Date: May 30, 2014 at 9.15 AM US CST (Fermilab local time)
Place: 1-866-740-1260 (ReadyTalk line) (Host : Krzysztof)
Soon Yung Jun
(Summary by Krzysztof)
We discussed Boyana new results related to the introduction of a
helper class G4xyd in G4PhysicsVector. The suggestions were to reorder
the classes and functions according to a metric which would weight
them according to the usage (number of instructions?). Krzysztof
subsequently noted that the branch was done of off v10.0 and not v10.1
so this is what is should be compared to.
We discussed some of the elements of the paper. We decided not to
mention missing initializer lists as most of them do not have a big
computational impact as most of the objects are created quite
infrequently. We also decided not to investigate inlining of
G4Physics2DVector::FindBinLocation at this time. We shall point to
some of the TAU based tables on our redmine project web site as not
all of them will "fit" into the report as there are to many of them.
Andrea suggested by e-mail that some (but probably not all) of the
G4UniversalFluctuation::SampleFluctuations fractional time increase may
be due to inlining of the G4Poisson function.
For the next meeting, Boyana will try to redo the G4xyd tables in:
also trying to point to the line numbers which caused the cache
misses. Krzysztof will try draft the report summary. We also need to
integrate the current TAU sections with the rest of the report