Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 Speed up ALICE simulation with TFluka Ernesto Lopez Torres.

Similar presentations


Presentation on theme: "1 Speed up ALICE simulation with TFluka Ernesto Lopez Torres."— Presentation transcript:

1 1 Speed up ALICE simulation with TFluka Ernesto Lopez Torres

2 2 Problem HIJING Central event takes ~ 60 h with TFluka ~ 4 h with Geant3

3 3 Why TFluka takes so long time? First ideas: Energy transport cuts on region basics is not allow for hadrons, so we have 1MeV cut for all volumes There is not cut on number of track steps (for GEANT we have set max. 30 000 steps) More detail physics?

4 4 Where is Fluka spending time? Some primary particle create a lot secondaries (10 5 - 10 6 ) 1.Small bug on HMPID: by error someone comment the initialization of the array with the quantum efficiency properties. QEff = 10 31 instead of = 1 T0 HMPID Fix on the CVS

5 5 Where is Fluka spending time? # Primaries with (transp. time > 30s): 657 (1.1%) with a total time: 151067 s = 41.9632 h (63.8%) Transp. time(seg.)Volume name 134101.00 (37.3h) ALIC 6970.29 ZNST 1600.35 ZNSL 1351.01 ZPST 1338.15QB28 766.762QF29 650.828ZPSL 594.577ZDC 401.73QBPM 376.511QS29 192.871ZNG4 190.431ZNG3 177.951ZNG1 177.051ZNG2 168.76HUWA TFluka HIJING Pb-Pb 5.5TeV central event 56140 particles (Laptop Centrino 2GHz 1Gb RAM) Total transport time: 236389s (65.6 h)

6 6 Where is Fluka spending time? Small hadronic cut (1MeV) on ZDC calorimeter (PIPE, etc) allow the develop of detail hadronic showers. Those hadrons escape the volume and develop the EM part also in ALIC from AliZDCv2::CreateMaterials() // Thresholds for showering in the ZDCs i = 1; //tantalum gMC->Gstpar(idtmed[i], "CUTGAM",.001); gMC->Gstpar(idtmed[i], "CUTELE",.001); gMC->Gstpar(idtmed[i], "CUTNEU",.01); gMC->Gstpar(idtmed[i], "CUTHAD",.01); …… i = 5; //lead gMC->Gstpar(idtmed[i], "CUTGAM",.001); gMC->Gstpar(idtmed[i], "CUTELE",.001); gMC->Gstpar(idtmed[i], "CUTNEU",.01); gMC->Gstpar(idtmed[i], "CUTHAD",.01); …. // Avoid too detailed showering along the beam line i = 7; //iron with energy loss (ZIRON) gMC->Gstpar(idtmed[i], "CUTGAM",.1); gMC->Gstpar(idtmed[i], "CUTELE",.1); gMC->Gstpar(idtmed[i], "CUTNEU", 1.); gMC->Gstpar(idtmed[i], "CUTHAD", 1.);

7 7 Where is Fluka spending time? ZDC (ZEM), PIPE, Compensator DIPO =>ALIC AliMC::Stepping() timer.Stop() TH3F->Fill(x,y,z,timer.CpuTime());... timer.Start() } T0 ZEM

8 8 Where is Fluka spending time? (ZDC (ZEM), PIPE, Compensator DIPO) ->ALIC 1100 cm ALIC pi- P=92Gev #secondaries=17626 transport time=134s (2.3 min)

9 9 Where is Fluka spending time? ZDC and PIPE => ALIC TH3F->Fill(x,y,z,timer.CpuTime()); ZDC ZNST ZNSL ZPST

10 10 Where is Fluka spending time? ZDC and PIPE -> ALIC

11 11 To reduce the transport time in ALIC we introduced new volumes with high Energy cuts for gammas an electrons (0.1GeV) ALIC ALIZ Shielding around dipole compensator DCAR PIAR

12 12 We change from to

13 13 Final Results CVSWith new volumes TFluka65 h9 h Geant34 h2 h # Primaries with (transp. time > 30s): 33 (0.05%) with a total time: 8497 s = 2.3 h (25.5%) Transp. time(seg.) Volume name 3841.32ZNST 1209.11ZPST 924.061ZNSL 818.353ALIZ 645.909ZPSL 297.304ZDC 104.881ZNG4 103.681ZNG3 100.791ZNG2 100.521ZNG1 53.16ZNF3 51.78ZNF4 50.12ZNF2 49.53ZNF1 3841.32ZNST

14 14 TODO Check how the changes could affect the ZDC performance.

15 15 END

16 16 Where is Fluka spending time? ALIC ZDC

17 17 Final FLUKA GEANT


Download ppt "1 Speed up ALICE simulation with TFluka Ernesto Lopez Torres."

Similar presentations


Ads by Google