[Frost] Inconsistent results of using eloss at CMU and at JLab

[Eugene Pasyuk]

Hi Dao,

This is odd behavior. The function 'cylPath' is defined in file targets.F
I discovered one potential problem in there. Some of the variables in it are meant to be static and defined in DATA statement, however SAVE statement is missing in this function. Depending on fortran compiler implementation it may or may not cause a problem. Anyways, I added this SAVE statement and committed updated version to CVS.

All those who use eloss fro g9 and g14 should update the package and recompile it.
I already did it for clasg9 and clasg14 group accounts.

-Eugene

----- Original Message -----
> From: daoh at andrew.cmu.edu
> To: "Eugene Pasyuk" <pasyuk at jlab.org>
> Sent: Tuesday, April 22, 2014 12:43:18 PM
> Subject: Inconsistent results of using eloss at CMU and at JLab
> 
> Dear Eugene,
> 
> I have a subtle inconsistent result from Eloss at JLab and at CMU. In
> particular, I built the eloss package identically (using rootbeer
> makefile) at the lab and at CMU. I then used two million event data
> for
> testing. The first 1.5 million gives me identical result (to 5
> significant
> figures) for the eloss at the lab and at CMU. However, the last 0.5
> million gives me inconsistent result with the result at CMU
> systematically
> smaller than the "correct" one at the lab. My suspicion is that
> somehow
> the program at CMU gets overflowed and some of the return values got
> cut
> off. Could you give me some idea where you think I should take a
> look.
> By the way, I see the function cylPath got called a lot but I could
> not
> find where it got defined. Could you also tell me where to look.
> Thank you in advance.
> Sincerely,
> Dao Ho
> 
>