Boolean Function Representation based on disjoint-support decompositions. Λ

Similar documents
, each of which is a tree, and whose roots r 1. , respectively, are children of r. Data Structures & File Management

Present state Next state Q + M N

Boolean Function Representation Based on Disjoint-Support Decompositions.

ECE COMBINATIONAL BUILDING BLOCKS - INVEST 13 DECODERS AND ENCODERS

1 Introduction to Modulo 7 Arithmetic

Outline. 1 Introduction. 2 Min-Cost Spanning Trees. 4 Example

Constructive Geometric Constraint Solving

Outline. Computer Science 331. Computation of Min-Cost Spanning Trees. Costs of Spanning Trees in Weighted Graphs

Cycles and Simple Cycles. Paths and Simple Paths. Trees. Problem: There is No Completely Standard Terminology!

Outline. Binary Tree

EE1000 Project 4 Digital Volt Meter

CSE 373: More on graphs; DFS and BFS. Michael Lee Wednesday, Feb 14, 2018

0.1. Exercise 1: the distances between four points in a graph

(2) If we multiplied a row of B by λ, then the value is also multiplied by λ(here lambda could be 0). namely

Module graph.py. 1 Introduction. 2 Graph basics. 3 Module graph.py. 3.1 Objects. CS 231 Naomi Nishimura

Seven-Segment Display Driver

CS September 2018

An undirected graph G = (V, E) V a set of vertices E a set of unordered edges (v,w) where v, w in V

Designing A Concrete Arch Bridge

CS 461, Lecture 17. Today s Outline. Example Run

COMPLEXITY OF COUNTING PLANAR TILINGS BY TWO BARS

Garnir Polynomial and their Properties

CSE 373. Graphs 1: Concepts, Depth/Breadth-First Search reading: Weiss Ch. 9. slides created by Marty Stepp

QUESTIONS BEGIN HERE!

The University of Sydney MATH2969/2069. Graph Theory Tutorial 5 (Week 12) Solutions 2008

ECE 407 Computer Aided Design for Electronic Systems. Circuit Modeling and Basic Graph Concepts/Algorithms. Instructor: Maria K. Michael.

learning objectives learn what graphs are in mathematical terms learn how to represent graphs in computers learn about typical graph algorithms

Similarity Search. The Binary Branch Distance. Nikolaus Augsten.

Organization. Dominators. Control-flow graphs 8/30/2010. Dominators, control-dependence. Dominator relation of CFGs

CS 241 Analysis of Algorithms

Why the Junction Tree Algorithm? The Junction Tree Algorithm. Clique Potential Representation. Overview. Chris Williams 1.

Algorithmic and NP-Completeness Aspects of a Total Lict Domination Number of a Graph

Exam 1 Solution. CS 542 Advanced Data Structures and Algorithms 2/14/2013

16.unified Introduction to Computers and Programming. SOLUTIONS to Examination 4/30/04 9:05am - 10:00am

QUESTIONS BEGIN HERE!

V={A,B,C,D,E} E={ (A,D),(A,E),(B,D), (B,E),(C,D),(C,E)}

b. How many ternary words of length 23 with eight 0 s, nine 1 s and six 2 s?

CSC Design and Analysis of Algorithms. Example: Change-Making Problem

Paths. Connectivity. Euler and Hamilton Paths. Planar graphs.

Complete Solutions for MATH 3012 Quiz 2, October 25, 2011, WTT

Planar Upward Drawings

Trees as operads. Lecture A formalism of trees

MAT3707. Tutorial letter 201/1/2017 DISCRETE MATHEMATICS: COMBINATORICS. Semester 1. Department of Mathematical Sciences MAT3707/201/1/2017

V={A,B,C,D,E} E={ (A,D),(A,E),(B,D), (B,E),(C,D),(C,E)}

Problem solving by search

CSE 373: AVL trees. Warmup: Warmup. Interlude: Exploring the balance invariant. AVL Trees: Invariants. AVL tree invariants review

Multipoint Alternate Marking method for passive and hybrid performance monitoring

Graph Isomorphism. Graphs - II. Cayley s Formula. Planar Graphs. Outline. Is K 5 planar? The number of labeled trees on n nodes is n n-2

Using the Printable Sticker Function. Using the Edit Screen. Computer. Tablet. ScanNCutCanvas

S i m p l i f y i n g A l g e b r a SIMPLIFYING ALGEBRA.

N=4 L=4. Our first non-linear data structure! A graph G consists of two sets G = {V, E} A set of V vertices, or nodes f

Math 61 : Discrete Structures Final Exam Instructor: Ciprian Manolescu. You have 180 minutes.

A Graph-Based Synthesis Algorithm for AND/XOR Networks 1. Yibin Ye Kaushik Roy. a minimum ESOP of a function except for those with

Integration Continued. Integration by Parts Solving Definite Integrals: Area Under a Curve Improper Integrals

Section 10.4 Connectivity (up to paths and isomorphism, not including)

# 1 ' 10 ' 100. Decimal point = 4 hundred. = 6 tens (or sixty) = 5 ones (or five) = 2 tenths. = 7 hundredths.

(a) v 1. v a. v i. v s. (b)

XML and Databases. Outline. Recall: Top-Down Evaluation of Simple Paths. Recall: Top-Down Evaluation of Simple Paths. Sebastian Maneth NICTA and UNSW

12/3/12. Outline. Part 10. Graphs. Circuits. Euler paths/circuits. Euler s bridge problem (Bridges of Konigsberg Problem)

5/9/13. Part 10. Graphs. Outline. Circuits. Introduction Terminology Implementing Graphs

CS61B Lecture #33. Administrivia: Autograder will run this evening. Today s Readings: Graph Structures: DSIJ, Chapter 12

DFA Minimization. DFA minimization: the idea. Not in Sipser. Background: Questions: Assignments: Previously: Today: Then:

CS200: Graphs. Graphs. Directed Graphs. Graphs/Networks Around Us. What can this represent? Sometimes we want to represent directionality:

Graph Contraction and Connectivity

Round 7: Graphs (part I)

12. Traffic engineering

Graphs. Graphs. Graphs: Basic Terminology. Directed Graphs. Dr Papalaskari 1

A 4-state solution to the Firing Squad Synchronization Problem based on hybrid rule 60 and 102 cellular automata

Greedy Algorithms, Activity Selection, Minimum Spanning Trees Scribes: Logan Short (2015), Virginia Date: May 18, 2016

5/7/13. Part 10. Graphs. Theorem Theorem Graphs Describing Precedence. Outline. Theorem 10-1: The Handshaking Theorem

OpenMx Matrices and Operators

A Simple Code Generator. Code generation Algorithm. Register and Address Descriptors. Example 3/31/2008. Code Generation

Numbering Boundary Nodes

Limits Indeterminate Forms and L Hospital s Rule

Tableau algorithms defined naturally for pictures

Tangram Fractions Overview: Students will analyze standard and nonstandard

BDD-BASED LOGIC OPTIMIZATION SYSTEM

Outline. Circuits. Euler paths/circuits 4/25/12. Part 10. Graphs. Euler s bridge problem (Bridges of Konigsberg Problem)

CMSC 451: Lecture 2 Graph Basics Thursday, Aug 31, 2017

Decimals DECIMALS.

