Outline Digital CMOS Design Arithmetic Operators Adders - PowerPoint PPT Presentation
Outline Digital CMOS Design Arithmetic Operators Adders Comparators Shifters Multipliers Square rooting Pirouz Bazargan Sabet Digital Design February 2010 Square root An real number using floating point representation
Outline Digital CMOS Design Arithmetic Operators Adders Comparators Shifters Multipliers Square rooting Pirouz Bazargan Sabet Digital Design February 2010
Square root An real number � � using floating point representation � � Find a real number � � such as � � ( ) 2 + ε = x y o Calculation cannot be implemented in hardware o Need iterative operation Pirouz Bazargan Sabet Digital Design February 2010
Square root o Direct method digit-by-digit 2 o Indirect method − y = x 0 resolve Pirouz Bazargan Sabet Digital Design February 2010
Square root - direct Find x such as x = y [ [ ( ) 0 E ∈ M 1 , 2 = − × × 1 2 y M With E ( ) 0 = − × × 1 2 2 y M E odd ? ( ) 0 ′ [ [ 2 E ′ = − × × ′ y 1 M 2 ∈ M 1 , 4 With ( ) 0 ′ E ′ = = − × × x y 1 M 2 ( ) [ [ 0 ′ E = − × × x 1 X 2 ∈ 1 , 2 X With Pirouz Bazargan Sabet Digital Design February 2010
Square root - direct ( ) 0 ′ [ [ 2 E ′ = − × × ′ y 1 M 2 ∈ M 1 , 4 With x = Find x such as y ( ) [ [ 0 ′ E = − × × x 1 X 2 ∈ X 1 , 2 With − n � i = × X x 2 i = i 0 Iterate on i and evaluate x i Pirouz Bazargan Sabet Digital Design February 2010
Square root - direct Pirouz Bazargan Sabet Digital Design February 2010
Square root o Direct method digit-by-digit 2 o Indirect method − y = x 0 resolve Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect Resolving a non ( ) f x y linear equation tangent ( ) = f x 0 initial guess x new estimation Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect ( ) = f x 0 Resolving a non linear equation x Taylor series in the neighborhood of 0 1 ( ) ( ) ( )( ) ( )( ) 2 � ′ ′ ′ = + − + − + f x f x f x x x f x x x 0 0 0 0 0 2 ( ) ( ) ( )( ) 1 st order : ′ ≈ + − f x f x f x x x 0 0 0 Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect ( ) = f x 0 Resolving a non linear equation Iterative resolution starting from an initial guess x 0 ( ) ( ) ( )( ) ′ ≈ + − f x f x f x x x 0 0 0 ( ) ( ) ( )( ) ′ = + − = f x 0 f x f x x x 0 0 0 0 ( ) − f x 0 = + x x ( ) 0 ′ f x 0 Newton-Raphson method Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect Resolving x = y ( ) = f x 0 x = y Find a function f such as for ( ) 2 = − f x x y ( ) ( ) ( )( ) ′ ≈ + − f x f x f x x x 0 0 0 ( ) ( ) ( ) 2 ≈ − + − f x x y 2 x x x 0 0 0 ( ) � � 2 − y x 1 y ( ) � � = f x 0 0 = + = + x x x x � � 0 0 � � 2 x 2 x 0 0 Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect Resolving x = y � � 1 y � � Each iteration = + x x � � + i 1 i � � 2 x i division !! Hard to implement ( ) 2 = − f x x y Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect Resolving x = y 1 ( ) ( ) = = = f f u x 0 0 u x = y Find a function f such as for y ( ) − 2 = − f u u y ( ) ( ) ( )( ) ′ ≈ + − f u f u f u u u 0 0 0 ( ) ( ) − − ( ) 2 3 ≈ − − − f u u y 2 u u u 0 0 0 ( ) − 2 − u y 1 ( ) ( ) 2 = f u 0 0 = + u u = − u u 0 3 u y 0 − 3 0 2 u 2 0 Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect Resolving x = y 1 ( ) 2 Each iteration = − u u 3 u y + i 1 i i 2 multiply !! 1 y = = = = ⋅ u x y u y y y Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect 1 Resolving x = = y u y [ [ ( ) 0 E ∈ M 1 , 2 = − × × 1 2 y M with E ( ) 0 = − × × 1 2 2 y M E odd ? ( ) 0 ′ [ [ 2 E ′ = − × × ′ y 1 M 2 ∈ M 1 , 4 with ( ) 0 ′ E ′ = − × × y 1 M 2 ′ M ( ) 0 ′ E = − × × 1 2 y ′ M Pirouz Bazargan Sabet Digital Design February 2010
Square root - indirect 1 Resolving x = = y u y Initial guess 1 [ [ ′ ∈ M 1 , 4 = u with M ′ ′ M ] ] ∈ u 0 . 5 , 1 4 2 u 1 1 0.5 0.707 Pirouz Bazargan Sabet Digital Design February 2010
Recommend
More recommend
Explore More Topics
Stay informed with curated content and fresh updates.