On Friday 26 January 2001 18:07, Tom Lane wrote:
> Curiously, this change exposed what I take to be a platform dependency
> in the int8 regress test. It was computing
> int8(float8(4567890123456789::int8)) and expecting to get back exactly
> 4567890123456789. However, that value is 53 bits long and so there is
> no margin for error in a standard IEEE float8 value. I find that at
> least on HP hardware, rint() treats the value as inexact and rounds to
> nearest even:
>
> regression=# select round(4567890123456788::float8) -
> 4567890123456780::float8; ?column?
> ----------
> 8
> (1 row)
>
> regression=# select round(4567890123456789::float8) -
> 4567890123456780::float8; ?column?
> ----------
> 8
> (1 row)
>
> regression=# select round(4567890123456790::float8) -
> 4567890123456780::float8; ?column?
> ----------
> 10
> (1 row)
>
> regression=#
>
> Whether this is a bug in rint or spec-compliant behavior is unclear, but
> I'll bet HP's hardware is not the only platform that behaves this way.
> Since I'm not eager to try to develop a new set of platform-specific
> int8 expected files at this late hour, I just diked out that test
> instead...
Here is what I get on Linux (PIII):
reaster=# select round(4567890123456788::float8) - 4567890123456780::float8;
?column?
----------
8
(1 row)
reaster=# select round(4567890123456789::float8) - 4567890123456780::float8;
?column?
----------
9
(1 row)
reaster=# select round(4567890123456790::float8) - 4567890123456780::float8;
?column?
----------
10
(1 row)
I'm not sure what the problem is either. The PIII has an 80-bit FPU but not
sure that matters. When there is no exponent, maybe only 52 bits are really
in the mantissa. If you try rounding numbers <= 4503599627370495 (2^52 - 1),
maybe you'll get expected results. The hidden bit is 0. Could be that round
or rint (whatever it is) always makes the hidden bit 1 when I think it should
only be 1 when the exponent is nonzero. I'm no float expert! :) Feel free
to correct me.
--
-------- Robert B. Easter reaster@comptechnews.com ---------
-- CompTechNews Message Board http://www.comptechnews.com/ --
-- CompTechServ Tech Services http://www.comptechserv.com/ --
---------- http://www.comptechnews.com/~reaster/ ------------
