-
Notifications
You must be signed in to change notification settings - Fork 14
Open
Description
A range of recent optimisations in XAD focused on significantly reducing overheads of using the XAD types when tape recording is not enabled. Purely for using xad::AReal<double> instead of double.
We would like to record and check benchmarks of this overhead compared to double using the full QuantLib test suite as an example and the QuantLib benchmarks. This should have all the individual QuantLib tests timed, with and without XAD types, and run a few times with averaging for more stable results.
Understanding this overhead with multiple compilers and operating systems would be highly beneficial to guide future optimisation work.
Ideally this should be added to the workflows so that it can easily be re-run with new versions.
Metadata
Metadata
Assignees
Labels
No labels