Freescale Semiconductor MCF54455 Reference Manual page 129

Enhanced Multiply-Accumulate Unit (EMAC)
Field
31–12 Reserved, must be cleared.
11–8 Product/accumulation overflow flags. Contains four flags, one per accumulator, that indicate if past MAC or
PAVn MSAC instructions generated an overflow during product calculation or the 48-bit accumulation. When a
MAC or MSAC instruction is executed, the PAVn flag associated with the destination accumulator forms the
general overflow flag, MACSR[V]. Once set, each flag remains set until V is cleared by a move.l, MACSR
instruction or the accumulator is loaded directly.
Bit 11: Accumulator 3
...
Bit 8: Accumulator 0
7 Overflow saturation mode. Enables or disables saturation mode on overflow. If set, the accumulator is set
OMC to the appropriate constant (see S/U field description) on any operation that overflows the accumulator.
After saturation, the accumulator remains unaffected by any other MAC or MSAC instructions until the
overflow bit is cleared or the accumulator is directly loaded.
6 Signed/unsigned operations.
S/U
In integer mode:
S/U determines whether operations performed are signed or unsigned. It also determines the accumulator
value during saturation, if enabled.
0 Signed numbers. On overflow, if OMC is enabled, an accumulator saturates to the most positive
(0x7FFF_FFFF) or the most negative (0x8000_0000) number, depending on the instruction and the
product value that overflowed.
1 Unsigned numbers. On overflow, if OMC is enabled, an accumulator saturates to the smallest value
(0x0000_0000) or the largest value (0xFFFF_FFFF), depending on the instruction.
In fractional mode:
S/U controls rounding while storing an accumulator to a general-purpose register.
0 Move accumulator without rounding to a 16-bit value. Accumulator is moved to a general-purpose
register as a 32-bit value.
1 The accumulator is rounded to a 16-bit value using the round-to-nearest (even) method when moved to
a general-purpose register. See
lower word of the destination register. The upper word is zero-filled. This rounding procedure does not
affect the accumulator value.
5 Fractional/integer mode. Determines whether input operands are treated as fractions or integers.
F/I 0 Integers can be represented in signed or unsigned notation, depending on the value of S/U.
1 Fractions are represented in signed, fixed-point, two's complement notation. Values range from -1 to
-15
1 - 2
Representation."
Round/truncate mode. Controls rounding procedure for move.l ACCx,Rx, or MSAC.L instructions when
4
R/T in fractional mode.
0 Truncate. The product's lsbs are dropped before it is combined with the accumulator. Additionally, when
a store accumulator instruction is executed (move.l ACCx,Rx), the 8 lsbs of the 48-bit accumulator
logic are truncated.
1 Round-to-nearest (even). The 64-bit product of two 32-bit, fractional operands is rounded to the nearest
40-bit value. If the low-order 24 bits equal 0x80_0000, the upper 40 bits are rounded to the nearest even
(lsb = 0) value. See
executed (move.l ACCx,Rx), the lsbs of the 48-bit accumulator logic round the resulting 16- or 32-bit
value. If MACSR[S/U] is cleared and MACSR[R/T] is set, the low-order 8 bits are used to round the
resulting 32-bit fraction. If MACSR[S/U] is set, the low-order 24 bits are used to round the resulting 16-bit
fraction.
5-5
Table 5-2. MACSR Field Descriptions
Section 5.3.1.1,
for 16-bit fractions and -1 to 1 - 2
Section 5.3.1.1, "Rounding". Additionally, when a store accumulator instruction is
Description
"Rounding". The resulting 16-bit value is stored in the
-31
for 32-bit fractions. See Section 5.3.4, "Data
Freescale Semiconductor