On Jul 8, 2008, at 1:13 AM, Cory Nelson wrote:
> On Mon, Jul 7, 2008 at 9:06 PM, Shane Harrelson <[EMAIL PROTECTED]>
> wrote:
>> Make sure SQLite isn't being compiled with -ffast_math on the the
>> Debian
>> side. That might cause problems.
>
> -ffast-math would not cause sqlite to bug out like described.
>
> it lets the compiler reorder floating point expressions in a way that
> might trivially change the output while being faster. so (a * b + a *
> c) would be transformed into (a*(b+c)), which could give different
> results due to rounding.
From the GCC manpage:
This option should never be turned on by any -O option
since it can
result in incorrect output for programs which depend on an
exact
implementation of IEEE or ISO rules/specifications for math
functions.
In SQLite, we have observed that NaN detection does not work with -
ffast-math. This problem cascades whenever NULL values and floating
point values are intermixed.
D. Richard Hipp
[EMAIL PROTECTED]
_______________________________________________
sqlite-users mailing list
[email protected]
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
