Presentation is loading. Please wait.

Presentation is loading. Please wait.

Escola Politécnica da Universidade de São Paulo GSEIS - LME Logic Synthesis in IC Design and Associated Tools Sequential Synthesis Wang Jiang Chau Grupo.

Similar presentations


Presentation on theme: "Escola Politécnica da Universidade de São Paulo GSEIS - LME Logic Synthesis in IC Design and Associated Tools Sequential Synthesis Wang Jiang Chau Grupo."— Presentation transcript:

1 Escola Politécnica da Universidade de São Paulo GSEIS - LME Logic Synthesis in IC Design and Associated Tools Sequential Synthesis Wang Jiang Chau Grupo de Projeto de Sistemas Eletrônicos e Software Aplicado Laboratório de Microeletrônica – LME Depto. Sistemas Eletrônicos Universidade de São Paulo

2 Escola Politécnica da Universidade de São Paulo GSEIS - LME Defined by the quintuple (,, S,, ). A set of primary inputs patterns. A set of primary outputs patterns. A set of states S. A state transition function : S S. An output function : S for Mealy models : S for Moore models. Defined by the quintuple (,, S,, ). A set of primary inputs patterns. A set of primary outputs patterns. A set of states S. A state transition function : S S. An output function : S for Mealy models : S for Moore models. Finite-State Machine Model

3 Escola Politécnica da Universidade de São Paulo GSEIS - LME Sequential Logic Synthesis Finite-State Machine F(X,Y,Z,, ) as State Transition Graph (STG): Circuit composed of : Combinational logic for and Set of registers (flip-flops) D D XY State Minimization State Encoding

4 Escola Politécnica da Universidade de São Paulo GSEIS - LME State Minimization Aims at reducing the number of machine states reduces the size of transition table. State reduction may reduce the number of storage elements. the combinational logic due to reduction in transitions Completely specified finite-state machines No don't care conditions. Easy to solve. Incompletely specified finite-state machines Unspecified transitions and/or outputs. Intractable problem.

5 Escola Politécnica da Universidade de São Paulo GSEIS - LME State Minimization for Completely-Specified FSMs Equivalent states Given any input sequence the corresponding output sequences match. Theorem: Two states are equivalent iff they lead to identical outputs and their next-states are equivalent. Equivalence is transitive Partition states into equivalence classes. Minimum finite-state machine is unique.

6 Escola Politécnica da Universidade de São Paulo GSEIS - LME Applicable input sequences All transitions are specified. Does not lead to any unspecified transition. Compatible states Given any applicable input sequence the corresponding output sequences match. Theorem: Two states are compatible iff they lead to identical outputs (when both are specified), their next-states are compatible (when both are specified). Compatibility is not an equivalency relation (not transitive). Applicable input sequences All transitions are specified. Does not lead to any unspecified transition. Compatible states Given any applicable input sequence the corresponding output sequences match. Theorem: Two states are compatible iff they lead to identical outputs (when both are specified), their next-states are compatible (when both are specified). Compatibility is not an equivalency relation (not transitive). State Minimization for Incompletely-Specified FSMs

7 Escola Politécnica da Universidade de São Paulo GSEIS - LME State Encoding of Finite-State Machines- 1 aka state assignment Given a (minimum) state table of a finite-state machine Find a consistent encoding of the states with one unique code to each state that preserves the cover minimality with minimum number of bits. with minimum delay (hard- usually neglected) The state set must be encoded while satisfying simultaneously both input and output constraints. Given s states, the number of coding bits is log 2 s b s The minimality depends on the combinational logic implementation (2-level? Multi-level?)

