lecture 03 – background + intro ai + dataflow and connection to ai eran yahav 1

PROGRAM ANALYSIS & SYNTHESIS

Lecture 03 – Background + intro AI + dataflow and connection to AI

Eran Yahav

2

Previously…

structural operational semantics trace semantics

meaning of a program as the set of its traces

abstract interpretation abstractions of the trace semantics

Example: Cormac’s PLDI’08

Example: Kuperstein PLDI’11

Today

abstract interpretation some basic math background

lattices functions

Galois connection A little more dataflow analysis

monotone framework

4

Trace Semantics

note that input (x) is unknown while loop? trace semantics is not computable

[y := x]1;[z := 1]2;while [y > 0]3 ( [z := z * y]4; [y := y − 1]5; )[y := 0]6

…

< 1,{ x42, y0, z0 } > < 2,{ x42, y42, z0 } >

< 3,{ x42, y42, z1 } > < 4,{ x42, y42, z1 }> …

[y := x]1 [z := 1]2

[y > 0]3

< 1,{ x73, y0, z0 } > < 2,{ x73, y73, z0 } >

< 3,{ x73, y73, z1 } > < 4,{ x73, y73, z1 }> …

[y := x]1 [z := 1]2

[y > 0]3

5

Abstract Interpretation

C C’concrete

’

set of states set of states

abstract state abstract state

abstract

S

S

C’’

6

Dataflow Analysis

Reaching Definitions The assignment lab: var := exp reaches lab’ if there

is an execution where var was last assigned at lab

1: y := x;2: z := 1;3: while y > 0 {4: z := z * y;5: y := y − 1 }6: y := 0

(adapted from Nielson, Nielson & Hankin)

{ (x,?), (y,?), (z,?) }

{ (x,?), (y,1), (z,?) }

{ (x,?), (y,1), (z,2) }

{ (x,?), (y,?), (z,4) }

{ (x,?), (y,5), (z,4) }

{ (x,?), (y,1), (z,2) }

{ (x,?), (y,?), (z,?) }

{ (x,?), (y,1), (z,2) }

7

Dataflow Analysis

Reaching Definitions The assignment lab: var := exp reaches lab’ if there

is an execution where var was last assigned at lab


(adapted from Nielson, Nielson & Hankin)

{ (x,?), (y,?), (z,?) }

{ (x,?), (y,1), (z,?) }

{ (x,?), (y,1), (z,2), (y,5), (z,4) }

{ (x,?), (y,?), (z,4), (y,5) }

{ (x,?), (y,5), (z,4) }

{ (x,?), (y,1), (z,2), (y,5), (z,4) }

{ (x,?), (y,6), (z,2), (z,4) }

{ (x,?), (y,1), (z,2), (y,5), (z,4) }

8

Dataflow Analysis

Build control-flow graph Assign transfer functions Compute fixed point

9

Control-Flow Graph


1: y:=x

2: z:=1

3: y > 0

4: z=z*y

5: y=y-1

6: y:=0

10

Transfer Functions

1: y:=x

2: z:=1

3: y > 0

4: z=z*y

5: y=y-1

6: y:=0

out(1) = in(1) \ { (y,l) | l Lab } U { (y,1) }

out(2) = in(2) \ { (z,l) | l Lab } U { (z,2) }

in(1) = { (x,?), (y,?), (z,?) } in(2) = out(1)in(3) = out(2) U out(5)in(4) = out(3)in(5) = out(4)in(6) = out(3)

out(4) = in(4) \ { (z,l) | l Lab } U { (z,4) }

out(5) = in(5) \ { (y,l) | l Lab } U { (y,5) }

out(6) = in(6) \ { (y,l) | l Lab } U { (y,6) }

out(3) = in(3)

11

System of Equationsin(1) = { (x,?), (y,?), (z,?) } in(2) = out(1)in(3) = out(2) U out(5)in(4) = out(3)in(5) = out(4)In(6) = out(3)out(1) = in(1) \ { (y,l) | l Lab } U { (y,1) }out(2) = in(2) \ { (z,l) | l Lab } U { (z,2) }out(3) = in(3)out(4) = in(4) \ { (z,l) | l Lab } U { (z,4) }out(5) = in(5) \ { (y,l) | l Lab } U { (y,5) }out(6) = in(6) \ { (y,l) | l Lab } U { (y,6) }F: ((Var x Lab) )12 ((Var x Lab) )12

