The 21064 Alpha processor does not support infinity and NaN values in hardware. The use of infinity and NaN values slows these systems based on this processor because these values must be calculated in software.
To get maximum floating-point performance for applications that do not deal with infinity and NaN values, these values cause an exception by default, to provide an opportunity to correct what is normally an error condition. This behavior is the default of the compiler.
For applications that require infinity and NaN values without exceptions, the default compiler must be changed by using the /QAieee switch, and the Windows NT default behavior must be changed by calling _ _controlfp(_EM_INEXACT, _MCW_EM), which will disable all exceptions except for inexact results. Inexact exceptions are not generated unless the /QAieee2 compiler option is used.