Yehuda Lindell Bar-Ilan University

SAMPLE PAGES. Primary. Primary Maths Basics Series THE SUBTRACTION BOOK. A progression of subtraction skills. written by Jillian Cockings

Logic Redundancy Identication. September 14, target specic faults and are sometimes referred to as. \fault-independent methods.

5/1/2018. Huffman Coding Trees. Huffman Coding Trees. Huffman Coding Trees. Huffman Coding Trees. Huffman Coding Trees. Huffman Coding Trees

Graphs. CSC 1300 Discrete Structures Villanova University. Villanova CSC Dr Papalaskari

Liqiong Wei, Zhanping Chen, and Kaushik Roy. Yibin Ye and Vivek De. logic circuits more feasible.

BDD-BASED LOGIC OPTIMIZATION SYSTEM

New challenges on Independent Gate FinFET Transistor Network Generation

for Performance Potsdam, NY Urbana, IL can be connected to the inputs and outputs of the LUT to

Math 166 Week in Review 2 Sections 1.1b, 1.2, 1.3, & 1.4

SOLVED EXAMPLES. be the foci of an ellipse with eccentricity e. For any point P on the ellipse, prove that. tan

Fundamental Algorithms for System Modeling, Analysis, and Optimization

Grade 7/8 Math Circles March 4/5, Graph Theory I- Solutions

Chapter 18. Minimum Spanning Trees Minimum Spanning Trees. a d. a d. a d. f c

RAM Model. I/O Model. Real Machine Example: Nehalem : Algorithms in the Real World 4/9/13

NP-Completeness. CS3230 (Algorithm) Traveling Salesperson Problem. What s the Big Deal? Given a Problem. What s the Big Deal? What s the Big Deal?

The University of Sydney MATH 2009

MULTIPLE-LEVEL LOGIC OPTIMIZATION II

Steinberg s Conjecture is false

DUET WITH DIAMONDS COLOR SHIFTING BRACELET By Leslie Rogalski

Transcription:

Booln Funtion Rprsnttion s on isjoint-support ompositions. Λ Vlri Brto n Murizio Dmini Diprtimnto i Elttroni Inormti Univrsitá i Pov, Vi Grnio 6/A, 353 Pov, ITALY -mil: rin@i.unip.it mini@i.unip.it Tl: +39 49 827 7829 F: +39 49 827 7699 Astrt Th Multi-Lvl Domposition Dirms (s) prsnt in this ppr provi nonil rprsnttion o Booln untions whil mkin pliit isjoint-support omposition. This rprsnttion n irtly mpp to nonil multi-lvl t ntwork o loi untion with AND/OR or NOR-only (NAND-only) ts. Usin s w r l to ru th mmory ouption with rspt to tritionl ROBDDs or svrl nhmrk untions, y omposin loi untions rursivly into simplr - n mor onivisil - omponnts. Bus o this proprty, nlysis o th rphs llow us to somtims intiy nw n ttr vril orrin or svrl nhmrk iruits. W pt th proprtis o s to usul in svrl ontts, most notly loi synthsis, thnoloy mppin, n squntil hrwr vriition. Introution Ru, Orr Binry Dision Dirms (ROB- DDs) [] r proly th most powrul t strutur known so r or th mnipultion o lr loi untions, n or this rson thy hv om prvsiv in loi synthsis n vriition nvironmnts [2, 3, 4, 5]. Onoin rsrh is ttmptin to tn thir ppliility to othr omins, suh s th solution o rph prolms n intr-linr prormmin [6, 7]. Still, som ky iniinis (n ponntil lowup or som lsss o untions, th unpritility o th ROBDD siz n shp with rspt to th vril orrin hosn, t...) motivt n inrsin rsrh tivity in this r. Rsrh irtions inlu in prtiulr: Eiint implmnttions [8, 9], vlopmnt o orrin huristis [,, 2], n ltrntiv rprsnttions ltothr [3, 4, 5, 6, 7]. In [8], th uthors prsnt n ition to th si ROBDD rprsnttion, s on th nloy o ROBDDs with trministi init utomt. Th nw rprsnttion ws ountrprt o nontrministi utomton (hn possily mor ompt), in whih untion root t h ROBDD no ws rprsnt s loi OR o simplr, isjoint-support omponnts. This rsrh ws prtilly support y th ESPRIT III Bsi Rsrh Prormm o th EC unr ontrt No. 972 (Projt GEPP- COM) n y CNR rnt # 95.26.CT7 In this ppr, w to th si ROBDD rprsnttion th pility o isovrin th prsn o ritrry, multipl-lvl tr ompositions o untions. Th rprsnttion shrs with ROBDDs noniity, irtyli rph strutur, n rursiv onstrution thniqu. Unlik ROBDDs, howvr, nos my rprsnt not only two-input MUXs, ut lso unlimit-nin OR / AND (or NAND-only, NOR-only) ts. It is worth notin tht, us o t-lik nos, our rprsnttion is ssntilly multipl-lvl iruit. Throuh th us o multipl-lvl NOR-only (or NAND-only) omposition, w mintin onstnt-tim omplmnttion; n us o th tr omposition, th rprsnttion is siniintly lss orr-snsitiv thn ROBDDs. In prtiulr, w intii lss o untions or whih th rprsnttion is totlly inpnnt rom th vril orr hosn, n or whih som iiult prolms (lik, Booln NPN mthin [9]) n solv in linr tim. Ths turs rprsnt sustntil improvmnt ovr th work [8], whr sinl-lvl OR omposition ws us, n omplmnttion ws iiult. Eprimntlly, w oun tht th nw rprsnttion is mmorywis siniintly mor ompt thn ROBDDs, us omposl untions n shr omponnts. Mor intrstinly, howvr, th nw rprsnttion ivs us som systmti n t insiht on th rol o th input vrils o loi untion. This insiht is rr to spil-purpos huristis (suh s ynmi rorrin) in th s o ROBDDs. Th rst o th ppr is orniz s ollows. Stions 2 n 3 introu isjoint-support omposition n s, rsptivly. Stion 4 sris th prours us or mnipultion, n vntully Stion 5 prsnts th primntl rsults. Proos o Thorms r rr to th Appni, or th sk o rility. 2 Disjoint support omposition Th rprsnttion prsnt in this ppr is s on th notion o tr omposition o untion. In this stion, w introu th si initions onrnin this omposition. W onsir th omposition o untions into th NOR (NAND, OR, AND) o isjoint-support suuntions, whnvr possil. This notion will l to rursiv (.. tr) omposition styl n to th inition o Multi Lvl Domposition Dirms (s).

