Analog Devices ADSP-2106x SHARC User Manual page 530

ALU Floating-Point
B Compute Operations
Fn = (Fx + Fy)/2
Syntax:
Fn = (Fx + Fy)/2
Function:
Adds the floating-point operands in registers Fx and Fy and divides the
result by 2, by decrementing the exponent of the sum before rounding.
The normalized result is placed in register Fn. Rounding is to nearest
(IEEE) or by truncation, to a 32-bit or to a 40-bit boundary, as defined by
the rounding mode and rounding boundary bits in MODE1. Post-rounded
overflow returns ±Infinity (round-to-nearest) or ±NORM.MAX (round-to-
zero). Post-rounded denormal results return ±Zero. A denormal input is
flushed to ±Zero. A NAN input returns an all 1s result.
Status flags:
AZ Is set if the post-rounded result is a denormal
(unbiased exponent < –126) or zero, otherwise cleared
AU Is set if the post-rounded result is a denormal, otherwise cleared
AN Is set if the floating-point result is negative, otherwise cleared
AV Is set if the post-rounded result overflows
(unbiased exponent > +127), otherwise cleared
AC Is cleared
AS Is cleared
AI Is set if either of the input operands is a NAN, or if they are
opposite-signed Infinities, otherwise cleared
B – 28
www.BDTIC.com/ADI