On Tue, 16 Sep 2008, Mindaugas Kavaliauskas wrote: Hi Mindaugas,
> I've tested a few memory managers (MM) with tests\memtst.prg. All test is > using BCC55 compiler, singe thread application, FM statistics switched off, > 2.2GHz Celeron, 256MB RAM. > Three MM were tested: BCC55, Win32 Local*(), DL MM (see: > http://gee.cs.oswego.edu/dl/html/malloc.html). > The results are: > BCC WIN DL > single large memory blocks: 1.27 1.80 2.56 > many large memory blocks: 13.11 12.97 12.69 > large memory block reallocation: 2.47 3.30 0.33 Very interesting. BTW this are my results for Linux with GLIBC 2.8 on Phenom 2.1Ghz: single large memory blocks: 0.55 many large memory blocks: 3.07 large memory block reallocation: 0.07 > P.S. BCC with its MM does not crash on memtst.prg! Probably earlier memory allocation in two previous tests changed the internal memory strategy or simply created in memory manager enough number of memory 'wholes' that that were used for intermediate allocations so the realloc was always working on the same memory area only enlarging it (no new memory was mapped after main array block). best regards, Przemek _______________________________________________ Harbour mailing list Harbour@harbour-project.org http://lists.harbour-project.org/mailman/listinfo/harbour
