On Aug 12, 2008, at 2:15 PM, Noah Hart wrote: > I understand that memory, disk and processor all play a part in the > speed test results. > > However, I am looking for some benchmarks that would give ballpark > figures for the results. > > Are people willing to post their speed test results, along with OS, > CPU > speed and Memory characteristics?" > > Such as > > SQLite 3.6.1 running under "Vista 32 bit, 2.4GHz, 3G ram, 5000rpmDisk" > > speed1-insert1... 326279489 uS 10.8195909732183 row/s
Right off the bat, I notice that I'm getting 101871.8 rows/s under Linux. Almost 10,000 times faster. The difference here is that the "testfixture" program that runs this test is compiled with -DSQLITE_NO_SYNC=1 which disables the fsync() system call on unix, but does not disable the equivalent FlushBuffers() system call on win32. The -DSQLITE_NO_SYNC=1 compile- time option is the equivalent of hard-coding "PRAGMA synchronous=OFF". That's great for testing, but might not be such a good idea for production programs that need to survive power failures. If you want to do a meaningful speed comparison, then speed1.test should probably be modified to set PRAGMA synchronous=OFF. D. Richard Hipp [EMAIL PROTECTED] _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