12

System of Equations

in(1) = { (x,?), (y,?), (z,?) } in(2) = out(1)in(3) = out(2) U out(5)in(4) = out(3)in(5) = out(4)In(6) = out(3)out(1) = in(1) \ { (y,l) | l Lab } U { (y,1) }out(2) = in(2) \ { (z,l) | l Lab } U { (z,2) }out(3) = in(3)out(4) = in(4) \ { (z,l) | l Lab } U { (z,4) }out(5) = in(5) \ { (y,l) | l Lab } U { (y,5) }out(6) = in(6) \ { (y,l) | l Lab } U { (y,6) }

F: ((Var x Lab) )12 ((Var x Lab) )12

in(1)={(x,?),(y,?),(z,?)} , in(2)=out(1), in(3)=out(2) U out(5), in(4)=out(3), in(5)=out(4),In(6) = out(3)out(1) = in(1) \ { (y,l) | l Lab } U { (y,1) }, out(2) = in(2) \ { (z,l) | l Lab } U { (z,2) }out(3) = in(3), out(4) = in(4) \ { (z,l) | l Lab } U { (z,4) }out(5) = in(5) \ { (y,l) | l Lab } U { (y,5) }, out(6) = in(6) \ { (y,l) | l Lab } U { (y,6) }

In(1) {(x,?),(y,?),(z,?)}

== == ==

In(2) {(y,1)} {(x,?),(y,1),(z,?)}

==

In(3) {(z,2),(y,5)} {(z,2),(y,5)} {(z,2),(z,4),(y,5)}

In(4) {(z,2),(y,5)}

In(5) {(z,4)} {(z,4)} {(z,4)}

In(6) {(z,2),(y,5)}

Out(1)

{(y,1)} {(x,?),(y,1),(z,?)}

== ==

Out(2)

{(z,2)} {(z,2)} {(z,2),(y,1)} ==

Out(3)

{(z,2),(y,5)} {(z,2),(y,5)}

Out(4)

{(z,4)} {(z,4)} {(z,4)} {(z,4)}

Out(5)

{(y,5)} {(y,5)} {(z,4),(y,5)} {(z,4),(y,5)}

Out(6)

{(y,6)} {(y,6)} {(y,6)} {(y,6)}

…

13


RD RD’ when i: RD(i) RD’(i)

…

RD F(RD) F(F(RD)) F(F(F(RD))) F(F(F(F(RD))))

In(1) {(x,?),(y,?),(z,?)}

== == ==

In(2) {(y,1)} {(x,?),(y,1),(z,?)}

==

In(3) {(z,2),(y,5)} {(z,2),(y,5)} {(z,2),(z,4),(y,5)}

In(4) {(z,2),(y,5)}

In(5) {(z,4)} {(z,4)} {(z,4)}

In(6) {(z,2),(y,5)}

Out(1)

{(y,1)} {(x,?),(y,1),(z,?)}

== ==

Out(2)

{(z,2)} {(z,2)} {(z,2),(y,1)} ==

Out(3)

{(z,2),(y,5)} {(z,2),(y,5)}

Out(4)

{(z,4)} {(z,4)} {(z,4)} {(z,4)}

Out(5)

{(y,5)} {(y,5)} {(z,4),(y,5)} {(z,4),(y,5)}

Out(6)

{(y,6)} {(y,6)} {(y,6)} {(y,6)}

System of Equations

14

Monotonicity


RD F(RD) F(F(RD)) F(F(F(RD))) F(F(F(F(RD))))

…

Out(1)

{(y,1)} {(x,?),(y,1),(z,?)}

{(y,1)} ….