8 Escola Politécnica da Universidade de São Paulo GSEIS - LME State Encoding Aimed to 2-Level Traditional PLA solution, checking dependencies (outputs and next state signas) w.r.t. inputs and present state signals Maximize number of common cubes betwwen covers. State Transition Table (STT) PS NS x=0 x=1 z x=0 x=1 AA D0 1 BA C0 CC B0 DC A0 1 y1y2y1y2 Y 1 Y 2 x=0 x=1 z x=0 x=1 A0000 100 1 B0100 110 C1111 010 D1011 000 1 y1y2y1y2 Y 1 Y 2 x=0 x=1 z x=0 x=1 A0000 110 1 B0100 100 C1010 010 D1110 000 1 Y 1 =x´ y 1+ x y 1 ´ Y 2 =x´ y 1+ x y 2 z = x y 2 ´ Y 1 =x´ y 1+ x y 1 ´ Y 2 = x y 2 ´ z = x y 1 ´ y 2 ´ + x y 1 y 2 Transition/Output Table (TOT)

9 Escola Politécnica da Universidade de São Paulo GSEIS - LME Mustang Algorithm State encoding for multilvel logic Developed in UC Berkeley upon MIS Motivation: traditional state encoding for PLAs brings sub-optimal results when applied to multilevel logic STG ou STT Transition/Output Table Two-Level Format Multilevel Format Encoder for PLAs Mustang Sub-optimal Optimal Sub-optimal

10 Escola Politécnica da Universidade de São Paulo GSEIS - LME Traditional Encoding for PLA 6 cubes 16 gates 13 intermediate literals

11 Escola Politécnica da Universidade de São Paulo GSEIS - LME Traditional Encoding for Multilevel 7 cubes 15 gates 12 intermediate literals

12 Escola Politécnica da Universidade de São Paulo GSEIS - LME Mustang State Encoding Logic network representation. Area: # of literals. Encoding based on cube-extraction heuristics Rationale When two (or more) states have a transition to the same next-state Keep the distance of their encoding short. Extract a large common cube. When two (or more) states have a transition from the same next-state Keep the distance of their encoding short (wishing common 1´s). Extract a large common cube.

13 Escola Politécnica da Universidade de São Paulo GSEIS - LME Example- 1 2-input, 5-state FSM (3-bits). s1 s3 with input i 1 i 2 s2 s3 with input i 1 i 2. Encoding s1 000 = abc. s2 001 = abc s3 110 the common-cube refers to A and B. Transition A = B = i 1 i 2 abc + i 1 i 2 abc = ab (i 1 i 2 c+ i 1 i 2 c) 8 literals instead of 10. For Nb: number of bits Nd: Hamming distance (Nb-Nd)=3-1=2 is the number of common literals

14 Escola Politécnica da Universidade de São Paulo GSEIS - LME 2-input, 5-state FSM (3-bits). s3 s1 with input i 1 i 2. s3 s2 with input i 1 i 2. Encoding s1 011 s2 001 s3 110 = abc Transition C= i 1 i 2 abc + i 1 i 2 abc = abc (i 1 i 2 + i 1 i 2 ) 7 literals instead of 10. Example- 2 the common-cube refers to C For Nb: number of bits Nd: Hamming distance (Nb-Nd)=3-1=2 is the number of terms with common cubes

15 Escola Politécnica da Universidade de São Paulo GSEIS - LME Mustang Algorithm Approach Constructin Graph (V, Em, W(Em)), where V- set of vertices. Each vertex represent a state Em- set of arcs connecting vertices (states). Each connection represent the relationship between two states and their coding W(Em) –weigths of arcs. The weigth is the gain associated to each two states, assuming their enconding is close (with low Hamming distance) The weight carries different meanings depending if the proximity in encoding refers to the present states (example 1) or next states (example 2)

16 Escola Politécnica da Universidade de São Paulo GSEIS - LME Mustang Algorithm Implementation- 1 Fanout-oriented algorithm –Consider state fanout i.e. next states and outputs. –Assign closer codes to pair of states that have same next state transition or same output. –Maximize the size of the most frequent common cubes. Weight Em between to states k and l given as M P k,l Average(divide by 2): we do not know how many 1s ´there exist in the next state

