Related inks:

• Filter's configurations and bias/acceptance tests used for this timing study

Figure 1: BFC filter processing time for accepted events
Note an extra peaks around 23/33/43 seconds for the QCD sample
(they also present but less pronounced in gamma-jet sample)
I found that these are processing times needed for the first accepted by filter events
(not always the time for the first processed event and depends on filter acceptance rate).
Processing is much longer due to time needed to initiaize additional stuff in bfc makers
and it is ignored in the total cpu time estimate

Figure 1b:
1st row: starsim time per partonic pt bin
2nd row: bfc time for accepted events per partonic pt bin
3rd row: bfc time for rejected events per partonic pt bin

Average starsim/bfc timing (ignoring times in Fig. 1 with more that 18 seconds):

Gamma-jets   bfc:acc     starsim    bfc:rej
pt=25-55       6.49765   20.2376    0.139911
pt=15-25      4.04562    11.8268     0.095903
pt=9-15        4.84475    11.4816     0.112344
pt=6-9         6.19909     13.0528     0.143109
pt=4-6         4.86546     9.8856      0.114983
pt=3-4         5.07415     10.404       0.12363
pt=2-3         4.04254     9.04627    0.0995413

QCD           bfc:acc       starsim       bfc:rej
pt=25-55   6.18626     14.9944     0.13752
pt=15-25   6.20077     12.9668     0.126722
pt=9-15     6.848         13.7706     0.140625
pt=6-9      5.29513     10.3708      0.110363
pt=4-6      6.29547     11.5318      0.127077
pt=3-4      4.45859     10.1538      0.0986285
pt=2-3      5.26187     13.9151      0.114372