…

…

out(1) = {(y,1)} when (x,?) out(1)in(1) \ { (y,l) | l Lab } U { (y,1) }, otherwise(silly example just for illustration)

15

Convergence

0

1…

2

n

0

1

…

2

n

0

1

…

2

16

Least Fixed Point

We will see later why it exists For now, mostly informally…F: ((Var x Lab) )12 ((Var x Lab) )12

RD RD’ when i: RD(i) RD’(i)

F is monotone: RD RD’ implies that F(RD) F(RD’)

RD = (, ,…,)

F(RD), F(F(RD) , F(F(F(RD)), … Fn(RD)

Fn+1(RD) = Fn(RD)

RD

F(RD)

F(F(RD))

…

Fn(RD)

17

Partially Ordered Sets (posets) Partial ordering is a relation

: L L { true, false} that is: Reflexive (l: l l) Transitive (l1,l2,l3: l1 l2 l2 l3 l1 l3) Anti-symmetric ( l1 l2 l2 l1 l1 = l2)

A partially ordered set (L, ) is a set L equipped with a partial ordering

For example: ((S), )

Captures intuition of implication between facts l1 l2 will intuitively mean l1 l2 Later, in abstract domains , when l1 l2 , we will say that l1 is

“more precise” than l2 (represents fewer concrete states)

18

Bounds in Posets

Given a poset (L, ) A subset Y of L has l L as an upper bound if l’ Y : l’ l A subset Y of L has l L as a lower bound if l’ Y : l’ l

(we write l’ l when l l’) A least upper bound (lub) l of Y is an upper bound of Y

that satisfies l l’ whenever l’ is another upper bound of Y A greatest lower bound (glb) l of Y is a lower bound of Y

that satisfies l l’ whenever l’ is another lower bound of Y

For any subset Y L If the lub exists then it is unique, and denoted by Y (join)

We often write l1 l2 for { l1, l2} If the glb exists then it is unique, and denoted by Y (meet)

We often write l1 l2 for { l1, l2}

19

Complete Lattices

A complete lattice (L, ) is a poset such that all subsets have least upper bounds as well as greatest lower bounds

In particular = = L is the least element (bottom) = L = is the greatest element (top)

20

Complete Lattices: Examples

{1}{2}

{3}

{1,2}{1,3}

{2,3}

{1,2,3}

- +

0

21

Complete Lattices: Examples

[0,0][-1,-1][-2,-2]

[-2,-1]

[-2,0]

[1,1] [2,2]

[-1,0] [0,1] [1,2]

…

[-1,1] [0,2]

[-2,1] [-1,2]

[-2,2]

……

[2,]…

[1,]

[0,]

[-1,]

[-2,]

…

…

…

…

[- ,]

…

…

[- ,-2]…

[-,-1]

[- ,0]

[-,1]

[- ,2]

…

…

…

…

22

Moore Families

A Moore family is a subset Y of a complete lattice (L,) that is closed under greatest lower bounds Y’ Y: Y’ Y Hence, a Moore family is never empty Always contains least element Y and a greatest

element Why do we care?

We will see that Moore families correspond to choices of abstractions (as targets of upper closure operators)

More on this… later…

23

Moore Families: Examples

{1}{2}

{3}

{1,2}{1,3}

{2,3}

{1,2,3}

{2}

{1,2} {2,3}

{1,2,3}

{1,2,2}

24

Chains

In program analysis we will often look at sequences of properties of the form l0 l1 … ln

Given a poset (L, ) a subset Y L is a chain if every two elements in Y are ordered l1,l2 Y: (l1 l2) or (l2 l1) When Y is a finite subset we say that the chain is a finite chain

A sequence of elements l0 l1 … L is an ascending chain A sequence of elements l0 l1 … L is a descending chain L has finite height when all chains in L are finite A poset (L, ) satisfies the ascending chain condition (ACC) if

every ascending chain eventually stabilizes, that is s N: n N: n s ln = ls

25

Chains: Example Posets

-

-2

…

-1

0

0

1

…2

… …

- 0-1 1

(a) (b) (c)

26

Properties of Functions

A function f: L1 L2 between two posets (L1,1) and (L2, 2) is

Monotone (order-preserving) l,l’ L1: l 1 l’ f(l) 2 f(l’) Special case f: L L, l,l’ L: l l’ f(l)

f(l’)

Distributive (join morphism) l,l’ L1 f(l l’) = f(l) f(l’)

27

Function Properties: Examples

{1}{2}

{3}

{1,2}{1,3}

{2,3}

{1,2,3}

28


{2}

{1,2} {2,3}

{1,2,3}

{2}

{1,2} {2,3}

{1,2,3}

29


{2}

{1,2} {2,3}

{1,2,3}

{2}

{1,2} {2,3}

{1,2,3}

30


{1}{2}

{3}

{1,2}{1,3}

{2,3}

{1,2,3}

{1}{2}

{3}

{1,2}{1,3}

{2,3}

{1,2,3}

31

Fixed Points

Consider a complete lattice L = (L,,,,,) and a monotone function f: L L

An element l L is A fixed point iff f(l) = l A pre-fixedpoint iff l f(l) A post-fixedpoint iff l f(l)

