CSE 246: Computer Arithmetic

Algorithms and Hardware Design

Winter 2005
Lecture 2: Redundant and
residue number systems

Instructor:
Prof. Chung-Kuan Cheng
Topics:
 Redundant Number Systems
 Residue Number Systems
 Mixed Radix Number Systems (as they
pertain to conversions of Residue
Number Systems)

CSE 246 2
Redundant Number Systems
 Examine number systems that can be
characterized by the 3-tuple (r, , )
 r : radix, [, ] : set of digits

 The redundancy of such a system is said to be


non-redundant if  -  = r

redundant if  -  > r
 minimally redundant if  -  = r + 1
 maximally redundant if  -  =2r - 1
 over redundant if  -  > 2r - 1

CSE 246 3
Redundant Number Systems
 Ex: (2,0,2)
 Redundant - furthermore, minimally redundant
4 2 1

Addition example: 0 0 0 0
0 0 1 1
0 0 2 2
0 1 0 2
1 0 2 (6) 0 1 1 3
0 1 2 4
+1 1 1 (7) 0 2 0 4
0 2 1 5
2 1 3 [3 not in radix] 0 2 2 6
1 0 0 4
+ 1 1
2 2 1 = 4*2 + 2*2 +1 = 13 = 6+7

CSE 246 4
Redundant Number Systems
 More addition examples:
110 121 101121 111222
+102 +101 +112210 +112222
212 222 111 20
11 20
11 20
21 11
222211 22
11220
11
2
1121220

CSE 246 5
Redundant Number Systems
 What has redundancy bought us?
 From the examples, it seems carry
propagation has been somewhat reduced
 Intuitively - redundancy acts as a buffer
against carry propagation
 However, carry propagation was not
eliminated in (2,0,2)
 Need to formalize the degree to which a
particular degree of redundancy eliminates
carry propagation

CSE 246 6
Redundant Number Systems
 To formalize:
xi+1 xi where xi + yi = r*ti+1 + wi
+ yi+1 yi
ti+1 wi
ti+2 wi+1

ti+1 + wi+1

we must bound ti+1 + wi+1 for all i to eliminate

carry propagation

CSE 246 7
Redundant Number Systems
As xi and yi are in [, ] and xi + yi = r * ti+1 + wi+1
1 2  r*ti+1 + wi  2

suppose there exist  and  such that

   ti+1  
we need
  -   wi+1   - 
to avoid propagation. Since this must be true for all i, whether the
subscript is i or i+1 is not important.

From 2 & 3 we have

4 r* +  -   r*ti+1 + wi+1  r* +  - 

From 1 & 4 we have

5
CSE 246
2  r* +  -  8
Redundant Number Systems
5 2  r* +  - 
gives the following bounds for  &  to insure no propagation

  /(r-1)
  /(r-1)

For (2,0,2):
  /(r-1) ->   2/(2-1)
  /(r-1) ->   0/(2-1)

which, from 3 means all wi must be in the range [0,0] and

there exist inputs requiring carry propagation

CSE 246 9
Redundant Number Systems
 How then to ensure that no carry propagation is needed?
 Change (r, , )

Consider (3,0,5)
  /(r-1) ->   5/(3-1)
  /(r-1) ->   0/(3-1) yields
 : 3 (first integer after 5/2)
:0
From 3, this gives a range on wi of [0,2]

Thus, when presented with multiple possibilities for

representing the sum of two digits, always choose terms in
[0,2] and there will be no carry propagation

CSE 246 10
Redundant Number Systems
Ex: for (3,0,5)
243
+435 5+3 = 8 could use 15, but choose from [0,2]
22 yielding 22 and no propagation
21
20
2232

CSE 246 11
Redundant Number Systems
 Advantages
 Constant time addition/subtraction!
 Disadvantages
 More space
 Comparison
 no unique forms/representations
 conversion to canonical form as expensive as
summation
 Uses:
 excellent for intermediate results of addition
 Multiplication
 DSP loops that have no comparisons
CSE 246 12
Residue Number Systems
 Define a Residue Number System as follows

For any given integer x, x=(x1|x2|...|xk)RNS(P1|P2|...|Pk)

where xi = x mod Pi and  i,j Pi is relatively prime to Pj
EX: 84 = (0|4|0)RNS(7|5|3)
1 = (1|1|1)RNS(7|5|3)
2 = (2|2|2)RNS(7|5|3)
3 = (3|3|0)RNS(7|5|3)

Residue numbers are not positional.

