Fons Rademakers wrote: > > I finally had some time to investigate this problem. > On my PPro 200 Linux box I was able to sort 1.000.000 > TObjStrings of the form: > [snip] > stored in a TOrdCollection or TObjArray in 85 seconds. About 130 MB was used for this test and > my machine has 256 MB. > > Anyway the 79 seconds for 1.000.000 strings is not > too bad. So now the question is: what happened in > your case where it took 400 CP seconds to sort 22.000 > strings. Fons and Jim, Here is my view on the problem: The QSORT algorithm is fine if you have enough memory, but it seems to drive the paging disks crazy if you don't. This morning I tried Fons's example on a NT machine with 96 MB but I gave up waiting for it to terminate !! Then I reduced the size to 500.000 strings (the space requirement is then 65-70 MB) and the unoptimised version of the example terminated in 39 seconds ! So with this algorithm the rule seems to be: Make sure you have enough REAL memory to hold the whole TObjArray. Sverre _________________________________________________ Sverre Jarp Email: Sverre.Jarp@Cern.CH IT Department Phone: +41.79.201.0134 CERN Phone at CERN: 16.0134 CH- 1211 Geneva 23 Fax: +41.22.767.7155 _________________________________________________
This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 00:26:19 MET