Dinition. Lt : B n! B not non-onstnt Booln untion o n vrils ; ; n. W sy tht pns on i i @=@ i is not intilly. W ll support o (init y S( )) th st o Booln vrils pns on. 2 Dinition 2. A st o non-onstnt untions ; k, k, with rsptiv supports S( i ) is ll isjointsupport NOR omposition o i: + + k = ; S( i ) S( j )=;; i 6= j() A isjoint support NOR omposition is miml i no untion i is urthr omposl in th OR o othr untions with isjoint support. W in isjoint support OR, AND, NAND ompositions in similr shion. W init y D NOR () ny suh miml omposition. 2 Empl. Th untion = ( + )( + ) hs th ollowin isjoint-support ompositions: 2 l AND: =( + ) n 2 =( + ); l NOR: =( + ) n 2 =( + ) ; l NAND n OR:. In th rst o th ppr, isjoint-support ompositions r rrr to s ompositions, or short. Morovr, w ous only on NOR omposition, s th rsults or th othr ompositions n otin rily y stnr Booln lr. 2. Tr ompositions Domposition n ppli rursivly to loi untions. In this s, w otin rprsnttion o F s on NOR tr. Empl 2. Th untion F =( + )( + + ) th omposition tr. Similrly w n in TD NAND n TD AND/OR. 2 Thorm () low stts n intuitiv ut rlvnt rsult. Thorm. For ivn untion, th ollowin proprtis hol:. thr is uniqu D NOR ; 2. thr is uniqu TD NOR. 2 2.2 Tr-omposl untions. Whn omposin untion, it my possil tht th lvs o th omposition ru only to primry inputs or thir omplmnts. This is th s, or mpl, o th untion F =(+)(++) =[(+) +(++) ]. Dinition 4. A loi untion (;:::; n ) is tromposl i th input suuntions o its TD NOR lon to th st ;:::; n ; ;:::; n, i.. th st o inputs vrils n thir omplmnts. 2 I untion is tr omposl, thn Thorm () inits tht its omposition tr TD NOR is nonil rprsnttion. Not vry untion,howvr, is tromposl. For instn, th untion F = + nnot rprsnt s th NOR o ny isjoint-support suuntions. Hn, NOR omposition trs r not univrsl rprsnttion styl. W n nlr, howvr, th st o tr-rprsntl untions s ollows. Somtims th omplmnt o non-omposl untion my omposl. In this s th omplmnt F = ( + )( + + ) is in tr-omposl. W n thus ploit th omposility o F in rprsntin th D NOR o F y simply ppnin NOT t t th root o th NOR tr. Fi. (2) shows th rprsnttion. Fiur 2. Rursiv omposition with NOT t root. Fiur. A rursivly omposl untion. is rursivly NOR omposl. From th irst omposition w otin = ( + ) n 2 = [ +( + ) + ( + ) ]. Ths untions r thn in omposl until rhin th input vrils, s rport in Fi. (). 2 Dinition 3. A tr omposition o loi untion is rursiv omposition o into NOR-only tr o suuntions, whr th untions t th inputs o h NOR r mimlly ompos. W init y TD NOR Th only rminin qustion is whthr th introution o NOT ts t th root prsrvs noniity, tht is, whthr tr-omposl untion n hv two omposition trs, on with NOT t th root n nothr without it. To this rr, th ollowin rsult hols: Thorm 2. I loi untion F is tr-omposl, thn its omplmnt F is not. 2 Bus o Thorm (2), NOT ts n ppr only t th root or th lvs o th TD NOR. Suppos, y ontrition, tht topoloy lik in Fi. (3) wr possil. In this s, w oul mr th NOR t N 2 with N. This 2

woul init tht tht w h not ompos mimlly th untion rprsnt y th NOR N. N2 Fiur 3. An impossil topoloy or NOR omposition trs. Bs on rursiv D NOR, w hv now prtion loi untions into thr lsss:. tr-omposl untions; 2. untions tr-omposl with th us o NOT ts; 3. untions not tr-omposl. In th nt stion, w sri how th notion o tr omposition n omposility is us or otinin hyri rprsnttion styl or ritrry loi untions. This rprsnttion will ontin NOR trs n BDD nos. W onlu this stion with som osrvtions on tromposl untions: Cnoniity n vril orrins. As mntion, or tr-omposl untions, th tr omposition is nonil. Morovr, unlik ROBDDs, th tr rprsnttion is trivilly inpnnt rom th vril orrin. In, vn i untion F is not ntirly tr-omposl, th knowl o prtil omposition inits orrin strtis or th input vrils o F.IF is omposl s, sy, F =( + 2), thn optiml orrins will pl ll th vrils o on top o thos o 2 (or vivrs), n th siz o th ROBDD o F will th sum o thos o n 2. Hn, it ollows in prtiulr tht th ROBDD o ny tr-omposl untion, with n optiml vril orrin, is linr in th numr o inputs. Booln mthin. Booln mthin is n importnt stp o thnoloy mppin [9]. It onsists o inin whthr two untions () n (y) oini tr rplin th input vrils y with prmuttion P o th input vrils. Vritions o th prolm inlu mthin moulo omplmnttion o som inputs n o on o th untions, n it is nm NPN-mthin (Ntion, Prmuttion, Ntion mthin). In nrl, this is iiult prolm, s it ntils numrtin th prmuttions o th input vtor n hkin th quivln o () with (P ). For tromposl untions, lrly mth ists i n only th trs rprsntin n r qul, pt possily or th prsn o input/ output invrtrs. Tr isomorphism n rri out in tim linr in th siz o th tr [2]. Mor nrlly, mthin n ist only i n ompos in shion similr to. Evn prtil omposility o is thus hlpul. N Booln mnipultion routins. ROBBD mnipultion routins r s on rursiv visit o th ROBDD untions. At h rursion, vril is slt n th otors ; r vlut. Rursion is m st us, y thir vry ntur, ROB- DDs llow onstnt-tim otorin. I untion is rprsnt y NOR tr, inst, thn otorin rquirs ssinin th vlu to th tr input n thn proptin th t (i.., simultin) towrs th output. This simultion tks tim proportionl to th tr pth. In th implmnttion stion it will sn how th knowl o omposition, howvr, hlps ompnstin this mor iiult otorin. 3 Multi-Lvl Domposition Dirms In this stion w ploit tr ompositions to riv nw hyri mol or rprsntin loi untions. W will rprsnt untions o th irst n son lss y tr o NOR ts. Funtions in th thir lss will rprsnt throuh th us o Shnnon pnsion with rspt to som vril, lin to BDD nos. W thn pply tr omposition n Shnnon pnsion in orr to h otor n rursivly. Empl 3. Th untion =( + + ) hs ) + ) Fiur 4. ) Tr omposition o th untion in Empl (3). ) Th sm untion with th ition o BDD nos. TD NOR s in Fi. (4.). Not tht in no s w oul urthr ompos + us o th isjoint support onstrint. Applyin Shnnon pnsion, in Fi. (4.) w otin TD NOR or h input o MUX. 2 Th nw strutur w prsnt in this ppr plors tr ompositions o ivn untion. Bus o its purpos w ll it Multi-Lvl Domposition Dirms,. W now in s s on TD NOR. In our rwins o rphs, irls rprsnt MUX vrtis, whil rrys o squrs rprsnt NOR vrtis. Dinition 5. A is irt yli rph, with l vrtis ll y Booln onstnt or vril n two kins o intrnl vrtis: NOR vrtis hv non-mpty st o outoin s, h pointin to. MUX vrtis hv two outoin s, n, n r ll y Booln vril. A ins rursivly loi untion with th ollowin ruls: 3

