Ticket #9254 (closed: invalid)
Optimize log output in Fit.cpp and CostFuncLeastSquares.cpp
Reported by: | Wenduo Zhou | Owned by: | Wenduo Zhou |
---|---|---|---|
Priority: | critical | Milestone: | Release 3.2 |
Component: | Framework | Keywords: | |
Cc: | taylorrj@… | Blocked By: | |
Blocking: | Tester: | Peter Peterson |
Description
Some recent profiling on GetDetOffsetsMultiPeak shows that the logging in Fit.cpp and CostFuncLeastSquares.cpp slow down the execution significantly.
Optimize the logging in these 2 classes.
Change History
comment:2 Changed 7 years ago by Russell Taylor
- Status changed from new to assigned
It would be good to capture (for future reference) the details of exactly why these changes are needed. That can be either in commit messages (preferred) or the ticket.
comment:4 Changed 6 years ago by Wenduo Zhou
- Status changed from assigned to inprogress
Replaced stringstream by boost::lexical_cast. Refs #9254.
Changeset: 17aae69975756a4c707d5e0b4f46982044f8ad66
comment:5 Changed 6 years ago by Wenduo Zhou
- Status changed from inprogress to verify
- Resolution set to invalid
The ticket is closed as 'invalid'. The reason is that with more speed test, it is found that the current implementation is almost as fast as deleting all the log output for Jacobian in CostFuncLeastSquare and fitting iteration status. Replacing the streaming to log object by either stringstream or boost::lexical_cast slows down the algorithm.
Optimized logging. Refs #9254.
The profiling shows that the << operator of Mantid log object is single threaded.
Changeset: f3524bd7936d40d7888309e4bcd110ab064eacbf