Fix(f) = set of all fixed points

Ext(f) = set of all pre-FP Red(f) = set of all post-FP

Red(f)

Ext(f)

Fix(f)

lfp

gfp

fn()

fn()

32

Tarski’s Theorem

Consider a complete lattice L = (L,,,,,) and a monotone function f: L L, then lfp(f) and gfp(f) exist and lfp(f) = Red(f) = Fix(f) gfp(f) = Ext(f) = Fix(f)

33

Kleene’s Fixedpoint Theorem Consider a complete lattice L = (L,,,,,) and a continuous

function f: L L then

lfp(f) = n N fn()

A function f is continuous if for every increasing chain Y L, f(Y) = { f(y) | y Y }

Monotone functions on posets satisfying ACC are continuous Every ascending chain eventually stabilizes

l0 l1 … ln = ln+1 = … hence ln is the least upper bound of { l0,l1,…,ln } , thus f(Y) = f(ln)

From monotonicity of f f(l0) f(l1) … f(ln) = f(ln+1) = … Hence f(ln) is the least upper bound of { f(l0),f(l1), … ,f(ln) } , thus { f(y) | y Y } = f(ln)

34

An Algorithm for Computing lfp Kleene’s fixed point theorem gives a

constructive method for computing the lfp

lfp(f) = n N fn()

l = while f(l) l do l = f(l)

In our case, can compute efficiently using a worklist algorithm for “chaotic iteration”

35

Finally…

36

Reaching Definitions…in(1) = { (x,?), (y,?), (z,?) } in(2) = out(1)in(3) = out(2) U out(5)in(4) = out(3)in(5) = out(4)In(6) = out(3)out(1) = in(1) \ { (y,l) | l Lab } U { (y,1) }out(2) = in(2) \ { (z,l) | l Lab } U { (z,2) }out(3) = in(3)out(4) = in(4) \ { (z,l) | l Lab } U { (z,4) }out(5) = in(5) \ { (y,l) | l Lab } U { (y,5) }out(6) = in(6) \ { (y,l) | l Lab } U { (y,6) }


l = while F(l) l do l = F(l)

37

Algorithm Revisited

Input: equation system for RD Output: lfp(F) Algorithm:

RD1 = , … , RD12 = While RDj Fj(RD1,…,RD12) for some j

RDj = Fj(RD1,…,RD12)

Idea: non-deterministically select which component of RD should make a step

Can make selection efficiently based on dependencies We have information on what equations depend on Fj that

has been updated, need only recompute these on every step

38

But…

where did the transfer functions come from? e.g., out(1) = in(1) \ { (y,l) | l Lab } U

{ (y,1) }

how do we know transfer functions really over-approximate the concrete operations?

39

Trace Semantics

note that input (x) is unknown while loop? trace semantics is not computable

[y := x]1;[z := 1]2;while [y > 0]3 ( [z := z * y]4; [y := y − 1]5; )[y := 0]6

…

< 1,{ x42, y0, z0 } > < 2,{ x42, y42, z0 } >

< 3,{ x42, y42, z1 } > < 4,{ x42, y42, z1 }> …

[y := x]1 [z := 1]2

[y > 0]3

< 1,{ x73, y0, z0 } > < 2,{ x73, y73, z0 } >

< 3,{ x73, y73, z1 } > < 4,{ x73, y73, z1 }> …

[y := x]1 [z := 1]2

[y > 0]3

40

Instrumented Trace Semantics on every assignment, record the

label in which the assignment was performed

41

Abstract Interpretation

Instrumented trace semantics


(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)(y,5)(z,4)(y,5)(y,6)

…

(x,?)(y,?)(z,?)(y,1)(z,2)(y,6)

42

Collecting Semantics

1: y:=x

2: z:=1

3: y > 0

4: z=z*y

5: y=y-1

6: y:=0

(x,?)(y,?)(z,?)(y,1)

(x,?)(y,?)(z,?) (y,1)(z,2)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)