2 l A trminl vrt t ll y Booln vril or onstnt nots th untion. l A MUX vrt m ll y Booln vril ins th untion: F m = F (m) +F (m) (2) l A NOR vrt n with k outoin s ins th untion: F n = + + k (3) whr i, i = ;:::;k is th untion in y th point y i. In, whil MUX vrtis orrspon to ROBDD nos, NOR vrtis r nw tur o this mol whih mphsizs th tr ompositions o th untion. Just lik ROBDDs, w impos rution ruls n orrin ruls to s in orr to otin mor ompt nonil rprsnttion: l Thr r no two intil surphs in th sm. l Thr r no vrtis with two or mor outoin s pointin t th sm. l W impos totl orrin twn vrils llin intrnl n trminl vrtis o. Eh pth rom root to trminl must trvrs susqunt MUX nos in rspt o this orrin n h vril is vlut t most on on h pth. ) ) Th ollowin rsult is irt onsqun o th noniity o tr ompositions n rution ruls. W thus stt it without proo, or th sk o onisnss: Thorm 3. Ru Orr Dompos s r nonil. 2 Th o untion mths multi-lvl loi iruit in th ovious wy. In Fi. (6.) n (6.) w rport n th orrsponin t-lvl ntwork. Du to this vint orrsponn, hrtr w will ll rph nos inirntly vrtis or ts. Empl 4. Fi. (6.) rprsnts th nonil ) ) Fiur 6. ) Th o th untion in Empl (4). ) A loi iruit viw o th. ) Its ROBDD. or th untion =( + )( + )+ with liorphil orrin o th vrils. In this s two istint surphs shr th NOR t rprsntin. This is not otn th s or simpl untions. For mor ompl untions it is mor likly to hppn. In Fi. (6.) shows th orrsponin ROBDD. Th hs ompos oth otors n until rhin th input vrils. 2. ) 3. Proprtis o s Fiur 5. Son rution rul. ) NOR vrtis. ) Mu vrtis. It is worth notin tht, unlik ROBBDs, th son rution rul rs irnt onsquns on th two kins o intrnl vrtis. As skth in Fi. (5), rution o MUX vrt implis th ltion o th no. This is not th s or NOR vrtis. In ition to ROBDD-lik ruls, in orr to rnt noniity w must impos omposition ruls: l th suuntions point y NOR vrt must hv isjoint support. Non o thm n ompos y D OR ; l untion is rprsnt y MUX i it is not omposl, nor its omplmnt. W onlu th stion y pointin som rsults on D NOR s tht r usul or th onstrution o mnipultion routins. Thorm 4. Suppos ; ; k isdosom untion. Thn, y rsin lmnts rom th st, th nw st is lso D. 2 Thorm 5. I D NOR ( ) = ; ; k [p; ;p h n D NOR () =; ; l [p; ;p h, whr i 6= j ;i =; l; j =; k, thn:. D NOR ( ) =p; ;p h [ D NOR ([( + ::: + k ) ( + ::: + l ) ] ). 2. D NOR ( + ) = p; ;p h [ D NOR ([( + ::: + k ) +( + ::: + l ) ] ) 3. Lt not vril not in th support o or. Thn: 4

2 Λ D NOR ( + ) = p; ;p h [ D NOR ([ ( + ::: + k ) + ( + + l ) ] ) OR (ml op, ml op2) D(op) = D(op) D(op2); 2 D(op) = D(op) n D(op); 3 D(op2) = D(op2) n D(op); Thorm 6. Lt not vril, 62 S(), n suppos = + 4 i (trminl s) rturn (D(op) [. Thn, 5 rs = omp lookup(op, op2); trminl vlu); 6 i (rs!= NULL) rturn (D(op) [ rs); D NOR ( )=[D NOR () 7 (4) = top vr(op, op2); 8 i (top vr(op) == ) 2 9 op l = vltop(op, ); op r = vltop(op, ); ls op l = op; op r = op; 3.. Complmnttion Th o tr-omposl untion is trivilly2 i (top vr(op2) == ) symmtri s NOR tr, possily with NOT t t th root. This llows 3 lt=or (op l, op2 l); riht=or (op r, op2 r); onstnt tim n sp omplmnttion. 4 rs = ml in (lt, riht, ); It is wll known [8] tht or MUX nos, th insrtion 5 omp insrt (op, op2, rs); o NOT ts (i.. omplmnt s), n ris noniity 6 rturn (D(op) [ rs); prolms. To t roun this prolm w us NOT t rution ruls similr to thos o [8]. Ths r pit in Fi. (7). Fiur 8. Psuoo o OR() Fiur 7. Equivlnt s 4 mnipultion routins As w hv sn, this mol hs som o th ROBDD turs. Amon ths, t strutur tht n mnipult throuh rursiv prours. Th t strutur w implmnt rlizs vrtis uniormly with n-tupls, th irst lmnt in n intr, ll th othrs in pointrs to othr s. In th irst lmnt w no th typ o no (i.., MUX or NOR vrt), th numr o lmnts in th n-tupl (or MUX nos it is lwys 2) n th top vril o th untion rprsnt. W mintin th strutur in stron nonil orm, i.., two quivlnt untions r intii y th sm pointr, y th milir hshin mhnism. W hv thn implmnt Booln oprtion routins. As n mpl, Fi. (8) rports th psuo-o or th loi OR o two untions. Rows, 2 n 3 r th pplition o Thorm 5, s 2. W sk ommon lmnts in th oprns n rmov thm rom th rursiv oprtion. This rmovl n rsult in str ution us w hv simplr oprns. D(op) inits th st o lmnts o th omposition o op. In NOR vrt op, it is th st o ll outoin pointrs (n init st oprtion o irn). Th situtions or whih op is MUX is spil s. For uniorm mnmnt o th strutur n untions rprsnt, w init s D(op) o MUX vrt op, pointr to th omplmnt o th untion root t op. W lso mintin omput tl, lik tht o stnr ROBDD prours, whr w stor prtil rsults. Th rmovl o ommon suuntions lso hlps voiin th ovrill o this tl us w n ploit th nri sinl ntry o th tl F + G = H or rtrivin rsults o vry oprtion (F + ) +(G + ) =(H + ) whn vris, whih onsquntly ns not stor. I th srh in th omput tl ils, w strt rursion. First o ll w in th top vril o th oprns, whih is immit u to its noin in th irst lmnt o th t strutur. Prour vltop(, vlu) rturns th o untion =vlu ssumin is th top vr. o. This stp orrspons to tkin otors in ROBDDs. Atr rursion, ml in() rts rom top vr. n its otors. W now nlyz in mor til ths thr stps, nmly, trminl ss, otorin, n rtion. Trminl ss n vlus pn on th oprtion w r pplyin. For th Booln OR w roniz th ollowin situtions: trminl s rturn vlu op=, op2= op=, op2= op2, op op = op2 op 9; 2 DSD(op ); 2 DSD(op2 ) Prour vltop() is rsponsil or otorin. Its psuo-o is rport in Fi. (9, n Fi. () shows its oprtion. vltop() rursivly os own th tr omposition until it rhs th MUX no ll with th top vril o th, n it tks its otor. In Fi. (.), this is th MUX ll y. Rturnin up rom rursion, it sustituts NOR vrtis with nwly nrt ons, whil mintinin noniity (th sh ts o Fi. (.)). 5

