xxrich@alliant1.lerc.nasa.gov (Rich Rinehart) (07/26/90)
Can anyone offer a good explaination as to why the user time portion from etime can vary greatly (we just saw 2x difference) between runs. I can justify away some of it but not as much as what we are seeing. What are other peoples experiences? -rich -- ----------------------------------------------------------------------------- Rich Rinehart | phone: 216-433-5211 NASA Lewis Research Center | email: xxrich@alliant1.lerc.nasa.gov -----------------------------------------------------------------------------
dereks@aggie.sgi.com (Derek Spears) (07/28/90)
Is your job locked into a CE? If not (as in the case of a program compiled under pcc), the program could float onto IP's, which run much slower. This random "floating about" to different type processors could account for etime differences. Derek Spears dereks@aggie.sgi.com
xxrich@alliant1 (Rich Rinehart) (07/31/90)
In article <11052@odin.corp.sgi.com> dereks@aggie.sgi.com (Derek Spears) writes: > > Is your job locked into a CE? If not (as in the case of a program >compiled under pcc), the program could float onto IP's, which run much >slower. This random "floating about" to different type processors could >account for etime differences. good point, but my question was regarding complex work (see subject). I have also noticed much variance (2x ?) in user time on individual ce's (FORTRAN jobs though, which will not run on the ip's). -- ----------------------------------------------------------------------------- Rich Rinehart | phone: 216-433-5211 NASA Lewis Research Center | email: xxrich@alliant1.lerc.nasa.gov -----------------------------------------------------------------------------