This is a repeat of the study I did last April: Sti comparison between SL16b for HitReUseOn. This time I am testing in DEV as of 2016-11-16, and comparing Sti, StiCA, and Sti+Hit-ReUse in their current states, but still using the same 500 events from Run 13 pp510 run 14079008.

First, some track counts. In this table, I quote a fractional increase in TPC tracks relative to standard Sti: a fractional increase of 0.50 means that there are 50% more tracks, e.g. 100 standard Sti tracks and 150 tracks when re-using TPC hits.

	Sti	Sti+ReUse	fractional increase	StiCA	fractional increase	distributions for Sti (black), Sti+ReUse (red), and StiCA (blue)
total global tracks	2339469	1820840	-0.22	2135956	-0.09
"good" global tracks	1085507	1055555	-0.03	1175753	0.08
"good" primary tracks	77077	76687	-0.01	97162	0.26
globals passing my cuts	118311	155032	0.31	159665	0.35	(not available)

Documentation on what are "my cuts" is available here.

I made three track-by-track comparisons, using the sample of globals passing my cuts from the table above, and I include the April results as a comparison point:

	A=Sti, B=Sti+ReUse	A=Sti, B=StiCA	A=StiCA, B=Sti+ReUse	A=Sti, B=Sti+ReUse (April)
Matches	107696	96549	126680	129040
Unmatched from A	10615	21762	32985	2572
Unmatched from B	47336	63116	28352	89699

We see that there were far fewer unmatched standard Sti tracks in April. Also, while StiCA finds more tracks that pass my cuts than Sti+ReUse, fewer of the StiCA tracks match with standard Sti tracks. StiCA and Sti+ReUse do show a rather large increase in matched tracks between each other, and it is particularly worth noting that the 126680 matched tracks between these two is a larger number than the 118311 tracks found by standard Sti (implying that StiCA and Sti+ReUse are similarly finding lots of tracks that standard Sti has missed).

Here are some detailed plots comparing quantities track-by-track...

1) Sti vs. Sti+ReUse

The first plot shows three distributions: matched tracks (black), unmatched tracks from Sti+ReUse (red), unmatched tracks from Sti (blue); the second plot shows the ratio of red/black, i.e. the fractional increase of tracks due to turning on re-use.

q/pT
η
φ (defined here as azimuthal angle of pointing [or momentum] vector near primary vertex)
zvertex
signed DCA
Nhits

2) Sti vs. StiCA

The first plot shows three distributions: matched tracks (black), unmatched tracks from StiCA (red), unmatched tracks from Sti (blue); the second plot shows the ratio of red/black, i.e. the fractional increase of tracks due to turning on StiCA.

q/pT
η
φ (defined here as azimuthal angle of pointing [or momentum] vector near primary vertex)
zvertex
signed DCA
Nhits

2) StiCA vs. Sti+ReUse

The first plot shows three distributions: matched tracks (black), unmatched tracks from Sti+ReUse (red), unmatched tracks from StiCA (blue); the second plot shows the ratio of red/black, i.e. the fractional increase of tracks due to Sti+ReUse over StiCA.

q/pT
η
φ (defined here as azimuthal angle of pointing [or momentum] vector near primary vertex)
zvertex
signed DCA
Nhits

One final note on reconstruction time (same node): 32.9 CPU-sec/event with Sti+ReUse vs. 22.0 CPU-sec/event for standard Sti, and 22.1 CPU-sec/event for StiCA.


-Gene