Λ Λ vltop (ml op, ooln vlu) i (op is MUX no) rturn (op.vlu); Fiur. An mpl o vltop() pplition ml in(ml lt, ml riht, top vr ) i (lt == riht) rturn (lt); 2 i = lmnt o op suh tht op.topvr = op.i.topvr; 2 i (riht == ) 3 opr = vltop(op.i, vlu); 3 nw vrt = in or rt(,, ); 4 D(op) = D(op) n op.i; 4 D(rs) = nw vrt [ riht; 5 D(rs) = D(opr) [ D(op); 5 rturn(rs); 6 rturn (rs); 6 i (riht == ) similrly 7 i (lt == or lt == ) symmtri s Fiur 9. Psuoo o vltop() 8 i (lt ρ D(riht)) 9 D(riht) = D(riht) n lt; nw vrt = ml in(, riht, ); [ D(rs) = nw vrt lt; 2 rturn(rs); 3 i (riht ρ D(lt)) symmtri s 4 D(op) = D(lt) D(riht); 5 i ( D(op) = ; ) rturn( in or rt(lt, riht, ) ); 6 D(lt) = D(lt) n D(op); 7 D(riht) = D(riht) n D(op); 8 nw vrt = ml in(lt, riht, ); 9 D(rs) = D(op) [ nw vrt; 2 rturn(rs); Fiur. Psuoo o ml in() Th o o vltop() works s ollows: Lin hks or n-o-rursion-s, i.. rhin o MUX no rom whih w n tk th rqust otor. Othrwis w hv to in th ritil lmnt in our NOR vrt list to us or oin own on lvl. Lin 3 mks th rursiv ll with this ritil lmnt. Atr rursion w sustitut th ritil lmnt in th list with th rturn rph. For mpl i th ritil lmnt ws MUX vrt w sustitut it with its otor. Whil oin this work w my hv to mr list n/or hk or spil ss (or mpl i th rturn rph is th onstnt, w simply rturn th onstnt ) n mintin noniity (rution ruls). Prour ml in() is skth in Fi. (). It uils tryin to isovr vry possil ommon trm rom th two otors. First o ll, it hks or simpl ss (rows 2 to 8). Thy r pplition o Thorm 6. For mpl, rows 2 to 5 min th sitution or riht =, i.., th untion to nrt is = lt. With NOR suh untion is ivn y =( + l + + l n ) (l i r th omponnts o lt). W hv rprsnt ths trminl ss in Fi. (2). in or rt() provis th rtion or rtrivl o MUX or trminl vrt whil kpin up to t uniqu tl similr to tht o ROBDD. In rows 8-3 w hk or on o th two nrl ss, whr non o th otors is onstnt. I th omplmnt o on otor is ontin in th othr s uniqu lmnt, thn thr is tr omposition. r + (r + r2 + + r n ) whr r;r2;:::;r n r th omponnts o th riht. This is quivlnt to: r +[ + (r2 + + r n ) ] W hv rport this s in Fi. (3.). Lins 4-2 l with th othr nrl s. Hr w hv to srh or ommon lmnts twn lt n riht n to tor thm out. This pplis s 3 o Thorm 5. Ths stps r skth in Fi. (3.). As mntion, vltop() n ml in() rpl otorin n th si in or rt() oprtions in ROBDDs. Whil oprtions r trivil onstnttim in ROBDDs, thy my tk O() tim in s, whr nots th tr pth. To this rr, w osrv tht is oun y th numr o vrils n it is rthr smll in prti (lwys 3 or lss or th synthsis nhmrks). Morovr, s OR is ppli to pirs o nos own in th rph, th support st o suuntions will hv wr lmnts n so th numr o lls to vltop(). Empl 5. W hv rport in Fi. (4) miml pth tr omposl untion = ((( + 2)3 + 4)5 + :::. 2 5 s vrsus ROBDDs In this stion w prsnt som omprisons in rprsntin untions with s n ROBDDs. 6

