Beruflich Dokumente
Kultur Dokumente
12/9/14
12/9/14
12/9/14
12/9/14
12/9/14
12/9/14
12/9/14
10
12/9/14
11
12/9/14
12
12/9/14
13
14
y[n] = i=0N-1h[i]x[n-i] !
Note that an input array of size N is all that can
be used for x[n]!
Computation of y[n] requires x[n],,x[n-N+1]
but y[n+1] requires x[n+1]..,x[n-N+2] (x[n-N
+1] is no longer needed)!
12/9/14
15
16
12/9/14
17
12/9/14
18
12/9/14
19
12/9/14
20
12/9/14
21
12/9/14
22
12/9/14
23
Filtering
DFT:
!
It
direct implementation!
24
Nlog2N
25
12/9/14
26
12/9/14
27
12/9/14
28
29
12/9/14
30
12/9/14
31
12/9/14
32
12/9/14
33
12/9/14
34
12/9/14
35
12/9/14
36
12/9/14
12/9/14
12/9/14
12/9/14
12/9/14
12/9/14
12/9/14
10
000000
00000
----------------------------------------11111010 = -610
If we go back to the 4 bit representation the result wil
Example:
Compute
in fixed point on 4 bi
Multiplication
of
-3
athe
nd
6product
gives
1101 = -310
0110 = 610
----------------------------------------00000000
1111101
111101
00000
----------------------------------------11101110 = -1810
11
7+6-8 = 5
0111+
7+
----------------------------------------00000000
1111101
111101
00000
----------------------------------------11101110 = -1810
nd be 11102 = -21
If we
go back to the 4 bit representation the result will
----------------------------------1101+
-3+ (overflow)
1000
-8
----------------------------------0101
5 (correct)
12
12/9/14
"!
12/9/14
13
12/9/14
14
12/9/14
12/9/14
16
12/9/14
17
12/9/14
18
19
To
20
0.110 = 0.7510
--------------------------------------00000000
1111100
111100
00000
--------------------------------------11.1010002 = -0.37510
Supplementary sign bit
We go back to the 4 bits representation:
In this1|1.101|000
example, there is no
truncation. 1.101
right! 10 which is a
-0.375
2 =1.101
2 = is
d)
12/9/14
21
"!
12/9/14
22
When
23
12/9/14
24
12/9/14
25
12/9/14
26
12/9/14
27
12/9/14
28
29
12/9/14
30
12/9/14
31
The
32
The
33
"
"
34
Floating
35
"
"
"
36
Power
37
Floating
38
12/9/14
39
Assume
12/9/14
40
This
41
When
42
12/9/14
43
Extensive
12/9/14
44
12/9/14
45
46
It
47
12/9/14
Feature(s) Benefit(s)!
A 32 x 16-bit Instruction buffer queue!
Two 17-bit x17-bit MAC units!
One 40-bit ALU!
One 40-bit Barrel Shifter!
One 16-bit ALU!
Four 40-bit accumulators!
Twelve independent buses:!
Three data read buses!
Two data write buses!
Five data address buses!
One program read bus!
One program address bus!
User-configurable IDLE Domains!
12/9/14
!
!
12/9/14
!
!
12/9/14
decodes
Queue
holds
64
bytes
at
a
time
Block
local
repeats
executed
within
Speculative
fetches
during
conditional
calls/
branches
Identies
code
boundaries
(8,16,24,32,40
and
48
bits)
Determines
parallel/non-parallel
Sends
data
to
D
and
A
units
12/9/14
12/9/14
12/9/14
All
10
12/9/14
11
12/9/14
12
12/9/14
13
12/9/14
14
12/9/14
15
12/9/14
16
12/9/14
17
18
19
12/9/14
20
12/9/14
21
22
12/9/14
23
12/9/14
24
12/9/14
25
12/9/14
26
12/9/14
27
12/9/14
28
12/9/14
29
12/9/14
30
12/9/14
31
12/9/14
32
12/9/14
33
34
12/9/14
35
12/9/14
36
12/9/14
37
12/9/14
38
12/9/14
39
12/9/14
40
If
41
12/9/14
42