The use of the Win32 /UNDERFLOW option ensures that the first occurrence of underflow in an arithmetical computation is treated as a failure and is not ignored as would otherwise be the case. A large number of occurrences of underflow during execution can result in long execution times because of the way in which the underflow condition is treated. If an underflow is trapped, the message
ERROR: Floating point arithmetic underflow
is output and the interactive debugger is entered. If underflows occur during program execution and the /UNDERFLOW option is not used, a message is output at the end of the run specifying the number of underflows that have occurred.
See also Underflows.