(x,?)(y,?)(z,?)(y,1)(z,2)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)

(x,?)(y,?)(z,?)(y,1)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)(y,5)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)

(x,?)(y,?)(z,?)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)

(x,?)(y,?)(z,?)(y,1)(z,2) (y,6)

43

System of Equationsin(1) = { (x,?)(y,?)(z,?) } in(2) = out(1)in(3) = out(2) U out(5)in(4) = out(3)in(5) = out(4)In(6) = out(3)out(1) = { (y,1) | in(1) }out(2) = { (z,2) | in(2) }out(3) = in(3)out(4) = { (z,4) | in(4) }out(5) = { (y,5) | in(5) }out(6) = { (y,6) | in(6) }

G: ((Trace) )12 ((Trace) )12

Trace = (Var x Lab)* TR

G(TR)

G(G(TR))

…

Gn(TR)

…

lfp(G)

44

Galois Connection

C (C)

(A) A

(C) A C (A)

45

C(C)

(A)A

Reaching Definitions: Abstraction

DOM() = variables in SRD()(v) = l iff rightmost pair for v in is

(v,l)

(C) = { (v,SRD()(v)) | v DOM() C }

(A) = { | v DOM() : (v,SRD()(v) ) A }

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(y,6)

(x,?),(z,4),(y,6)

(x,?),(z,4),(y,6)

(x,?),(z,2),(y,6)

(x,?),(z,2),(z,4),(y,6)

Set of Traces Set of (var,lab) pairs

46

C(C)

(A)A

Reaching Definitions: Concretization

DOM() = variables in SRD()(v) = l iff rightmost pair for v in is

(v,l)

(C) = { (v,SRD()(v)) | v DOM() C }

(A) = { | v DOM() : (v,SRD()(v) ) A }

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)(z,4)(y,5)(y,6)

(x,?),(z,4),(y,6)

(x,?)(y,?)(z,?)(y,1)(z,2)(z,4)(y,5)…(z,4)(y,5)(y,6)

…

47

Best (Induced) Transformer

C

A

C’

A’

How do we figure out the abstract effect S# of a statement S ?

?

S#

S

48

Sound Abstract Transformer

C

A

C’

A’

S#S

Aopt

S (A) S#(A)

49

Soundness of Induced Analysis

RD

F(RD)

F(F(RD))

…

lfp(F)

TR

G(TR)

G(G(TR))

…

Gn(TR)

…lfp(G)

(lfp(G))

(lfp(G)) lfp( G ) lfp(F)

50

Back to a bit of dataflow analysis…

51

Recap

Represent properties of a program using a lattice (L,,,,,)

A continuous function f: L L Monotone function when L satisfies ACC

implies continuous Kleene’s fixedpoint theorem

lfp(f) = n N fn()

A constructive method for computing the lfp

52

Some required notation

blocks : Stmt P(Blocks)blocks([x := a]lab) = {[x := a]lab}blocks([skip]lab) = {[skip]lab}blocks(S1; S2) = blocks(S1) blocks(S2)blocks(if [b]lab then S1 else S2) = {[b]lab} blocks(S1) blocks(S2)blocks(while [b]lab do S) = {[b]lab} blocks(S)

FV: (BExp AExp) Var Variables used in an expressionAExp(a) = all non-unit expressions in the arithmetic expression asimilarly AExp(b) for a boolean expression b

53

Available Expressions Analysis

For each program point, which expressions must have already been computed, and not later modified, on all paths to the program point

[x := a+b]1;[y := a*b]2;while [y > a+b]3 ( [a := a + 1]4; [x := a + b]5

)

(a+b) always available at label

3

54

Available Expressions Analysis Property space

inAE, outAE: Lab (AExp) Mapping a label to set of arithmetic

expressions available at that label

Dataflow equations Flow equations – how to join incoming

dataflow facts Effect equations - given an input set of

expressions S, what is the effect of a statement

55

inAE (lab) = when lab is the initial label { outAE (lab’) | lab’ pred(lab) }

otherwise

outAE (lab) = …Block out (lab)

[x := a]lab in(lab) \ { a’ AExp | x FV(a’) } U { a’ AExp(a) | x FV(a’) }

[skip]lab in(lab)

[b]lab in(lab) U AExp(b)

