[Ecls-list] Another question
Juan Jose Garcia-Ripoll
juanjose.garciaripoll at googlemail.com
Sun Dec 18 20:12:10 UTC 2011
On Sun, Dec 18, 2011 at 8:52 PM, Paul Bowyer <pbowyer at olynet.com> wrote:
> I've just gotten a test copy of the 64-bit version of PCLinuxOS running on
> another computer and as part of checking it out, I have tried to set up
> ECL on it. Everything seems to have gone correctly and ECL installed
> apparently without problems, but when I ran the tests, I experienced more
> failures in the ansi.log file for the 64-bit install than I get with the
> 32-bit install that I have been using.
Some of the ANSI tests are not very "correct" I would say. The tests check
on the accuracy of certain functions, but what is accurate is something to
be discussed. The differences between 32- and 64-bit platforms, or between
operating systems in the same machine, are due to how the C library
computes those mathematical functions. Some of them have enough accuracy
for long-double types, some don't. Some work internally with double-float
precision and have more accuracy than what the test suite expects. Some
signal overflows, some don't. etc, etc.
The other two tests
are also probably due to the fact that the code uses internally double
types for the products, exceeding the precision of the interpreted code. I
also see them in my computer.
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the ecl-devel