) 5 4 3 ) 2 Fiur 4. A miml tr omposl untion Fiur 2. Intiition o D urin trvrsl - trminl ss 2 3 4 2n- 2n Fiur 5. Th PAD or th untions F n 5.2 Tsts on nhmrk iruits ml H ml ml H ml H ml I ml J ) ) ml H Fiur 3. Intiition o D urin trvrsl - nrl ss 5. Eponntil rowth In this sustion w ontrst s with ROBDDs with rspt to prtiulr lss o orr-snsitiv untions, nmly, th untions: F n =( + 2)(3 + 4) (2n + 2n) (5) It is wll known tht with n impropr orrin o th vrils (or mpl, plin th o-ll vrils up top) rsults in ROBDD or F n with ovr 2 n nos []. Morovr, in spit o th simpliity o th untion, most vril orrins or F n n prov. Th or th untion is shown in Fi. (5). It onsists o two-lvl NOR iruit, rrlss o th orr hosn or th vrils i n it is lwys linr. Empl 6. Consir th untion =(A + B) + ( + AB), with n orrin o vrils plin on top. Sin = 6= =, ny ROBDD hs th spt shown in Fi. (6.). In nrl, w my think o s whr th two otors look lik untion n o Eq. (5), ut with irnt omintion o prouts. Any orrin o ; A; ; B whih optimizs on rnh is oun to suoptiml or th othr rnh o th ROBDD. Fi. (6.) illustrts th or th sm untion. Both rnhs r utomtilly ompos optimlly. 2 ml I ml J W hv ompr our nw mol with ROBDDs in numr o nhmrk iruits in trms o mmory ouption n CPU tim n to uil th output untion rphs. Th nhmrks r ivi in thr stions: multilvl iruits, two-lvl n thir stion tstin th omintionl prt o synhronous iruits. All ths nhmrks om rom th IWLS9 nhmrk suit [2]. Th vril orrin hosn or ths iruits ws otin y pplyin th Brkly orrin [3]. No vril rorrin took pl, howvr, urin th ution o ny pk. W hv implmnt our mol n tst it inst th Crni-Mllon ROBDD pk. W rri out omprisons on th tul mmory ouption. W ssum r-on implmnttions, in whih in prtiulr h ROBDD no tks thr mhin wors. Morovr ROB- DDs hv omplmnt s. With rrs to vrtis, w ssum n implmnttion whr h no on- ) A B A A ) A B A B Fiur 6. ) ROBDD strutur or th untion o Empl (9). ) strutur or th sm untion. 7

sists o n rry. As mntion, th irst lmnt stors inormtions out th no, whil othr lmnts r pointrs. This mol lso implmnts NOT ts throuh omplmnt s. CPU-tim ws tkn on HP Vtr 5/33 with 48Myts o RAM. From Tl, s turn out to mor ompt on vr o 8%. Som nhmrks iv prtiulrly oo rsults, or mpl omp n pir, nhmrks whih TD NOR is vry tiv in omposin output untions until rhin input vrils or vry simpl untions. Th CPU tim is lwys ttr or ROBDDs. Empirilly w hv oun th ollowin thr rsons: l W mk intrnl rursions in th onstrutions o s (vltop() n ml in()). Thus th numr o lls to ky prours or h omputtion is hihr. l W hv to mn rrys tht in nrl hv mor lmnts thn ROBDDs. For mpl, hsh untions r mor ompl n lso storin n rtrivin rom omput tl n uniqu tl ns mor tim. l Th strutur w us llows multipl pths rom rtin no (NOR). On th othr hn, with ROBDD th pth is uniqu. This is similr to simultion throuh NFA opposit to DFA. W hv lso implmnt ynmi rorrin in our mol with sitin-s lorithm [2]. Ovr ROBDDs, w hv th vnt to know mor out oo vril orr irtly rom th t strutur. In tl 2 w mk omprisons usin or h nhmrk th orr ivn y sitin (intrstinly, it is irnt or th two mols). Vril orrin took pl only t th n o ution. Rsults show tht, tr sitin, s improv slihtly urthr ovr ROBDDs. W think this is us urin sitin w ploit our ttr knowl o th untion s strutur n n voi to o throuh orrins tht iv smll vnt ut lok urthr improvmnts. 6 Conlusions n utur work s hv prov thmslvs iint in mkin pliit th Ds o loi untions. This proprty llows us to rh mor ompt, lil n roust rph-s rprsnttion. Morovr, this rprsnttion is losly rlt to multipl lvl iruit, n is mor inormtiv on th rol o th support vrils o untion. W pt ths proprtis to usul in ivrs pplitions, most notly thnoloy mppin or omintionl iruits n Booln mthin /rhility nlysis or vriition / ATPG in squntil iruits. 7 Appni Bnhmrk ROBDD RATIOS nos mm nos mm nos mm MultiLvl lu2 25 65 26 59 62.7% 8.5% lu4 685 255 5 77 34.% 6.% p6 7 353 93 3377 29.7% 4.% p7 555 665 23 979 4.3% 7.% 9 8 543 75 452 4.3% 2.% C355 45922 37766 4456 523 4.% -8.3% C432 378 93534 647 82676 93.% 3.% C499 45922 37766 4456 523 4.% -8.3% 8 33 399 96 388 38.5% 2.8% C88284 38523 973 3476 4.% 22.4% ht 5 45 86 42 74.4% 6.9% CM5 5 533 285 66 79.3% 43.8% CM52 383 49 284 6 34.9% 8.4% omp 5476 6428 434 459 6.8% 26.% ount 24 62 87 73 9.% -2.9% DES 358 94524 2885 966.8% 4.3% mpl2 869 267 223 362 289.7% 9.4% r 24 62 53 458 284.9% 33.6% r2 374 42 349 472 7.9% 6.4% k2 28336 858 27437 8634 3.3% -.5% ll 38 44 63 284 9.% 45.8% Ars 457 37 456 372.2% -.% pir 428 23384 853 2664 4.7% 363.% plr8 44 432 98 392 46.9%.2% rot 2537 376 7796 27463 6.8% 37.% st 8 354 49 239 4.8% 48.% trm 638 94 54 54 34.3% 254.4% too lr 796 2288 4876 853 45.5% 7.3% ttt2 25 65 5 565 78.3% 8.8% v 4345 335 423 3235 3.4% -.5% 297 389 223 67 48.6% 32.9% 4 683 249 477 825 43.2% 2.3% TwoLvl lu4.pl 97 359 8 3294 49.4% 9.% p5.pl 2679 837 88 5259 46.2% 52.8% lip.pl 226 678 48 664 52.7% 2.% 64.pl 44 4323 66 244 283.3% 79.8% mis2.pl 37 4 34 294 32.9% 39.8% mis3.pl 3 393 84 3929 59.8% -.7% so2.pl 55 465 48 39 222.9% 45.8% v2.pl 44 332 52 2429.8% 28.9% F.S.M. 769 237 8 785 55.7% 29.2% 2 375 25 44 729 752.3% 54.3% 3 29 387 27 37 377.8% 22.% 4 248 744 39 497 535.9% 49.7% 5 9 357 23 25 47.4% 42.2% 7 44 432 28 38 44.3% 4.3% s96 3387 6 226 9523 52.8% 6.7% s238 387 926 28 8998 53.% 2.9% s423 278 3824 53 336 25.2% 5.% s344 68 54 97 454 73.2%.% s4252 456 76 34.% 34.% s526 89 567 98 482 92.9% 7.6% s73 93 279 228 48 296.% 83.% s838 3 948 668 2.7% 34.7% s953 474 422 2 32 35.8% 9.2% Tl. ROBDD vs. in siz n prormn with Brkly orrin Proo o Thorm. Th proo o th irst ssrtion ollows y ontrition: W ssum th istn o two istint D NOR, nmly, ; ; p n ; ; q, n show tht this ls 8

