aboutsummaryrefslogtreecommitdiff
path: root/builtin/profiler/init.lua
Commit message (Collapse)AuthorAge
* Use a settings object for the main settingsShadowNinja2017-05-06
| | | | | | | This unifies the settings APIs. This also unifies the sync and async registration APIs, since the async registration API did not support adding non-functions to the API table.
* Builtin/profiler: Replace game profiler (#4245)Tim2016-07-12
Use the setting "profiler.load" to enable profiling. Other settings can be found in settingtypes.txt. * /profiler print [filter] - report statistics to in-game console * /profiler dump [filter] - report statistics to STDOUT and debug.txt * /profiler save [format [filter]] - saves statistics to a file in your worldpath * txt (default) - same treetable format as used by the dump and print commands * csv - ready for spreadsheet import * json - useful for adhoc D3 visualizations * json_pretty - line wrapped and intended json for humans * lua - serialized lua table of the profile-data, for adhoc scripts * /profiler reset - reset all gathered profile data. This can be helpful to discard of any startup measurements that often spike during loading or to get more useful min-values. [filter] allows limiting the output of the data via substring/pattern matching against the modname. Note: Serialized data structures might be subject to change with changed or added measurements. csv might be the most stable, due to flat structure. Changes to the previous version include: * Updated and extended API monitoring * Correct calculation of average (mean) values (undistorted by idleness) * Reduce instrumentation overhead. * Fix crashes related to missing parameters for the future and occasional DIV/0's. * Prevent issues caused by timetravel (overflow, timejump, NTP corrections) * Prevent modname clashes with internal names. * Measure each instrumentation individually and label based on registration order. * Labeling of ABM's and LBM's for easier classification. Giving several ABM's or LBM's the same label will treat them as one. Missing labels will be autogenerated based on name or registration order. * Configurable instrumentation and reporting. Skip e.g. builtin if you don't need it. * Profile the profiler to measure instrumentation overhead.