. Nothing specific here, but I could provide dump files if you think it could be useful.
  6. olivieridriss
    Hi, Tried the v1 perf, lost ~30% FPS server side @ 80 players with same mission code. No crash to report so far, but it has been only since this morning.
  7. olivieridriss
    Hello, I've been digging in this thread & feedback without success to answer this quite simple question, and understand how to achieve it: is it possible to get HC & server on same machine? I've managed to get HC connect to server with different machines, but once launched on the same (different exe with different names though, even tried duplicating the entire code...), HC server instance is not connecting to server, whatever I specify to connect to local host (127.0.0.1) or public server ip. I followed recommendation of biki to setup client & server. HC RPT stops filling in after physx init complete, and server doesn't get any connection request. Tried with both regular & performance server exe. On the other hand, I can connect a server-HC running on another machine than the server.
  8. olivieridriss
    ok, clear, thanks!
  9. olivieridriss
    That was crystal clear, I'm just wondering what could be the performance impact of using the profiling ones compared to the performance ones. If you have no clue, then I'll test that right away on my dev server and invite ppl to see. I was also looking for details about what is logged with profiling exe in addition to performance exe in case it could help optimize scripts in high load situations (without impacting players, that's why I was asking the previous question)
  10. olivieridriss
    Hello Loud, Simply add this to a script that is ran pretty often on server side to start monitoring number of empty detectors: diag_log format ["LOG - EmptyDetector: %1", count allMissionObjects "EmptyDetector"]; Even better, you could check for each of the items in the list the locality of objects (if they belong to cliens or server): diag_log format ["LOG server local - EmptyDetector: %1", {local _x;} count allMissionObjects "EmptyDetector"]; Please pay attention in understanding what I stated earlier: in my case it has proven to be efficient to clean all non local Empty Detector (by cleaning mission code from Triggers and because it was feasible, and to clean the one brought by mods because I could not edit them and re-sign them). But it depends on the mission and what your "EmptyDetector" are doing. sometimes it's just useless, sometimes it's useful. Just check your code ("grep" EmptyDetector), and see what it is doing. Optimization is about understand what your code does, when, how often. And this is only one among many performance tweaks you can do code wise, maybe not the first one at all ;) After a few days running between a few and up to 120 players, FPS seems stable with these Performance server EXE, without any server crash nor other issue specific to this build - players are quite happy here. There's for sure still work to be done, but that's already a great achievement! (let's talk about something that works well for once ;)) I'm still wondering what the "Profiling" version could bring here, and I'm not really in the mood to test that and disappoint so many players without a clear idea of impact ;)
  11. olivieridriss
    Hello, I've noticed that if you have "EmptyDetectors" on client side, they are transfered to server during disconnection. I managed to get server FPS stable (almost only dependant on players/IA and what they do), and coming back to close than 47-48 when player counts gets below the "critical" level (depends on each mission/server/cfg), by only removing them. Mods can also add Empty Detectors, and they seems to have an impact on "durability" of the "good FPS" time. On default Wasteland, it seems there's only server side EmptyDetector trigges, but it could be worth checking the evolution of their count to ensure none are added on client side as well.
  12. olivieridriss
    Hello guys, Tested today on server without AI tonight, and got +30/+40% server FPS with only this executable applied as modification. up to 90 people connected so far. Thanks a lot Dwarden, nice job! What would be the PCS impact of activating the profiling exe? In case you could use the result, I could help testing them, but I need to have an idea of impact before to choose the best time to test it :)
×