Bnhmrk ROBDD RATIOS nos mm nos mm nos mm MultiLvl lu2 64 492 8 474 39.% 3.8% lu4 429 287 349 25 22.9% 2.9% p6 669 27 537 247 24.6% -6.5% p7 484457 76 25.7% 89.5% 9 65 495 58 388 84.5% 27.6% C355 346 938 343 95622.4% -4.4% C432 3 39 2478 54-47.5% -64.7% C499 346 938 343 95622.4% -4.4% 8 3 64 238 56.2% 26.% C88 6969 297 4294 4392 62.3% 45.3% ht 9 273 9 376.% -27.4% CM5 7 5 7 52.% -.9% CM52 9 57 6 46 8.8% 23.9% omp 52 456 66 38 3.3% 9.7% ount 82 246 87 339-5.7% -27.4% DES 955 28545 858 28898 8.% -.2% mpl2 646 938 83 6 253.% 82.7% r 86 558 4 337 353.7% 65.6% r2 957 587 79 389 47.7% 54.% k2 426 4278 67 3796 3.% 2.7% ll 94 282 42 232 23.8% 2.6% Ars 52 56 456 372.% 9.8% pir 632 896 4283 5485 4.8% 6.9% plr8 3 39 64 324 3.% 2.4% rot 769 227 378 485 9.6% 49.5% st 53 59 33 87 6.6% -5.% trm 7 32 53 283.9% 3.4% too lr 3 3339 578 2529 92.6% 32.% ttt2 58 474 6 38 63.3% 24.4% v 534 62 289 648 84.8% -2.8% 682 246 22 4 459.% 3.8% 4 682 246 26 28 25.7% 8.4% TwoLvl lu4.pl 79 237 55 2675 53.4% -.4% p5.pl 44 432 88 3935 63.5% 9.8% lip.pl 7 5 66 276 57.6% 84.8% 64.pl 732 296 66 244 9.% -8.7% mis2.pl 333 32 286 246.9% 6.4% mis3.pl 76 22885 7 3.8% 94.9% so2.pl 33 399 45 289 95.6% 38.% v2.pl 44 242 59 43 6.7% 28.2% FSM 639 97 4 568 54.4% 22.3% 2 37 44 7 74.9% 56.% 3 29 387 27 36 377.8% 22.5% 4 239 77 36 477 563.9% 5.3% 5 8 324 23 249 369.6% 3.% 7 28 384 28 3 357.% 27.6% s96 86 248 38 862 53.5% 29.9% s238 83 2439 3 82 6.4% 34.6% s344 64 492 8 395 5.% 24.6% s4279 537 75 337 38.7% 59.3% s526 47 44 69 387 3.% 4.% s73 747 224 73 33.8%.9% s838 293 879 47 665 99.3% 32.2% s953 498 494 58 8 25.2% 38.2% Tl 2. ROBDD vs. in siz n prormn tr ynmi rorrin nssrily to th violtion o som proprtis o th untions i or i. It is not rstritiv to ssum tht th two sts i; i ir us 6= i;i =; ;p:sin i; i r oth ompositions o, it must : + + p = + + q (6) or quivlntly, + + p = + + q : (7) Sin ll untions i hv isjoint support, it is possil to in n ssinmnt o th vrils in S( 2);S( 3); ;S( q) suh tht i =;i =2; ;q. Noti tht th vrils in S( ) hv not n ssin ny vlu. Corrsponin to this prtil ssinmnt, Eq. (7) oms: Λ + + Λ p = (8) In Eq. (8), Λ i nots th rsiu untion otin rom i with th ormntion prtil ssinmnt. W n now istinuish svrl ss, pnin on th ssumptions on th strutur o th lt-hn si o Eq. (8). Cs ). Th lt-hn si rus to onstnt. Hn, is onstnt, inst th ssumptions. Cs 2). Th lt-hn si ontins two or mor trms. Sin ths trms must hv isjoint support, is urthr omposl, inst th ssumptions. Cs 3). Th lt-hn si rus to sinl trm. It is not rstritiv to ssum this trm to Λ. I = Λ, thn w hv =, inst th ssumption tht irs rom ny i. Hn, it must Λ 6=, n S( )=S( Λ ) ρ S( ) stritly: (9) W now show tht lso this s ls to ontrition. Consir son ssinmnt, zroin ll untions j;j 6=. Eq. (7) now rus to = Λ + + Λ q : () By th sm rsonins rri out so r, th r.h.s. o Eq. () n ontin only on trm. W now show tht this trm must. Λ I, y ontrition, = Λ j ;j 6=, thn y Eq. (9) on woul hv S( )=S( Λ j ) S( ) () inst th ssumption o i; j in isjoint-support. Hn, it must =. Λ In this s, y rsonins similr to thos lin to Eq. (9), w t S( )=S( Λ ) ρ S( ) stritly (2) whih ontrits Eq. (9). Hn, nnot ir rom ny i, n th irst point is prov. Th proo o th son sttmnt ollows y pplyin rursivly D NOR to h o i. Sin h D is uniqu, th tr omposition is lso uniqu n th Thorm is prov. 2 Proo o Thorm 2. By ontrition. Suppos w hv untion F tht is omposl s F =( + 2) with S( ) S( 2)=; n suh tht F is lso omposl s F =( + 2) with S( ) S( 2)=;. W hv to prov ontrition in th quivln: ( + 2) = + 2 (3) For sk o rility, w in =, = 2, =, = 2 n ontrit: = + : (4) 9

