I said:
> I had success with this test:
BTW, some digging in the assembly code shows that the actual problem is
this: instead of emitting "x / 3600.0", with -ffast-math the compiler
emits the equivalent of "x * (double) (1.0 / 3600.0)". It's the
last-bit inaccuracy of the latter constant that's killing us.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 2: you can get off all lists at once with the unregister command
(send "unregister YourEmailAddressHere" to [EMAIL PROTECTED])
