Project

General

Profile

Meeting

Date: February 14, 2014 at 9.15 AM US CST (Fermilab local time)
Place: 1-866-740-1260 (ReadyTalk line) (Host : Krzysztof)

Participants:

Andrea Dotti
Krzysztof Genser
Boyana Norris
Soon Yung Jun

Discussion

(Summary by Krzysztof)

We had mainly discussed recent results related to inlining functions of G4UrbanMscModel

SampleDisplacement()
LatCorrelation()
ComputeTheta0(double, double)
SimpleScattering(double, double)

As opposed to the geant4 v10.r00 where all of them inlined leading to about 2.5% cpu improvement
this time the last one was not inlined by the gcc C++ compiler.

Detailed profiling results from Soon follow:


Legend:
Sample = [Particle].[PhysicsList].[Energy(GeV)].[BField(T)]
r01        = 10.0-ref-01  
r01em   = 10.0-ref-01 + emstand-V10-00-09

-------------------------------------------------------------------------------------------
sample (CPU time/event      r01         r01em     %change
-------------------------------------------------------------------------------------------
higgs.FTFP_BERT.1400.4   351.68     351.99      0.09  
-------------------------------------------------------------------------------------------
e-.FTFP_BERT.1.0              0.0170     0.0172     1.18  
e-.FTFP_BERT.5.0              0.0856     0.0863    0.82  
e-.FTFP_BERT.10.0            0.1697     0.1714     1.00  
e-.FTFP_BERT.50.0            0.8384     0.8462     0.93 
------------------------------------------------------------------------------------------- 
e-.FTFP_BERT.1.4               0.0227      0.0229    0.88  
e-.FTFP_BERT.5.4               0.1139      0.1143     0.35  
e-.FTFP_BERT.10.4             0.2275      0.2284    0.40  
e-.FTFP_BERT.50.4             1.1309      1.1324     0.13  
-------------------------------------------------------------------------------------------
pi-.FTFP_BERT.1.0              0.0297     0.0294    -1.01  
pi-.FTFP_BERT.5.0              0.1301     0.1296    -0.38  
pi-.FTFP_BERT.10.0            0.2506     0.2496    -0.40  
pi-.FTFP_BERT.50.0            1.1084     1.1107      0.21  
-------------------------------------------------------------------------------------------
pi-.FTFP_BERT.1.4              0.0342      0.0339   -0.88  
pi-.FTFP_BERT.5.4              0.1503     0.1496   -0.47  
pi-.FTFP_BERT.10.4            0.2889     0.2886   -0.10  
pi-.FTFP_BERT.50.4            1.3408     1.3306    -0.76  
-------------------------------------------------------------------------------------------
pi-.QGSP_BERT.1.4              0.0335     0.0341    1.79  
pi-.QGSP_BERT.5.4              0.1464     0.1458   -0.41  
pi-.QGSP_BERT.10.4            0.2858     0.2861    0.10  
pi-.QGSP_BERT.50.4            1.2849     1.2881     0.25  
-------------------------------------------------------------------------------------------
pi-.QGSP_BIC.1.4               0.0344     0.0331    -3.78  
pi-.QGSP_BIC.5.4               0.1479     0.1492    0.88  
pi-.QGSP_BIC.10.4             0.2877     0.2884    0.24  
pi-.QGSP_BIC.50.4             1.2884     1.3018    1.04  
-------------------------------------------------------------------------------------------
anti_proton.FTFP_BERT.1.4      0.0788     0.0783   -0.63  
anti_proton.FTFP_BERT.5.4      0.2038     0.2020   -0.88  
anti_proton.FTFP_BERT.10.4    0.3394      0.3396    0.06  
anti_proton.FTFP_BERT.50.4    1.4147      1.4074   -0.52  
-------------------------------------------------------------------------------------------
proton.FTFP_BERT.1.4           0.0259     0.0255   -1.54  
proton.FTFP_BERT.5.4           0.1504     0.1513    0.60  
proton.FTFP_BERT.10.4         0.2944      0.2930   -0.48  
proton.FTFP_BERT.50.4         1.3699     1.3697   -0.01 
-------------------------------------------------------------------------------------------

plots and tables are available at:
https://g4cpt.fnal.gov/perfanalysis/g4p/index_others_10.0.html

Here is the previous result from Soon:

10.0.r00a (inlining the 4 shorter functions of G4UrbanMscModel mentioned above)
compared to 10.0.


#    10.0    10.0.r00a  %change    sample
 352.5200    347.4600   -1.44     higgs.FTFP_BERT.1400.4     
   0.0174      0.0169   -2.87     e-.FTFP_BERT.1.0           
   0.0872      0.0850   -2.52     e-.FTFP_BERT.5.0           
   0.1739      0.1695   -2.53     e-.FTFP_BERT.10.0          
   0.8578      0.8380   -2.31     e-.FTFP_BERT.50.0          
   0.0227      0.0224   -1.32     e-.FTFP_BERT.1.4           
   0.1137      0.1124   -1.14     e-.FTFP_BERT.5.4           
   0.2277      0.2248   -1.27     e-.FTFP_BERT.10.4          
   1.1249      1.1121   -1.14     e-.FTFP_BERT.50.4          
   0.0301      0.0296   -1.66     pi-.FTFP_BERT.1.0          
   0.1321      0.1295   -1.97     pi-.FTFP_BERT.5.0          
   0.2532      0.2481   -2.01     pi-.FTFP_BERT.10.0         
   1.1348      1.1067   -2.48     pi-.FTFP_BERT.50.0         
   0.0334      0.0331   -0.90     pi-.FTFP_BERT.1.4          
   0.1489      0.1471   -1.21     pi-.FTFP_BERT.5.4          
   0.2919      0.2885   -1.16     pi-.FTFP_BERT.10.4         
   1.3396      1.3248   -1.10     pi-.FTFP_BERT.50.4         
   0.0335      0.0331   -1.19     pi-.QGSP_BERT.1.4          
   0.1475      0.1460   -1.02     pi-.QGSP_BERT.5.4          
   0.2862      0.2826   -1.26     pi-.QGSP_BERT.10.4         
   1.2903      1.2732   -1.33     pi-.QGSP_BERT.50.4         
   0.0329      0.0322   -2.13     pi-.QGSP_BIC.1.4           
   0.1487      0.1471   -1.08     pi-.QGSP_BIC.5.4           
   0.2888      0.2850   -1.32     pi-.QGSP_BIC.10.4          
   1.2964      1.2797   -1.29     pi-.QGSP_BIC.50.4          
   0.0792      0.0780   -1.52     anti_proton.FTFP_BERT.1.4  
   0.2032      0.2014   -0.89     anti_proton.FTFP_BERT.5.4  
   0.3423      0.3375   -1.40     anti_proton.FTFP_BERT.10.4 
   1.4130      1.3964   -1.17     anti_proton.FTFP_BERT.50.4 
   0.0255      0.0252   -1.18     proton.FTFP_BERT.1.4       
   0.1526      0.1504   -1.44     proton.FTFP_BERT.5.4       
   0.2955      0.2913   -1.42     proton.FTFP_BERT.10.4      
   1.3764      1.3603   -1.17     proton.FTFP_BERT.50.4

On the review report topic, Krzysztof uploaded its latex skeleton to the repository.