Hi Sergei, The problem you have reported last week is now understood. I modified TTree::CloneTree to set the branch compression level to the file compression level. This forced all the branches in the Tree to be compressed. As a consequence, stress wrote/read less data and it consumed more time. In my new version, I leave the branch comprression level to the value in the original Tree. I assumed you were running on a Pentium II 400 Mhz (right!). The new results are: stress : Total I/O = 418.9 Mbytes, I = 289.6, O = 129.3 stress : Compr I/O = 388.3 Mbytes, I = 267.3, O = 121.1 stress : Real Time = 106.94 seconds Cpu Time = 95.91 seconds ****************************************************************** * ROOTMARKS = 144.4 * Root 2.23/08 991115/1650 ****************************************************************** Rene Brun Sergei Avvakumov wrote: > > Hi, > > Execution times of the 'stress' test for the two versions of ROOT > (from root_v2.23.Linux.2.2.3.egcs111.tar.gz) differ by a lot on my PC : > > 2.23/06: > ****************************************************************** > * Linux viola 2.2.13 #6 Mon Nov 8 14:53:28 EST 1999 i686 unkn > ****************************************************************** > stress : Total I/O = 425.9 Mbytes, I = 296.6, O = 129.3 > stress : Compr I/O = 407.7 Mbytes, I = 286.6, O = 121.1 > stress : Real Time = 112.40 seconds Cpu Time = 98.92 seconds > ****************************************************************** > * ROOTMARKS = 138.6 * Root 2.23/06 991021/850 > ****************************************************************** > > 2.23/08: > ****************************************************************** > * Linux viola 2.2.13 #6 Mon Nov 8 14:53:28 EST 1999 i686 unkn > ****************************************************************** > stress : Total I/O = 397.5 Mbytes, I = 277.8, O = 119.7 > stress : Compr I/O = 350.5 Mbytes, I = 248.2, O = 102.3 > stress : Real Time = 151.18 seconds Cpu Time = 144.61 seconds > ****************************************************************** > * ROOTMARKS = 99.0 * Root 2.23/08 991102/1137 > ****************************************************************** > > Is there any explanation for this? According to release notes, some > performance enhancing changes were introduced in version 2.23/07, > so the effect should have been opposite. > > Thanks, > > Sergei Avvakumov
This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 00:43:42 MET