Available Expressions Analysis

From now on going to drop the AE subscript when clear from context

56

Transfer Functions1: x = a+b

2: y:=a*b

3: y > a+b

4: a=a+1

5: x=a+b

out(1) = in(1) U { a+b }

out(2) = in(2) U { a*b }

in(1) = in(2) = out(1)in(3) = out(2) out(5)in(4) = out(3)in(5) = out(4)

out(4) = in(4) \ { a+b,a*b,a+1 }

out(5) = in(5) U { a+b }

out(3) = in(3) U { a+ b }

[x := a+b]1;[y := a*b]2;while [y > a+b]3 ( [a := a + 1]4; [x := a + b]5

)

57

Solution

1: x = a+b

2: y:=a*b

3: y > a+b

4: a=a+1

5: x=a+b

in(2) = out(1) = { a + b }

out(2) = { a+b, a*b }

out(4) =

out(5) = { a+b }

in(4) = out(3) = { a+ b }

in(1) =

in(3) = { a + b }

58

Kill/Gen

Block out (lab)

[x := a]lab in(lab) \ { a’ AExp | x FV(a’) } U { a’ AExp(a) | x FV(a’) }

[skip]lab in(lab)

[b]lab in(lab) U AExp(b)

Block kill gen

[x := a]lab

{ a’ AExp | x FV(a’) }

{ a’ AExp(a) | x FV(a’) }

[skip]lab

[b]lab AExp(b)

out(lab) = in(lab) \ kill(Blab) U gen(Blab)

Blab = block at label lab

59

Why solution with largest sets?

1: z = x+y

2: true

3: skip

out(1) = in(1) U { x+y }in(1) = in(2) = out(1) out(3)in(3) = out(2)

out(3) = in(3)

out(2) = in(2)[z := x+y]1;while [true]2 ( [skip]3;)

in(1) =

After simplification: in(2) = in(2) { x+y }

Solutions: {x+y} or

in(2) = out(1) out(3)

in(3) = out(2)

60

Reaching Definitions Revisited

Block out (lab)

[x := a]lab

in(lab) \ { (x,l) | l Lab } U { (x,lab) }

[skip]la

b

in(lab)

[b]lab in(lab) Block kill gen

[x := a]lab

{ (x,l) | l Lab } { (x,lab) }

[skip]lab

[b]lab

For each program point, which assignments may have been made and not overwritten, when program execution reaches this point along some path.

61

Why solution with smallest sets?

1: z = x+y

2: true

3: skip

out(1) = ( in(1) \ { (z,?) } ) U { (z,1) }in(1) = { (x,?),(y,?),(z,?) } in(2) = out(1) U out(3)in(3) = out(2)

out(3) = in(3)

out(2) = in(2)[z := x+y]1;while [true]2 ( [skip]3;)

in(1) = { (x,?),(y,?),(z,?) }

After simplification: in(2) = in(2) U { (x,?),(y,?),(z,1) }

Many solutions: any superset of { (x,?),(y,?),(z,1) }

in(2) = out(1) U out(3)

in(3) = out(2)

62

Live Variables

For each program point, which variables may be live at the exit from the point.

[ x :=2]1; [y:=4]2; [x:=1]3; (if [y>x]4 then [z:=y]5 else [z:=y*y]6); [x:=z]7


Live Variables

1: x := 2

2: y:=4

4: y > x

5: z := y

7: x := z

6: z = y*y

3: x:=1

64

1: x := 2

2: y:=4

4: y > x

5: z := y

7: x := z


6: z = y*y

Live Variables

Block kill gen

[x := a]lab

{ x } { FV(a) }

[skip]la

b

[b]lab FV(b)

3: x:=1

65

1: x := 2

2: y:=4

4: y > x

5: z := y

7: x := z


6: z = y*y

Live Variables: solution

Block kill gen

[x := a]lab

{ x }

{ FV(a) }

[skip]lab

[b]lab FV(b)

3: x:=1

in(1) =

out(1) = in(2) =

out(2) = in(3) = { y }

out(3) = in(4) = { x,y }

in(7) = { z }

out(7) =

out(6) = { z }

in(6) = { y }

out(5) = { z }

in(5) = { y }

in(4) = { x,y }

out(4) = { y }

66

Why solution with smallest set?

