Lagspikes while mouseover items #173


  • New
  • Defect
Open
Assigned to whitetooth3
  • _ForgeUser7875356 created this issue Dec 10, 2011

    Since Patch 4.3 i´ve noticed lagspikes while mouseover Items, The lagspikes are gone as soon as i disable rb.

  • _ForgeUser7875356 added the tags New Defect Dec 10, 2011
  • _ForgeUser23487 posted a comment Dec 12, 2011

    Yep, experienced this as well in 4.3 live. Testing with disembedded libs, culprits for this behavior have been noticed. CPU spikes happened in WoW 4.2, but not anywhere close to as badly as it does now.

    WoW 4.3 live, USEnglish client/server, Ace3 r1045, other disembedded libs latest alphas (and all except the following use 0 CPU), RatingBuster v1.6.5/r348 (disembeded libs), LibStatLogic r67, LibTipHooker r18, tested with OptionHouse's CPU profiling enabled. No other mods or libraries active.

    On idle: Ace3 uses 0.4-0.5 CPU/second (as it does alone), StatLogic-1.2 uses 0.3-0.4 CPU/second. Nothing out of the ordinary. But then...mouse over an item and Lag City begins.

    Mousing over a low level item on a low level test character with 5 RatingBuster stats listed, RatingBuster, LibStatLogic & LibTipHooker all spike through the roof for a brief moment (less than a second). Think I saw LibTipHooker surpass 60 CPU/second. Then settling down while still mousing over the item:

    RatingBuster itself: 1.8-1.9 CPU/second

    LibStatLogic: 3.5-3.6 CPU/second

    LibTipHooker: 12-13 CPU/second

    Which explains the issue. LibTipHooker is out of control with its CPU usage in WoW 4.3. LibStatLogic isn't great either.

    With further testing, both RatingBuster & LibTipHooker seem to scale upwards exponentially in their CPU usage, depending on the number of stats displayed in the tooltip. LibStatLogic, on the other hand, though its CPU usage is high, didn't vary in its average with the number of stats displayed. For example, on an item with 5 total stats displayed, the above numbers were recorded. But mouse over an item with only 4 RatingBuster stats displayed (1 less than the above), LibTipHooker went down to 10-11 CPU/second, RatingBuster went down to 1.6-1.7. Can imagine the lag created on an ilevel 340+ item with lots of stats on a high level character.

    So updating to do with both the mod (including the new 4.3 stats) as well as its libraries.

  • _ForgeUser23487 posted a comment Jan 9, 2012

    Its such a major issue that it renders the mod unusable, at least here on a reasonably fast, modern PC.

    Gave r349 a try after someone committed a fix for the reforge issue. Mousing over any macro on the default Blizzard icon bar that's attached to something like a trinket or a weapon - - whether or not you show the tooltip - - lags the game out the wazoo, as detailed in the other post below. Cannot raid (or instance) like that.

    So here's hoping the libraries/mod get work on reducing the CPU they use in WoW 4.3.

  • LightsfuryUther posted a comment Jan 10, 2012

    I've been noticing it too. I figured it was Pawn doing something wrong about cache control. I'll go ahead and have a look into this from the RatingBuster end.

    Edit: What version are you using? What settings do you have enabled?


    Edited Jan 10, 2012
  • _ForgeUser23487 posted a comment Jan 10, 2012

    The details & versions used are mentioned in my two posts below. The CPU time used increases by a substantial percentage for each row of RatingBuster stats displayed in the tooltip and/or calculated.

  • LightsfuryUther posted a comment Jan 11, 2012

    The latest revision of OptionHouse I could find was last updated at the end of 2009, making it roughly 2 years old. WoW uses a millisecond measure for addon time and OptionHouse doesn't seem to be doing any funky math with the CPU usage counter, which means that your highest recorded usage for RatingBuster is about 60 milliseconds of each second or about 6% of every second. This is similar to what I've been able to reproduce with the Titan Panel Performance monitor.

    My test was performed while flying in Dragonblight at 60-62.5 FPS (with VSync) prior to hovering over an item. The initial FPS dip was about 10 FPS for no more than one second per item tested before my FPS returned to the pre-hover value. This was with a full (zero-value stats included) Stat Summary.

    As it stands, I'm not able to reproduce this problem. If you have anything further to share, please post it.


To post a comment, please login or register a new account.