11) [Windows Server Build] :: Download Latest and Stable Server Build (Overwrites existing .exe's, includes .dll's)
12) [Windows Server .dll's] :: Download Pre-Requisite Server .dll's
- NPC:NPCToNPCAggroTimerMin 500 (Now 6000) 6 seconds
- NPC:NPCToNPCAggroTimerMax 2000 (Now 60000) 60 seconds
- Database version 9089 will take care of this update automatically only if you used the default values
- The CPU cost of NPC's checking the entire entity list to cast beneficial spells (Heals/Buffs) becomes extremely high when
higher NPC count zones exist (Based off of process profiling)
- Distance checks for every single NPC to every single other NPC who are casting beneficial spells occur every .5 - 2 seconds unless
npc_spells dictates other values, which most of the time it does not
- Zones that once fluctuated from 1-8% CPU with no activity (Idle but players present) now idle at .5% based on my testings due
to this change in conjunction with the past few performance commits, these are zones that have 600-800 NPC's in them
- These values normally are overidden by the spells table (npc_spells), fields (idle_no_sp_recast_min, idle_no_sp_recast_max)
- Added Rate limit the rate in which signals are processed for NPC's (.5 seconds instead of .01 seconds)
Added Perl Export Settings which should heavily reduce the Perl footprint
- Normally when any sub EVENT_ gets triggered, all kinds of variables have to get exported every single time an event is triggered and
this can make Perl very slow when events are triggered constantly
- The two most taxing variable exports are the item variables ($itemcount{} $hasitem{} $oncursor{}) and qglobals ($qglobals{})
- qglobals can pose to be an issue quickly when global qglobals build up, it is highly recommend to use the GetGlobal() and SetGlobal()
methods instead as they don't reference the hashmap $qglobals{} that is rebuilt every single time a sub event is triggered
- A stress test conducted with 10,000 samples shows an excess of time taken to export variables: http://i.imgur.com/NEpW1tS.png
- After the Perl Export Settings table is implemented, and all exports are shut off you see the following test result:
http://i.imgur.com/Du5hth9.png
- The difference of eliminating uneeded exports brings the overhead and footprint of 10,000 triggers from 54 seconds to 2 seconds
- In a 10,000 sample test (10,000 sub event triggers), exporting item variables adds 12 seconds alone, when item variables are only needed in
EVENT_ITEM and EVENT_SAY a majority of the time if at all
- In a 10,000 sample test (10,000 sub event triggers), exporting qglobals with approximately 1,000 global qglobals in the database creates
about 11-20 seconds of delay on its own (Depending on hardware of course)
- I've written a parser that has determined which of these exports are needed in which sub routines and have turned off all of the unneeded
exports in sub routines that do not need them and used it to create the default table that will be installed in the database.
- The export table is called 'perl_event_export_settings' and it resembles the following structure and contains all current 81 EVENTS
- If an entry doesn't exist in this table and a new subroutine is added to the source, all exports will be on by default for that routine
+----------+-----------------------------------------+-----------------+------------+-------------+-------------+--------------+
| event_id | event_description | export_qglobals | export_mob | export_zone | export_item | export_event |
+----------+-----------------------------------------+-----------------+------------+-------------+-------------+--------------+
| 0 | EVENT_SAY | 1 | 1 | 1 | 1 | 1 |
| 1 | EVENT_ITEM | 1 | 1 | 1 | 0 | 1 |
| 2 | EVENT_DEATH | 1 | 1 | 1 | 0 | 1 |
| 3 | EVENT_SPAWN | 1 | 1 | 1 | 0 | 1 |
| 4 | EVENT_ATTACK | 0 | 1 | 1 | 0 | 1 |
| 5 | EVENT_COMBAT | 1 | 1 | 1 | 0 | 1 |
+----------+-----------------------------------------+-----------------+------------+-------------+-------------+--------------+
- If a change is made to this table while the server is live and running, you can hot reload all zone process settings via:
#reloadperlexportsettings
- For those who wonder what "exports" are, they are reference to variables that are made available at runtime of the sub event, such as:
(export_qglobals) (Heavy) : $qglobals https://github.com/EQEmu/Server/blob/master/zone/embparser.cpp#L916
(export_item) (Heavy) : $itemcount{} $hasitem{} $oncursor{} https://github.com/EQEmu/Server/blob/master/zone/embparser.cpp#L1103
(export_zone) : $zoneid, $instanceid, $zoneln etc. https://github.com/EQEmu/Server/blob/master/zone/embparser.cpp#L1083
(export_mob) : $x, $y, $z, $h, $hpratio etc. https://github.com/EQEmu/Server/blob/master/zone/embparser.cpp#L1032
(export_event) : (event specific) IE: EVENT_SAY ($text) https://github.com/EQEmu/Server/blob/master/zone/embparser.cpp#L1141
Add new ruletypes 'Character:UseOldRaceExpPenalties' and 'Character:UseOldClassExpPenalties'
When set to true, adds in bonuses/penalties that existed in Classic
Class-based penalties were disabled during Velious, while race-based penalties were not removed until TSS
This method uses the original SOE method of increase exp needed to level, rather than decreasing exp gained
Updated mechanics to be consistent with live regarding how invisible breaks when the client is the target of a spell.
Invisible will drop whenever a client is hit with a detrimental spell, regardless of if resisted, if it does damage or AOE.
Hide skill now also follows the same rules as above.
Implemented support for Rogue AA - Nerves of Steel which gives a chance for hide NOT to break
when client is hit with an AOE spell.
Spells with this flag will not generate any aggro
Note: DOT portions of spells wilth this flag STILL generate aggro.
Example Harminous Arrow (Ranger AA) 16127
219 not_extendable - > not_focusable - No focus are applied to these spells
217 maxtargets -> no_heal_damage_item_mod - Not applied to these spells.
232 -> no_remove -> Can not click off these spells even if beneficial
209 powerful_flag -> no_resist -> Unresistable spell
Determines the maximum chance this spell has to critical hit.
Ie. If set to 15, the spell will never critical more then 15% of the time
regardless of your characters innate chance to critcal from AAs.