1: x>1

2: skip

out(1) = in(2) U in(3)out(2) = in(1)out(3) =

out(3) =

in(2) = out(2)

while [x>1]1 ( [skip]2;)[x := x+1]3;

After simplification: in(1) = in(1) U { x }

Many solutions: any superset of { x }

3: x := x + 1

in(3) = { x }

in(1) = out(1) { x }

67

Monotone Frameworks

is or CFG edges go either forward or backwards Entry labels are either initial program labels or final

program labels (when going backwards) Initial is an initial state (or final when going backwards) flab is the transfer function associated with the blocks Blab

In(lab) = { out(lab’) | (lab’,lab) CFG edges }

Initial when lab Entry labels

otherwise

out(lab) = flab(in(lab))

68

Forward vs. Backward Analyses

1: x := 2

2: y:=4

4: y > x

5: z := y

7: x := z

6: z = y*y

1: x := 2

2: y:=4

4: y > x

5: z := y

7: x := z

6: z = y*y

{(x,1), (y,?), (z,?) }

{ (x,?), (y,?), (z,?) }

{(x,1), (y,2), (z,?) }

{ z }

{ y }{ y }

69

Must vs. May Analyses

When is - must analysis Want largest sets the solves the

equation system Properties hold on all paths reaching a

label (exiting a label, for backwards)

When is - may analysis Want smallest sets that solve the

equation system Properties hold at least on one path

reaching a label (existing a label, for backwards)

70

Example: Reaching Definition

L = (Var×Lab) is partially ordered by

is L satisfies the Ascending Chain

Condition because Var × Lab is finite (for a given program)

71

Example: Available Expressions L = (AExp) is partially ordered by is L satisfies the Ascending Chain

Condition because AExp is finite (for a given program)

72

Analyses Summary

Reaching Definitions

Available Expressions

Live Variables

L (Var x Lab) (AExp) (Var)

AExp

Initial { (x,?) | x Var}

Entry labels { init } { init } final

Direction Forward Forward Backward

F { f: L L | k,g : f(val) = (val \ k) U g }

flab flab(val) = (val \ kill) gen

73

Analyses as Monotone Frameworks Property space

Powerset Clearly a complete lattice

Transformers Kill/gen form Monotone functions (let’s show it)

74

Monotonicity of Kill/Gen transformers

Have to show that x x’ implies f(x) f(x’)

Assume x x’, then for kill set k and gen set g(x \ k) U g (x’ \ k) U g

Technically, since we want to show it for all functions in F, we also have to show that the set is closed under function composition

75

Distributivity of Kill/Gen transformers

Have to show that f(x y) f(x) f(y) f(x y) = ((x y) \ k) U g

= ((x \ k) (y \ k)) U g= (((x \ k) U g) ((y \ k) U g))= f(x) f(y)

Used distributivity of and U Works regardless of whether is U or

76

Constant Propagation

… …

- 0-1 1

State = (Var Z)

No information

Variable not a constant

77

Constant Propagation

L = ((Var Z) , ) 1 2 iff x: 1(x) 2(x) ordering in the Z

lattice

Examples: [x , y 42, z ] [x , y 42, z

73] [x , y 42, z 73] [x , y 42,

z ]

78

Constant PropagationA: AExp (State Z)

Aa1 op a2 = Aa1 op Aa2

Ax =(x) otherwise

If = An =

n otherwise

If =

Block out ()

[x := a]lab

If = [x Aa] otherwise

[skip]la

b

[b]lab

79

Example

[x :=42]1; [y:=73]2; (if [?]3 then [z:=x+y]4 else [z:=12]5

[w:=z]6

);

[X , y , z , w ]

[X 42, y , z , w ]

[X 42, y 73, z , w ]

[X 42, y 73, z , w ]

[X 42, y 73, z 115, w ]

[X 42, y 73, z 12, w ]

[X 42, y 73, z 12, w 12]

[X 42, y 73, z , w 12]

… …

- 0 12 115… …

80

Constant Propagation is Non Distributive

Consider the transformer f = [y=x*x]#

Consider two states 1, 2 1(x) = 1 2(x) = -1

(1 2)(x) = f(1 2) maps y to

f(1) maps y to 1f(2) maps y to 1f(1) f(2) maps y to 1

f(1 2) f(1) f(2)

lecture 03 – background + intro ai + dataflow and connection to ai eran yahav 1

Documents