Context-sensitive Analysis Part II Chapter 4 (up to Section 4.3)

Transcription

1 Context-sensitive Analysis Part II Chapter 4 (up to Section 4.3)

2 Attribute Grammars Add rules to compute the decimal value of a signed binary number

3 Two kinds of Attributes Synthesized attribute Bottom-Up flow of values Depends on values from the node itself, children, or constants Inherited attribute Top-down flow of values Depends on values from siblings, parent and constants

4 Synthesized Attributes Depends on values from the node itself, children, or constants.val.val False.neg false True.neg true Bottom-Up flow (Children) Constants

5 Synthesized Attributes Depends on values from the node itself, children, or constants.val.val False.neg false True.neg true Bottom-Up flow (Children) Constants

6 Inherited Attributes Depends on values from siblings, parent and constants 0.pos 0.pos +.pos 0.pos Top-down flow

7 Back to the Examples For.neg

8 Back to the Examples For.neg.pos.val

9 Back to the Examples For.neg.pos.val.pos.val

10 Back to the Examples For Number.val Number.neg.pos.val.pos.val

11 Back to the Examples For Number.val Number.neg.pos.val.pos.val

12 Back to the Examples For Number.val Number.neg.pos.val.pos.val

13 Back to the Examples For Number.val Number.neg.pos 0.val.pos.val

14 Back to the Examples For Number.val Number.neg.pos 0.val.pos.val

15 Back to the Examples For Number.val Number.neg true.pos 0.val.pos.val

16 Back to the Examples For Number.val Number.neg true.pos 0.val.pos.val

17 Back to the Examples For Number.val Number.neg true.pos 0.val.pos 0.val

18 Back to the Examples For Number.val Number.neg true.pos 0.val.pos 0.val

19 Back to the Examples For Number.val Number.neg true.pos 0.val.pos 0.val 2.pos º

20 Back to the Examples For Number Number.val.val º.neg true.pos 0.val.val º.pos 0.val 2.pos º

21 Back to the Examples For Number Number.val.val º One possible evaluation order:.neg true.pos 0.val.val º.pos 2.neg.pos 0.val 2.pos º 3.pos 4.val 5.val 6 Number.val Evaluation order must be consistent with the attribute dependence graph Other orders are possible

22 Attributes + Parse Tree Attributes associated with nodes in parse tree Rules are value assignments associated with productions Rules & parse tree define an attribute dependence graph Dependence graph must be non-circular (no cycles) This produces a high-level, functional specification

23 Using Attribute Grammars Attribute grammars can specify contextsensitive actions Take values from syntax Perform computations with values Insert type tests, type inference, logic,

24 Evaluation Methods Dynamic, dependence-based methods Build the parse tree Build the dependence graph Topological sort the dependence graph Define attributes in topological order Rule-based methods Analyze rules at compiler-generation time Determine a fixed (static) ordering Evaluate nodes in that order (treewalk) Oblivious methods Ignore rules & parse tree Pick a convenient order (at design time) & use it (passes, dataflow)

25 Back to the Example Number 0 For 0

26 Back to the Example Number val: neg: pos: 0 val: pos: val: pos: val: pos: val: pos: val: pos: val: 0 For 0

27 Back to the Example Number val: 5 Inherited Attributes neg: true pos: val: 4 pos: 0 val: 5 pos: 0 val: Note: downward flow (pointing arrows) of information pos: 2 val: 4 pos: val: 0 pos: 2 val: 4 0 For 0

28 Back to the Example Number val: 5 Synthesized attributes neg: true pos: 2 val: 4 pos: val: 4 pos: val: 0 pos: 0 val: 5 pos: 0 val: Note: upward flow (pointing arrows) of information and the flow from node s (self) attributes pos: 2 val: 4 0 For 0

29 Back to the Example Number val: 5 neg: true pos: val: 4 pos: 0 val: 5 pos: 0 val: If we show the computation... pos: 2 val: 4 pos: 2 val: 4 0 pos: val: 0 then peel away the parse tree... For 0

30 Back to the Example neg: true pos: 2 val: 4 pos: 2 val: 4 val: 5 pos: val: 4 0 pos: val: 0 pos: 0 val: 5 pos: 0 val: For 0 All that is left is the attribute dependence graph. This succinctly represents the flow of values in the problem instance. The dependence graph must be acyclic (no cycles!)

31 An Extended Example Grammar for a basic block

32 An Extended Example Grammar for a basic block

33 An Extended Example Grammar for a basic block

34 An Extended Example Grammar for a basic block Example basic block a = -5 b = a * 7 c = b / 2 d = a + b - c How many clock cycles will this block take to execute?

35 An Extended Example Grammar for a basic block Simple Attribute Grammar Estimate cycle count for the block of instructions Each operation has a COST Add them, bottom up Assume a load per value Assume no reuse

36 An Extended Example (continued) Adding attribution rules All these attributes are synthesized!

37 An Extended Example (continued) Adding attribution rules All these attributes are synthesized!

38 An Extended Example (continued) Adding attribution rules All these attributes are synthesized!

39 An Extended Example (continued) Adding attribution rules All these attributes are synthesized!

40 An Extended Example (continued) Adding attribution rules All these attributes are synthesized!

41 An Extended Example Properties of the example grammar All attributes are synthesized Þ S-attributed grammar Rules can be evaluated bottom-up in a single pass Good fit to bottom-up, shift/reduce parser Easily understood solution Seems to fit the problem well What about an improvement? (see backup slides) Values are loaded only once per block (not at each use) Need to track which values have been already loaded

42 Backup Slides

43 A Better Execution Model Adding load tracking Need sets Before and After for each production Must be initialized, updated, and passed around the tree This looks more complex!

44 A Better Execution Model Adding load tracking Need sets Before and After for each production Must be initialized, updated, and passed around the tree This looks more complex!

45 A Better Execution Model Load tracking adds complexity Every production needs rules to copy Before & After A sample production Lots of work, lots of space, lots of rules to write

46 An Even Better Model What about accounting for finite register sets? Before & After must be of limited size Adds complexity to Factor Identifier Requires more complex initialization Jump from tracking loads to tracking registers is small Copy rules are already in place Some local code to perform the allocation