Octave documentation on plotting question

I am referring to octave documentation for plotting. floating point numbers are represented with 32 bits (single precision ) or 64 bits (double precision).

What is the meaning of the following statements?

"Similarly, single precision variables can accurately represent only 6-9 base10 digits. If your data contains very fine differences (approximately 1e-8) these cannot be resolved with the OpenGL-based graphics toolkits and the gnuplot toolkit is required. "

See, e.g.: Single-precision floating-point format - Wikipedia

The fraction of IEEE single precision floating point numbers is 23 bits wide. That means the smallest (relative) step representable in that type is in the order of 1/2^23 ≈ 10^-7. I.e., approximately 7 decimal digits. (That might be different for subnormal numbers.)

1 Like