CL-PS7500FE
System-on-a-Chip for Internet Appliance
q Division of 0/0 or infinity/infinity.
q x REM y where x is infinity or y is 0.
q Square root of any number less than zero (but SQT(−0) is −0).
q Conversion to integer when overflow, infinity or NaN make it impossible. If overflow makes a conversion to
integer impossible, the largest positive or negative integer is produced (depending on the sign of the operand)
and Invalid Operation is signalled.
q CMFE, CNFE when at least one operand is a NaN.
DZ – Division by Zero
The division-by-zero exception occurs if the divisor is zero and the dividend a finite, non-zero number. A
correctly-signed infinity is returned if the trap is disabled.
OF – Overflow
The OFC flag is set whenever the destination format's largest number is exceeded in magnitude by what
would have been the rounded result if the exponent range were unbounded. The untrapped result
returned is either:
q the correctly signed infinity
q the format's largest finite number
depending on the rounding mode.
UF - Underflow
Two correlated events contribute to underflow:
1) Tininess
The creation of a tiny non-zero result smaller in magnitude than the format's smallest normalized number.
2) Loss of accuracy
A loss of accuracy due to denormalization that may be greater than would be caused by rounding alone.
If the underflow trap enable bit is set, the underflow exception occurs when tininess is detected, regard-
less of loss of accuracy. If the trap is disabled, then tininess and loss of accuracy must both be detected
for the underflow flag to be set (in which case inexact will also be signalled).
IX – Inexact
The inexact exception occurs if:
q the rounded result of an operation is not exact (different from the value computable with infinite precision)
q overflow has occurred while the OFE trap was disabled
q underflow has occurred while the UFE trap was disabled.
OFE or UFE traps take precedence over IXE.
June 1997
ADVANCE DATA BOOK v2.0
FLOATING-POINT COPROCESSOR PROGRAMMER’S
MODEL
167