CODEWITHSUNDEEP
matlabfloating-pointfortranprecision
BillyJean
BillyJean

Reputation: 1587

Precision of Fortran vs that of Matlab

In Fortran, I output the result of tanh(1) and I get the value 0.7615941763:

open(unit=2, file='test_digits.txt', ACTION="write")
write(2, '(1000F14.10)')( real(tanh(1.0)))

However, I now try to do the same in MatLAB and the output is 0.761594155955765. There is a difference at the 8th digit.

What is the reason for this precision-difference? Can I fix it somehow?

Upvotes: 2

Views: 234

Answers (1)

Alexander Vogt
Alexander Vogt

Reputation: 18098

Matlab is using double precision by default, you are using single precision floats! They are limited to 7-8 digits... If you use double precision as well, you will get the same precision: 

program test
  write(*,*) 'single precision:', tanh(1.0)
  write(*,*) 'double precision:', tanh(1.0d0)
end program

Output: 

 single precision:  0.761594176    
 double precision:  0.76159415595576485

Upvotes: 6

Related Questions