------- Comment #4 from sgk at troutmask dot apl dot washington dot edu 2005-10-29 20:27 ------- Subject: Re: Complex arithmetic on special cases is incorrect.
On Sat, Oct 29, 2005 at 08:09:45PM -0000, joseph at codesourcery dot com wrote: > > > * underflow, sign and efficiency bugs by rewriting I*y as > > * (0.0+I)*(y+0.0*I) and laboriously computing the full complex product. > > Note that the correct form is (0.0+I)*y, since I is (per C99+TC1+TC2) > _Complex_I, of complex type (Annex G imaginary types conflicting with the > normative standard, unless and until anything changes in this respect > following DR#323). But the usual arithmetic conversions as specified in > the standard do not convert both operands to complex, so one can be real > and one complex. > If I read Annex G correctly, the z = I*inf = NaN + I inf is going to really bad things because the NaN is going to propagate if z is used in further computations. Annex G says z is an infinity. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24581
