From: tcl-commits <tcl...@tc...> - 2013-03-04 15:40:12
|
Commit by dgp (branch: core-8-4-branch): New scheme for keeping the per-process tcl_precision value in sync without the need for mutex locks on every read. Uses adapted ProcessGlobalValue machinery backported from Tcl 8.5 where it's been working without reported problems. Thanks to Phil Brooks for reporting on tests which highlight the thread performance problems raised by the old scheme, and to Clif Flynt for further testing pointing the finger at tcl_precision locks as the main culprit. http://core.tcl.tk/tcl/info/02ef6b18c8ba8bec37414a1f0286f2264164b178 Changeset: ChangeLog generic/tclUtil.c |