fix offsets used in Amp3Body for cached values #264
Conversation
extern packages are not properly compiling
updating branch with latest master changes
|
@FlorianReiss, the OMP build test seems to be failing because of an error in src/goofit/FitManagerMinuit2.cpp. That looks a bit strange because as far as I can see there are no changes that should affect that piece of code. Would you have any idea of what would be causing it? |
no, I can't think of anything that would cause this. The only explanation I have that I did not see this issue when compiling is that I picked up Minuit from my local ROOT installation, while this test uses the external one, which might be a different version |
|
@danielsibemol I merged with master and fixed the conflicts. I don't quite understand why the tests are still failing. Could you please have a look? |
It seems to be having problems getting the proper members from minuit. I'll take a look and try to understand why |
…ooFit into freiss_fixAmp3BodyOffsets Getting latest change from remote
@FlorianReiss , I think think a small adjustment was necessary in the if conditions for minuit2 since our extern minuit2 was not updated yet. Let's see how the tests go. |
|
If I put #269 in, @FlorianReiss are you okay if I do a rebase & force push, or would you rather I do a traditional merge to update this PR with the contents of #269? |
|
With so many unsquashed commits, rebasing looked tricky - did a standard merge instead. |
|
@danielsibemol @henryiii sorry, I didn't realize that this MR was being hold up by a failing test. I modified the test for EventWeightedAddPdf to address this |
try to fix two problems when using cached values in Amp3Body
EventIndexused by thecalculatorsalways started at zero, so that it ran always over the events at the beginning of the event array when calculating the cached values. This is fixed by adding the option to define an offset inSetDataSize. The offset needs to be the number of preceding events when one tries to fit multiple samples.MEMCPYtocResonancesoverwrites the previous values if one has more instances ofAmp3Body. This is fixed using the already existingcacheToUseas offset. Added the functionresetCacheCounterto reset the counter used for the offset. This can be useful for plotting the results.The MR is made on top of other changes. For easier overlook of the changes, see here
EventWeightedAddPdfaddresses #262 #263