Residue numbers have unique representations mod  Pi

i.e. for (7|5|3) there are unique representations of 0-104.

CSE 246 13
Residue Number Systems
 Need a conversion system to/from binary

Binary #

RNS
ops RNS #
+,-,*

Binary #

 What benefit to doing operations in RNS?

CSE 246 14
Residue Number Systems
 Addition (subtraction is similar)

x+y = ((x1+y1)p | (x2+y2)p | ... | (xk+yk)p )RNS(P1|P2|...|Pk)

1 2 k

where xi = (x)p and yi = (y)p

i i

 Multiplication
x*y = ((x1*y1)p | (x2*y2)p | ... | (xk*yk)p )RNS(P1|P2|...|Pk)
1 2 k

where xi = (x)p and yi = (y)p

i i

 Division ?
 Hard. What does a fraction look like in RNS?
CSE 246 15
Residue Number Systems
 Advantages:
 Parallel processing of +,*,- on smaller numbers
 Adding more primes without increasing range allows for
use of fields to assist in fault tolerance (Ex: go from (7|
5|3) to (7|5|3|2) where last field is used for parity)
 Disadvantages:
 No division
 Comparison non-trivial
 Conversion costs

 How to do conversion?

CSE 246 16
Residue Number Systems
 Conversion: RNS -> Binary #

Given (x1|x2|...|xk)RNS(P1|P2|..|Pk)


k
 P j 
Binary number x =
 xi i

i 1  Pi

mod

P j

  Pj    P 
 ,P  i j See: Chinese
where i = inv  Pi i    Pi  1 Remainder Theorem
  Pi

One may consider, for the purposes of computation, all i as

being a pre-defined part of each particular RNS system
CSE 246 17
Residue Number Systems
 Conversion -> Bin# example:
84 = (0|4|0)RNS(7|5|3)

For (7|5|3):
(1* 5*3/7)7 = 1  1 : 1
(2* 7*3/5)5 = 1  2 : 1
(3* 7*5/3)3 = 1  3 : 2

x = (0*1*5*3*7/7 + 4*2*7*3*5/5 + 0*3*7*5*3/3)105

x = (0 + 4*2*21 + 0)105

x = (4*1*21)105 = 84105

CSE 246 18
Residue Number Systems
 Conversion -> Bin # example:
1 = (1|1|1)RNS(7|5|3)

(1* 5*3/7)7 = 1  1 : 1
(2* 7*3/5)5 = 1  2 : 1
(3* 7*5/3)3 = 1  3 : 2

x = (1*1*5*3*7/7 + 1*2*7*3*5/5 + 1*3*7*5*3/3)105

x = (1*1*15 + 1*2*21 + 1*3*35)105

x = (1*1*15+1*1*21+1*2*35)105=(106)105 = 1105

CSE 246 19
 Residue Number Systems
Binary # Conversion -> Chinese remainder theorem

RNS How to choose primes?

ops RNS #
+,-,*

Binary #
 Minimize:
 total length of primes (determining range)
 max length of any prime (determining HW
cost/delay per unit)
 # of primes (determining # of parallel units)

What about  Popular choices: 2r-1 (2,3 particularly popular)

Bin -> RNS?  Thm: 2a-1 & 2b-1 are relatively prime iff a & b
are relatively prime

CSE 246 20
Residue Number Systems
 Binary # to RNS conversion
 Uses a lookup table
Observation: for n digit binary number y
(yn-1,yn-2,...,y0)p = [(2n-1)p *yn-1+ (2n-2)p *yn-2+...+(20)p *y0]p
i i i i i

Use a table to store (2i)p i 2i 2i7 2i5 2i3

i
0 1 1 1 1
1 2 2 2 2
Example for (7|5|3) 2 4 4 4 1
3 8 1 3 2
(1011)=(x1|x2|x3)RNS(7|5|3)
x1 = (1+2+1)7 = 4
x2 = (3+2+1)5 = 1
x = (2+2+1)3 = 2
3
CSE 246 21
Residue Number Systems
 Comparison
 Could convert back and forth to/from binary.
 Another approach: convert to a mixed radix
system, as numbers in a mixed radix system
are comparable.

