
TM1300 Data Book
Philips Semiconductors
A-26
PRODUCT SPECIFICATION
Dual clipped multiply of signed 16-bit halfwords
SYNTAX
[ IF r
guard ] dspidualmul rsrc1 rsrc2
→ rdest
FUNCTION
if r
guard then {
temp1
← sign_ext16to32(rsrc1<15:0>) × sign_ext16to32(rsrc2<15:0>)
temp2
← sign_ext16to32(rsrc1<31:16>) × sign_ext16to32(rsrc2<31:16>)
if temp1 < 0xffff8000 then temp1
← 0x8000
if temp2 < 0xffff8000 then temp2
← 0x8000
if temp1 > 0x7fff then temp1
← 0x7fff
if temp2 > 0x7fff then temp2
← 0x7fff
r
dest<31:16>
← temp2<15:0>
r
dest<15:0>
← temp1<15:0>
}
ATTRIBUTES
Function unit
dspmul
Operation code
95
Number of operands
2
Modier
No
Modier range
—
Latency
3
Issue slots
2, 3
DESCRIPTION
As shown below, the dspidualmul operation computes two 16-bit clipped, signed products separately on the two
pairs of high and low 16-bit halfwords of r
src1 and rsrc2. Both products are clipped into the range [215–1..–215] (or
[0x7fff..0x8000]) and written into the corresponding halfwords of r
dest. All values are signed 16-bit integers.
The dspidualmul operation optionally takes a guard, specied in r
guard. If a guard is present, its LSB controls the
modication of the destination register. If the LSB of r
guard is 1, rdest is written; otherwise, rdest is not changed.
EXAMPLES
Initial Values
Operation
Result
r30 = 0x0020010, r40 = 0x00030020
dspidualmul r30 r40
→ r60
r60
← 0x00060200
r10 = 0, r30 = 0x0020010, r40 = 0x00030020 IF r10 dspidualmul r30 r40
→ r70
no change, since guard is false
r20 = 1, r30 = 0x0020010, r40 = 0x00030020 IF r20 dspidualmul r30 r40
→ r100 r100 ← 0x00060200
r50 = 0x80000002, r80 = 0x00024000
dspidualmul r50 r80
→ r90
r90
← 0x80007fff
r110 = 0x08000003, r120 = 0x00108001
dspidualmul r110 r120
→ r125
r125
← 0x7fff8000
0
1
3
r
src1
0
1
3
r
src2
0
3
r
dest
×
1
0
3
0
3
Two full-precision
32-bit signed products
Clip to [215–1..–215]
signed
SEE ALSO
dspidualabs dspidualadd
dspidualsub dspiabs
dspidualmul