W prtition th supports o ths untions in this wy: S = S() S() S = S() S() S = S() S() S = S() S() Som o th S ij n mpty. In th rst o th proo w show tht Eq. (4) implis tht th support o t lst on o,,, is mpty, inst th ssumptions. To this n, w will rwrit Eq. (4) unr irnt prtil ssinmnts o th vrils in S ij. For instn, y sltin n ssinmnt o S() suh tht =, w otin: = + (5) whr inits th untion otin y ssinin in th vrils o S with vlus stisyin =. Th support o is thn S. Similrly, w n hoos nothr ssinmnt in S() so tht =n otin: = + : (6) From Eqs. (5) n (6), w hv: + = =( + )( + ) : (7) W now in prssions or n. W vlut to zro, ruin th ov qution to: =( + )( + ) : (8) is otin y ssinin irst th vrils in S n thn thos in S. Du to th omplt ssinmnt, is onstnt (not nssrily ). Similrly or. So, in ruin th lst qution, w our ss:. = =. Thn =, i.. its support st is mpty inst th ssumptions. 2. =n =. Thn =. 3. =n =. Thn =. 4. = =. Thn =. Rptin th sm prour to Eq. (7) with th vlution o to, w hv th symmtri ss:. = =. Thn =. 2. =n =. Thn =. 3. =n =. Thn =. 4. = =. Thn =. Now w hv to prov th ontrition usin Eq. (7) or ll th possil omintions o ths ss.. = n =. + =( + )( + )= + (9) Th ontrition oms vint i, or mpl, w ssin =, = n =, whih ls to =, i.. S = ;. A son ssinmnt, =, =n =, ls to =, so tht lso S = ;. Thus S = S [ S = ;; woul hv to onstnt, ontrition. 2. = n =. + =( + )( + )= + (2) Sin = w know tht S = ; n isor. W onsir oth ss. I =th qution ov rus to: ( + )= + n vlutin =n =w in =, hn S = ;, n thror S() =S [ S = ;. I, inst, =w hv: + = + Assinin =n =w in =, i.. S = ;, n thn S() =S [ S = ;, inst th ssumptions. 3. = n =. Thn, S = ;, S = ; n S() = S [ S = ;, i.. is onstnt. 4. = n =. + =( + )( + )= + (2) n lso S = ; n S = ;. Sin n r onstnts, w onsir two ss: =. Thn + = + n vlutin =w in tht is onstnt, so tht S() =S [ S = ;, inst th ssumptions. I, inst, =,whv ( + )= + : Evlutin = w in =. Thn S = ; n S() =S [ S = ;. All othr situtions r rsolv y pplyin th sm rsonin s in lst ss. 2 Proo o Thorm 4. Consir rmovin sinl lmnt, sy,, rom th st. Th nw st, 2; ; k, is still omposition. It is lso miml, or i ny trm wr urthr omposl, thn th sm trm woul omposl in ; ; k, n ; ; k woul not D. 2 Proo o Thorm 5. W prov only th thir rsult, th othr ss in onptully similr. Clrly, th riht-hn si o th thir qution is NOR omposition. Thror, th only issu is its mimlity. Non o p ; ;p h n urthr ompos, or ls w woul ontrit th ssumption tht p ; p h ppr in, sy, D NOR ( ). Th only nit or urthr omposition is thn p h+ =[ ( + ::: + k ) + ( + + l ) ]. Suppos, y ontrition, tht p h+ hs D OR z ; ;z q with mor thn on lmnt. In this s, pprs in th support o only on untion z j, sy, z q. Hn, = ( + ) = = = (p + + p h + z + + z q + z q;=) = ( + ) = = = (p + + p h + z + + z q + z q;=) Sin th trms z ; ;z q ppr in n, thy nnot oini with ny o i; j. But thn n woul hv two istint D NOR s, lry prov impossil. 2 Proo o Thorm 6. Th riht-hn si is isjoint-support omposition. Its mimlity ollows rom th impossiility o rkin own or ny trm in D NOR () into sum o othr trms. 2 Rrns [] R. E. Brynt. Grph-s lorithms or ooln untion mnipultion. IEEE Trns. on Computrs, 35(8):677 69, Auust 986. [2] O. Court n J.C. Mr. A unii rmwork or th orml vriition o squntil iruits. In Pro. ICCAD, ps 26 29, Novmr 99.

[3] S. Mlik, A. R. Wn, R. K. Bryton, n A. Sniovnni- Vinntlli. Loi vriition usin inry ision irms in loi synthsis nvironmnt. In Pro. ICCAD, ps 6 9, Novmr 988. [4] Y. Mtsun n M. Fujit. Multi-lvl loi optimiztion usin inry ision irms. In Pro. ICCAD, ps 556 559, Novmr 989. [5] H. Touti, H. Svoj, B. Lin, R.K. Bryton, n A. Sniovnni-Vinntlli. Impliit stt numrtion o init stt mhins usin BDD s. In Pro. ICCAD, ps 3 33, Novmr 99. [6] F. Corno, P. Printto, n M. Sonz Ror. Usin symoli thniqus to in th mimum liqu in vry lr sprs rphs. In Pro. EDAC, ps 32 324, Mrh 995. [7] Y-T. Li n S. Sstry. E-vlu inry ision irms or multi-lvl hirrhil vriition. In Pro. DAC, ps 24 243, Jun 992. [8] K. S. Br, R. L. Rull, n R. E. Brynt. Eiint implmnttion o BDD pk. In Pro. DAC, ps 4 45, Jun 99. [9] H. Ohi, K. Ysuok, n S. Yjim. Brth-irst mnipultion o vry lr inry ision irms. In Pro. DAC, ps 48 55, Novmr 993. [] K. M. Butlr, D. E. Ross, R. Kpur, n M. R. Mrr. Huristis to omput vril orrins or th iint mnipultion o inry ision irms. In Pro. DAC, ps 47 42, Jun 99. [] S. J. Frimn n K. J. Supowit. Finin th optiml vril orrin or inry ision irms. IEEE Trns. on Computrs, 39:7 73, 99. [2] R. Rull. Dynmi vril orrin or orr inry ision irms. In Pro. ICCAD, ps 42 47, Novmr 993. [3] U. Kshull, E. Shurt, n W. Rosntil. Multilvl loi s on untionl ision irms. In EuroDAC, Proins o th Europn Conrn on Dsin Automtion, ps 43 47, Sptmr 992. [4] R. Drhslr, A. Sri, M. Thol, B. Bkr, n M. A. Prkowski. Eiint rprsnttion n mnipultion o o swithin untions s on orr kronkr untionl ision irms. In Pro. DAC, ps 45 49, Jun 994. [5] S.-I. Minto. Zro-supprss s or st mnipultion in omintoril prolms. In Pro. DAC, ps 272 277, Jun 993. [6] Y.-T.Li, M. Prm, n S. B. K. Vruhul. Ev-s lorithms or ilp, sptrl trnsorm n untion omposition. IEEE Trns. on CAD/ICAS, 3(8):959 975, Auust 994. [7] R. E. Brynt. Binry ision irms n yon: nlin thnolois or orml vriition. In Pro. ICCAD, ps 236 243, 995. [8] V. Brto n M. Dmini. Booln untion rprsnttion usin prlll-ss irms. In Sith Grt Lks Symposium on VLSI, Mrh 996. [9] F. Milhot n G. DMihli. Alorithms or thnoloy mppin s on inry ision irms n on ooln oprtions. IEEE Trns. on CAD/ICAS, ps 599 62, My 993. [2] A. Aho, J. E. Hoprot, n J. D. Ullmn. Th sin n nlysis o omputr lorithms. Aison Wsly, 974. [2] S. Yn. Loi synthsis n optimiztion nhmrk usr ui, vrsion 3.. MCNC, Jnury 99.