CSE 246 22
Mixed Radix Number
Systems
 We shall describe a Mixed Radix System
as follows:

x =(Zk-1|Zk-2|...Z0)MRS(Pk-1|Pk-2|...|P1)

x = Zk-1Pk-1Pk-2...P1 + Zk-2Pk-2Pk-1...P0 + ... + Z1P1+Z0

CSE 246 23
Mixed Radix Number
Systems
 Conversion from RNS to MRS

Given x = (xk-1|xk-2|...|x0)RNS(Pk-1|Pk-2|...|P0)

We want x=(Zk-1|Zk-2|...|Z0)MRS(Pk-2|Pk-3|...|P0)

Observation: The MRS digit Z0 is in units of 1, so

fewer primes needed in MRS than in RNS

CSE 246 24
Mixed Radix Number
Systems
 Question: what is the relationship
between x0 and Z0?

 Can it be found by simple inspection?

 Yes.
1 x = Z .
0 0
 Why?
 x0 is the residue left from x mod P0 - all other
terms are multiples of P0 -> x0 = Z0

CSE 246 25
Mixed Radix Number
Systems
 This yields
2 x-x0 = (x’k-1|x’k-2|...|x’1|-)RNS(Pk-1|Pk-2|...|P1|-)
= (Zk-1|Zk-2|...|Z1|0)MRS(Pk-2|Pk-3|...|P0)

where x’i = (xi-x0)p i

Note that this is the only change in MRS

CSE 246 26
Mixed Radix Number
Systems
 Which leads to
3 (x-x0)/P0 = (x”k-1|x”k-2|...|x”1|-)RNS(Pk-1|Pk-2|...|P1|-)
= (Zk-1|Zk-2|...|Z1) MRS (Pk-2|Pk-1|...|P1)

Z1 = x”1

and so forth. (Deduction, division, repeat)

However, it was earlier noted that division in RNS is

hard - yet here we are doing division.
The trick? In this case, we know that we will always
get integer results.
CSE 246 27
Back to RNS
 Division in RNS
x”i =xi * (P0-1)p i

where

(Pj-1)p is the multiplicative inverse of Pj with respect to Pi

i

Ex: (3-1)7 = 5 --> (3*(3-1)7)7 = 1

(3-1)5 = 2 --> (3*(3-1)5)5 = 1

CSE 246 28
Example: RNS -> MRS
Y=(1|3|2)RNS(7|5|3) = (Z2|Z1|Z0)MRS(5|3)
That is, Y = Z2*5*3+Z1*3+Z0

by 1 (see slide #25)

Z0 = x0 = 2

from 2 (see slide #26) we have

y-x0 = y-2 =(x’2|x’1|0)RNS(7|5|3) = (Z2|Z1|0)MRS(5|3)

x’2 = (x2-x0)p = (1-2)7 = 6
2

x’1 = (3-2)5 = 1
y-2 = (6|1|0)RNS(7|5|3)=(Z2|Z1|0)MRS(5|3)
CSE 246 29
Example continued.
3 (See slide #27) then gives

(y-x0)/P0= (y-2)/3 = (x”2|x”1|-)RNS(7|5|3) = (Z2|Z1)MRS(5)

'
 7 A  x 
 7 B  x2" 7
then
 2

 3 7
Q: How does one derive x”2 ?
A: “It’s hard.” One has to try values one by one up to the
modulus

7 A  x  21B  3x 
'
2 7
"
2 7
remembering that (3-1)7 = 5, (3-1)5 = 2
x”2= (5*6)7 = 2, x”1 =(1*2)5 = 2
(y-2)/3 = (2|2|-)RNS(7|5|3) = (Z2|Z1)MRS(5)
CSE 246 30
Ex. Cont.
Apply 1 again,
Z1=x”1=2
gives by 2
(y-x0)/P0 - x”1 = (x”’2|0|-)RNS(7|5|3) = (Z2|0)MRS(5)

x”’2 =(x”2-2)7 = (2-2)7 = 0

(0|0|-)RNS(7|5|3)=(Z0|0)MRS(5)

By 3
y  x0 " x””2 = (x”’2*(5-1)7)7 = 0 = Z2
 x1
P0
...
P1 Yields final result
(0|2|2)MRS(5|3)

CSE 246 31
Ex. Concluded.
Check correctness

(0|2|2)MRS(5|3) = 0*5*3+2*3+2 = 8
87 = 1, 85 = 3, 83 = 2 -> (1|3|2)RNS(7|5|3)

Correct!

 Closing remark / Moral of the examples -

“Inversion is key”

CSE 246 32