lmi
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [lmi] numeric_io_traits problem under Valgrind


From: Vadim Zeitlin
Subject: Re: [lmi] numeric_io_traits problem under Valgrind
Date: Tue, 27 Feb 2018 16:54:29 +0100

On Tue, 27 Feb 2018 02:44:15 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2018-02-27 02:07, Vadim Zeitlin wrote:
GC> [...]
GC> > http://www.network-theory.co.uk/docs/valgrind/valgrind_27.html
GC> > 
GC> >   > As of version 3.0.0, Valgrind has the following limitations in
GC> >   > its implementation of x86/AMD64 floating point relative to
GC> >   > IEEE754. Precision: There is no support for 80 bit arithmetic.
GC> [...snip their bloviation...]
GC> > 
GC> >  So this looks like a direct consequence of limitations in the current
GC> > Valgrind FP support implementation, because log10l() of the number above 
is
GC> > very close to 1e-16 and the loss/extra precision makes it go just
GC> > above/below it.
GC> 
GC> If there's some valgrind-is-being-used macro, we'll just use it
GC> to conditionalize the test.

 Yes, this can indeed be done, and thanks to C++17 __has_include, we can
even do it without changing anything at the build system level. The result
is still a bit ugly, but not as bad as it could have been, please see

        https://github.com/vadz/lmi/pull/72

 Thanks,
VZ


reply via email to

[Prev in Thread] Current Thread [Next in Thread]