There was a dedicated bechmark with a copy of the RooExponential that
didn't implement the BatchMode, just to see that using the BatchMode is
still faster because the remaining PDFs are implemented and because of
the less virtual function calls.
However, the fallback for PDFs that are not implementing
computeBatch() is now such a central piece of the BatchMode that has
proven to work over and over, it's not worth anymore to carry this
SlowRooExponential code around just for this test.
There was a dedicated bechmark with a copy of the RooExponential that didn't implement the BatchMode, just to see that using the BatchMode is still faster because the remaining PDFs are implemented and because of the less virtual function calls.
However, the fallback for PDFs that are not implementing
computeBatch()
is now such a central piece of the BatchMode that has proven to work over and over, it's not worth anymore to carry thisSlowRooExponential
code around just for this test.