I’m looking for the values of epsilon (the smallest step between two numbers), min (the smallest magnitude) and max (the greatest magnitude) for CUDA devices.
I.E the equivalents to FLT_EPSILON (DBL_EPSILON), FLT_MIN (DBL_MIN) and FLT_MAX (DBL_MAX) defined in <float.h> in gcc compilers.
Are there constants in some CUDA include file? Any manual explaining them? Any way to write a kernel to compute them?
Thanks in advance.