-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 09/01/12 08:01, Robert Hawkey wrote: > Is there a large difference in speed between the two?
Why would you expect there to be? As per Amdahl's law you need to work out how much of your program execution is spent in SQLite in the first place. For example if it is 5% then even if you optimised the SQLite code to be infinitely fast, your program would still take 95% of what it did before. Some time is going to be spent in fsync() calls depending on how much writing you do. The latency for those calls is very volatile on Android and would overwhelm any other measurements. You are doing pretty much the same thing in both cases: starting with Java code and values, and calling across into C. It is essentially the same C code running, and the values have to cross the Java C boundary somehow. As always benchmarks are not meaningful, unless done on data and usage patterns that substantially represent your circumstances. Turning this around, lets say I did measure and said that using the NDK was twice as fast for me. How would that actually be useful to you? Roger -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.11 (GNU/Linux) iEYEARECAAYFAk8LWX0ACgkQmOOfHg372QTJ+ACg5dGhW/jP9RbAmeXiSkppBs7v i+0AoNnVuuvBOybBnIDDK2iPFzRnPI1A =wd/F -----END PGP SIGNATURE----- _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
