| 1. |
WeightedCorsika |
Old MC standard |
New MC standard (test) |
New MC optimized set1a |
New MC optimized set2a |
New MC optimized set3a |
New MC optimized set4a |
New MC optimized set5a |
| 2a. |
Energy_Primary Cut |
600 GeV |
600 GeV |
3 TeV |
20 TeV |
40 TeV |
40 TeV |
80 TeV |
| 2b. |
Energy_Muons Cut (ecuts2) |
273 GeV |
273 GeV |
1.2 TeV |
3.0 TeV |
3.0 TeV |
5.0 TeV |
5.0 TeV |
| 2c. |
Energy_electrons Cut (ecuts3) |
0.003 GeV |
0.003 GeV |
500 GeV |
800 GeV |
800 GeV |
800 GeV |
800 GeV |
| 2d. |
Log10(LeadCscdEnergy Cut) |
- |
- |
2.0 |
2.0 |
2.0 |
2.0 |
2.0 |
| 3. |
DatasetId |
1625 ( 1541 ) |
2301 |
2471 |
2479 2487 ( 2480 2600 ) |
2493 |
2519 |
2512 |
| 4. |
Number of files |
100 000 files |
90 files |
100 files |
200 files (934, 9841 files) |
100 files |
991 files |
200 files |
| 4.a |
Number of gen corsika showers per file |
400 000 |
400 000 |
400 000 |
400 000 |
400 000 |
400 000 |
400 000 |
| 4.b |
dslope |
-1 |
-1 |
-1 |
-1 |
-1 |
-1 |
-1 |
| 5. |
NumberOfEvents per file (IC22 triggered) |
2800 |
2900 |
4000 (xxx) |
7400 |
10900 |
6500 |
3000 |
| 6. |
NumberOfEvents per file (Level2) |
900 |
935 |
1600 (xxx) |
3000 |
4500 |
2600 |
1100 |
| 7. |
NumberOfEvents per file (Cscd filter only at level2) |
xx |
216 |
450 |
912 (44222/50=884) |
1370 |
|
416 |
| 8. |
NumberOfEvents per file (level2 CscdFilter + log(RecoEn) gt 4) |
xx |
21 |
49 |
79 (3959/50=79.18) |
126 |
|
50 |
| 9. |
NumberOfEvents per file (level4) |
55000/100000=0.55 |
48/90 = 0.53 (8) |
101/100 = 1.0 |
578/200 = 2.89 (2617/934=2.80) |
412/100 = 4.12
| 2084/872 = 2.4 |
|
| 10. |
NumberOfEvents per file (level6) |
3240/100000=0.03 |
4/90 = 0.04 |
9/100 = 0.09 |
88/200 = 0.44 (415/934=0.44, 4181/9841=0.42) |
53/100 = 0.53 |
350/872 = 0.4 |
14/150= 0.09 |
| 11. |
NumberOfEvents per file (level6 log(RecoEn) gt 3.0) |
1153/100000=0.01 |
xx |
4/100 = 0.04 |
43/200 = 0.21 (164/934=0.18, 1807/9841=0.18) |
26/100 = 0.26 |
167/872 = 0.19 |
8/150 = 0.05 |
| 12. |
NumberOfEvents per file
(level4 + log(RecoEn) gt 2.0 ) |
xxx |
39/90 = 0.43 (7) |
93/100 =0.93 |
549 /200 = 2.75 (2461/934=2.63) |
391 /100 = 3.91 |
|
|
| 13. |
NumberOfEvents per file
(level4 + log(RecoEn) gt 2.6 ) |
xxx |
11/90 = 0.12 (4) |
38/100=0.38 |
279 /200 = 1.40 (1247/934=1.34) |
194 /100 = 1.94 |
|
|
| PDSF (e.g. 1 job only) |
RunTime (wallclock) |
xx |
pc1016 26700 s (7.4h) |
|
pc1016 a) job=2479.26 50573 s (14 h)
b) job=2479.97 27167 s (7.5h) |
|
|
|
| PDSF (e.g. 1 job only) |
RunTime (user CPU) (*) |
xx |
pc1016 16000 s (4.4h) |
|
pc1016 a) job=2479.26 47554 s (13.2 h)
b) job=2479.97 26380 s (7.3h) |
|
|
|
Running time on PDSF machines: 1 node= 2x4=8 cores, cpu speed = 2 GHz , total memory 16 GB
1 core=1 job slot
(*) USer CPU at pdsf : not correct ? to be checked with pdsf experts what is user CPU (does not look like cpu)
Test of standard corsika (a) large statistics older software set 1541 (black histogram) compared with (b) newer software (used for optimized corsika sets) set 2301 (green histogram):
Rate [Hz] vs log (MCPrimary_Energy) and ratio vs log (MCPrimary_Energy)
Conclusion: standard corsika sets (old and new software, settings the same) are consistent
10) After level6 (low statistics) (no energy cut)
10a) Rate versus log(Energy): black=standard corsika
10b) Effective Livetime ( icecube/200902001-v2 ) vs :
MCLeadCascade log(Energy)
, log(RecoEnergy) ,
MCPrimary log(Energy)
10c) Rate versus log(Energy): (same as 10a but only 3 high statistics histograms)
(left) Ratio=Rate(set 2a)/Rate(standard) vs log10(Primary_Energy) and (right) Ratio=Rate(set 4a)/Rate(standard) vs log10(Primary_Energy)
11) After level6 (low statistics) (logRecoEn gt 3.0 )
Note: In the analysis log(RecoEnergy) gt 4 cut is used at the final cut level
11a) Rate versus log(Energy):
11b) Effective Livetime ( icecube/200902001-v2 ) vs :
MCLeadCascade log(Energy)
, log(RecoEnergy) ,
MCPrimary log(Energy)
11c) Rate versus log(Energy): (same as 11a but only 3 high statistics histograms)
(left) Ratio=Rate(set 2a)/Rate(standard) vs log10(Primary_Energy) and (right) Ratio=Rate(set 4a)/Rate(standard) vs log10(Primary_Energy)
Run time (rough!) estimate:
Optimized corsika test samples have limited statistics, but we can see that for optimized and standard corsika rates are ~ consistent and similar energy ranges are
covered for standard and optimized corsika; MC statistics enhancement factor is ~ or more than 20
To get the same amount of statistics as for standard MC (100 000 jobs), we would need to run 4400 jobs (400 000 corsika showers per job)
At pdsf it would take a month or longer. We need more than what we have in stndard IC22 MC.
12) After level7 (no energy cut)
Note: In the analysis log(RecoEnergy) gt 4 cut is used at the final cut level
12a) Rate versus log(Energy):
12b) Effective Livetime ( icecube/200902001-v2 ) vs :
MCLeadCascade log(Energy)
, log(RecoEnergy) ,
MCPrimary log(Energy)