...> x=x+a1 write(*,*)x x=int2*a2+a1 write(*,*)x end program test With ifort this compiles fine, and produces: $a.out -61547.0000000000000000000000000000 0.000000000000000000000000000000000E+0000 0.000000000000000000000000000000000E+0000 I get the same results also using the hp=kind(1.0e0). g95 with the latter gives: a.out ...

...*a2 write(*,*)x x=x+a1 write(*,*)x x=int2*a2+a1 write(*,*)x end program test With ifort this compiles fine, and produces: $a.out -61547.0000000000000000000000000000 0.000000000000000000000000000000000E+0000 0.000000000000000000000000000000000E+0000 I get the same results also using the hp=kind(1.0e0). g95 with the latter gives: a.out -61547....

... some print routines are capable of printing the number to an arbitrary number of decimal places. In effect you've engineered a sophisticated example of, say, printing 2 as 2.0000000000000000000000000000... Yet the issue of precision still stands: [32]> (expt 2.0000000000000000d0 -256) 8.636168555094445d-78 [33]> (expt 2.0000000000000001d0 -256) 8.636168555094445d-78 2.0 is a floating ...