Be careful with icc compiler family. It is relatively frequent to get small differences between gcc and icc floating (double/float) computations.A specialist of floating point said that icc build faster programs fro two reasons: 1) Intel knows his processors, 2) they make excessive assumptions on floating point expression (more or less that addition is associative, something like fast-math option in gcc). t.
Le 28/06/2017 à 04:42, François Bissey a écrit : > > > On 28/06/17 10:51, François Bissey wrote: >> with the peculiar being: >> ********************************************************************** >> File "src/sage/rings/real_double.pyx", line 2296, in >> sage.rings.real_double.RealDoubleElement.arccos >> Failed example: >> i.arccos() == q >> Expected: >> True >> Got: >> False >> ********************************************************************** >> If anyone has an idea. > > Definitely numerical noise. > sage: q = RDF.pi()/3 > sage: i = q.cos() > sage: q - i.arccos() > -2.220446049250313e-16 > > Is this supposed to be exact? > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To post to this group, send email to sage-devel@googlegroups.com. Visit this group at https://groups.google.com/group/sage-devel. For more options, visit https://groups.google.com/d/optout.
<<attachment: tdumont.vcf>>
