Front page | perl.vmsperl |
Postings from June 2002
At 10:55 PM -0500 6/7/02, Craig A. Berry wrote:
>So the following gets us to do the fake underflow as well as fake
>overflow in the absence of IEEE math:
>
>--- numeric.c;-1 Fri Jun 7 16:40:25 2002
>+++ numeric.c Fri Jun 7 22:18:01 2002
>@@ -744,9 +744,8 @@
> * [1] Trying to establish a condition handler to trap floating point
> * exceptions is not a good idea. */
> #if defined(VMS) && !defined(__IEEE_FP) && defined(NV_MAX_10_EXP)
>- if (!negative &&
>- (log10(value) + exponent) >= (NV_MAX_10_EXP))
>- return NV_MAX;
>+ if ((log10(value) + exponent) >= (NV_MAX_10_EXP))
>+ return negative ? 0.0 : NV_MAX;
> #endif
>
> /* In UNICOS and in certain Cray models (such as T90) there is no
>[end of patch]
Urk. I found a case where this isn't right (or actually
lib/Math/Trig.t found a case where it isn't right):
$ perl -e "my $x = 100.00000000000000001; print $x;"
100
so far so good, but now add another zero after the decimal:
$ perl -e "my $x = 100.000000000000000001; print $x;"
0
In the second example we are definitely hitting the code that returns
zero if the exponent is too small. This only happens with a D_FLOAT
configuration, where NV_MAX_10_EXP will be 38. The other values are:
NUMERIC\S_mulexp10\value: 0.100000000000000E+21
NUMERIC\S_mulexp10\exponent: 18
NUMERIC\S_mulexp10\negative: 1
This is all a bit silly since there aren't 21 digits of precision in
a double, but there is also apparently nothing to keep it from
attempting to represent more digits than it actually can and it
shouldn't print 0 when the right answer is 100. It's past my bedtime
and I don't understand Perl_my_atof well enough to do anything about
this just yet.
--
________________________________________
Craig A. Berry
mailto:craigberry@mac.com
"... getting out of a sonnet is much more
difficult than getting in."
Brad Leithauser
Thread Previous
|
Thread Next