On 29/09/16 14:37 +0100, Andre Vieira (lists) wrote:
Hi Jonathan,
On 27/09/16 16:11, Jonathan Wakely wrote:
The test might not be very good, but tests some small integer values
and some other values where accuracy is lost for one or other of the
alternative implementations mentioned above. If this FAILs for some
32-bit targets we might need to adjust the tolerances or the
dg-options.
On arm-none-eabi I'm seeing a failure for the long double type and inputs:
{ 1e-2l, 1e-4l, 1e-4l, 0.010000999950004999375l }
The abs(frac) is higher than the toler: 1.73455e-16 vs 1e-16. Is that a
reasonable difference? Should we raise toler3 to 1e-15?
The last line is also too high:
{ 2147483647.l, 2147483647.l, 2147483647.l, 3719550785.027307813987l }
Yields a frac of: 1.28198e-16
Those are the only ones that pass the 1e-16 threshold.
Yes, it makes sense to raise it, although Ed's reworking the code and
tests anyway.