17 Escola Politécnica da Universidade de São Paulo GSEIS - LME Example- 1 For the only output (i=1) State s=0 to 3 (st0 to st3): P O 0,1 =0; P O 1,1 =2; P O 2,1 =3; P O 3,1 =2 For the next states s=0 to 3 (st0 to st3): Next state s0; present state s=0 to 3 (st0 to st3): P s 0,s0 =2; P s 1,s0 =1; P s 2,s0 =0; P s 3,s0 =0 Next state s1; present state s=0 to 3 (st0 to st3): P s 0,s1 =1; P s 1,s1 =1; P s 2,s1 =1; P s 3,s1 =0 Next state s2; present state s=0 to 3 (st0 to st3): P s 0,s2 =0; P s 1,s2 =1; P s 2,s2 =1; P s 3,s2 =1 Next state s3; present state s=0 to 3 (st0 to st3): P s 0,s3 =0; P s 1,s3 =0; P s 2,s3 =1; P s 3,s3 =1

18 Escola Politécnica da Universidade de São Paulo GSEIS - LME Example- 2

19 Escola Politécnica da Universidade de São Paulo GSEIS - LME Mustang Algorithm Implementation- 2 Fanin-oriented algorithm –Consider state fan-in i.e. present states and inputs. –Assign closer codes to pair of states that have transition from same present state or same input. –Maximize the frequency of the largest common cubes.

20 Escola Politécnica da Universidade de São Paulo GSEIS - LME Example- 1 For the inputs (i=1 to 2) Input i 1 ; next states=0 to 3 (st0 to st3): P ON 0,1 =2; P ON 1,1 =0; P ON 2,1 =3; P ON 3,1 =0 P OFF 0,1 =0; P OFF 1,1 =3; P OFF 2,1 =0; P OFF 3,1 =2 Input i 2 ; next states=0 to 3 (st0 to st3): P ON 0,2 =2; P ON 1,2 =1; P ON 2,2 =1; P ON 3,2 =1 P OFF 0,2 =1; P OFF 1,2 =1; P OFF 2,2 =1; P OFF 3,2 =0 For the next states s=0 to 3 (st0 to st3): Present state s0; next state s=0 to 3 (st0 to st3): P s 0,s0 =2; P s 1,s0 =1; P s 2,s0 =0; P s 3,s0 =0 Present state s1; next state s=0 to 3 (st0 to st3): P s 0,s1 =1; P s 1,s1 =1; P s 2,s1 =1; P s 3,s1 =0 Present state s2; next t state s=0 to 3 (st0 to st3): P s 0,s2 =0; P s 1,s2 =1; P s 2,s2 =1; P s 3,s2 =1 Present state s3; next state s=0 to 3 (st0 to st3): P s 0,s3 =0; P s 1,s3 =0; P s 2,s3 =1; P s 3,s3 =1

21 Escola Politécnica da Universidade de São Paulo GSEIS - LME Example- 2

22 Escola Politécnica da Universidade de São Paulo GSEIS - LME Algorithm- Graph Embedding Examine all state pairs –Complete graph with |V| = |S|. Add weight on edges –Model desired code proximity. –The higher the weight the lower the distance. Embed graph in the Boolean space. –Objective is to minimize the cost function Difficulties –The number of occurrences of common factors depends on the next-state encoding. –The extraction of common cubes interact with each other.

23 Escola Politécnica da Universidade de São Paulo GSEIS - LME Algorithm- simplified Choose one node v i such that is maximu where v j are adjacent nodes to v i. Code v i and and assign shortest distance codes to v j. Repeat until all graph is assigned.

24 Escola Politécnica da Universidade de São Paulo GSEIS - LME Example For 5 states, N b =3

25 Escola Politécnica da Universidade de São Paulo GSEIS - LME Results


Download ppt "Escola Politécnica da Universidade de São Paulo GSEIS - LME Logic Synthesis in IC Design and Associated Tools Sequential Synthesis Wang Jiang Chau Grupo."

Similar presentations


Ads by Google