Filter
Top Products
CM71-00105-1E FUJITSU MICROELECTRONICS LIMITED 457
FR81 Family
APPENDIX
APPENDIX C Supplemental Explanation about FPU Exception Processing
3. Overflow exception (Overflow)
This exception occurs when, during calculation, the exponent exceeds the maximum that can be
expressed in the specified format. If this exception occurs, the following operations are carried out.
[FCR:EEF:O=1]
Writing to the floating point register is prohibited.
The FCR:CEF:O flag is set to generate this exception.
[FCR:EEF:O=0]
As shown in the following Table C.2-1, the value is written to the floating point register based on the
rounding mode.
The FCR:ECF:O flag is set.
4. Underflow exception (Underflow)
This exception occurs when the rounding result is incorrect and the absolute value is less than the
minimum normalized number in the specified format. In concrete terms, this exception occurs when the
exponent is set to 0 while normalizing a significand or when the exponent is set to a negative value
during multiplication or division. This exception also occurs when a specific value is set to the minimum
normalized number after round processing while it is an unnormalized number before round processing
(pre-rounding rule). This exception does not occur when the result of non-round processing is correct and
set to zero. If this exception occurs, the following operations are carried out.
[FCR:EEF:U=1]
Writing to the floating point register is prohibited.
The FCR:CEF:U flag is set to generate this exception.
[FCR:EEF:U=0]
Zero is always stored in the floating point register. (Zero flush)
The FCR:ECF:U flag is set.
5. Inexact exception (Inexact)
This exception occurs when the rounding result is incorrect (including a case where an Underflow
exception is detected at FCR:EEF:U = 0 when a unnormalized number is flushed to zero) or when
overflow has occurred because an Overflow exception is invalid (including a case where overflow has
occurred as a result of round processing). If this exception occurs, the following operations are carried
out.
[FCR:EEF:X=1]
Writing to the floating point register is prohibited.
The FCR:CEF:X flag is set to generate this exception.
Table C.2-1 Output Result in Rounding Mode and at Overflow
Rounding mode
(FCR:RM)
Output result
Positive overflow Negative overflow
00
B
(Latest value) +∞ -∞
01
B
(Zero) +MAX -MAX
10
B
(+∞)+∞ -MAX
11
B
(-∞)+MAX-∞