1. Runtime Environments

2. Support of Execution  Activation Tree  Control Stack  Scope  Binding of Names –Data object (values in storage) –Environment (functions that map to stg) –State (funct that maps a stg location to the value held there)

3. Storage Allocation Strategies  Static –Names are bound to storage at compile time  Dynamic – names are bound to storage at run time

4. Storage Organization CODE CODE STATIC DATA STATIC DATA STACK STACK \/ \/ /\ /\ HEAP HEAP

5. Stack  Allocate activation record  Locals get new storage  Enters information into its fields

6. Activation Record (Stack Frames)  Returned value  Actual parameters  Optional control link  Optional access link  Saved machine status  Local data  Temporary data

7. Calling Sequence  Caller evaluates arguments  Caller stores return address in callee’s activation record  Caller stores stack top  Callee saves register values and status information for caller

8. Return Sequence.  Callee restores state of machine  Callee places return value next to the activation record of the caller  Restores top of stack pointer  Caller copies return value

9. Variable Length Data  Arrays –Stored after the activation record –The activation record does not have to allocate space for the

10. Dangling Reference  #include  #include  int *dangle();  main(){  int *p;  p = dangle(); printf("p = %d\n",*p);  sub(); printf("p = %d\n",*p);  }  int *dangle(){  int i = 23;  return &i;  }  sub(){  int i,j,k;  }

11. Heap vs Stack Allocation  Stack allocation cannot be used if –Local names must be retained after activation ends –Activation outlives caller

12. Heap Management Strategies  free list  first fit  best fit  worst fit

13. Garbage Collection  records not reachable  reclaim to allow reuse  performed by runtime system (support programs linked with the compiled code) (support programs linked with the compiled code)

14. Record Types  live – will be used in the future  not live – will not be used in the future  reachable – able to be accessed via programs

15. Types of Algorithms  Mark-And-Sweep Collection  Reference Counts  Copying Collection  Generational Collection  Incremental Collection

16. Mark-And-Sweep Collection  Program variables and heap records form a directed graph  Roots are the variables  node n is reachable if r -> … -> n r -> … -> n  Depth first search marks reachable nodes  Any node not marked is garbage

17. Cost of Garbage Collection  Depth first search takes time proportional to the number of reachable nodes  Sweep phase takes time proportional to the size of the heap

18. Maintaining Free Space  Create a list of free space  Search for a space of size N might be long  Maintain several free lists of differing sizes  External fragmentation a problem  Internal fragmentation can also be a problem

19. Reference Counts  Count the number of pointers pointing to each record  Store the reference count with each record  If p addresses an alternate record, decrement the old and increment the new  If count reaches 0, free record

20. When to Decrement Instead of decrementing the counts a record references when the record is placed on the free list, it is better to do this when the record is removed from the free list. Instead of decrementing the counts a record references when the record is placed on the free list, it is better to do this when the record is removed from the free list.

21. Why  Breaks the recursive decrementing work into shorter pieces  Compiler emits code to check whether the count has reached 0, but the recursive decrementing will be done only in one place, in the allocator

22. Problems with Reference Count  Cycles of garbage cannot be reclaimed  Incrementing the reference counts is very expensive

23. Solutions-Cycles, Expensive  Require the programmer to break the cycle  Combine reference counting with mark-sweep  No solution for it being expensive  Problems outweigh advantages, thus rarely used

24. Copying Collection  Reachable part is a directed graph with records as nodes, pointers as edges, and variables as roots  Copy the graph from “from-space” to “to-space”  Delete all “from-space”qq

25. Access to Nonlocal Names  Lexical scope without nested procedures –Allows nonlocals to be found via static addresses –Uses physical layout –All storage locations known at compile time –Functions can be passed as parameters

26. Lexical Scope Example main { /* main */ A.R. main p(); … p(); … } /* main */ A.R. p p{ control link main p{ control link main int n; no access link int n; no access link n = 1; … n = 1; … r(2); A.R. r r(2); A.R. r fun q{ /* inside of p */ contol link p fun q{ /* inside of p */ contol link p n = 5; /*n non-local non-global*/ access link p n = 5; /*n non-local non-global*/ access link p } … } … fun r(int n){ /* inside of p */ A.R. q fun r(int n){ /* inside of p */ A.R. q q(); control link r q(); control link r }/* r */ access link p }/* r */ access link p } … } …

27. Access Chaining Example main{ A.R. main p(); … p(); … fun p{ A.R. p fun p{ A.R. p int x; ctl link main, acc main int x; ctl link main, acc main q(); A.R. q q(); A.R. q fun q{ ctl link p, acc link p fun q{ ctl link p, acc link p r(); A.R. r r(); A.R. r fun r{ /* fun r */ ctl link q, acc link q fun r{ /* fun r */ ctl link q, acc link q x = 2; A.R. p x = 2; A.R. p if... then p(); ctl link r, acc link main if... then p(); ctl link r, acc link main } /* fun r */ A.R. q } /* fun r */ A.R. q } /* fun q */ ctl link p, acc link p } /* fun q */ ctl link p, acc link p } /* fun p */ A.R. r } /* fun p */ A.R. r } /* fun main */ ctl link q, acc link q

28. Passing Function Example main{ A.R. main q(); … q(); … fun p (fun a) { A.R. q fun p (fun a) { A.R. q a(); ctl link main, acc main a(); ctl link main, acc main } /* end p */ A.R. p } /* end p */ A.R. p fun q { ctl link q, acc main fun q { ctl link q, acc main int x; A.R. a int x; A.R. a x = 2; ctl link p, acc q x = 2; ctl link p, acc q p(r); p(r); fun r{ fun r{ printf( x ); printf( x ); } /* end r */ } /* end r */ } /* end q */ } /* end q */ } /* end main */

29. Dynamic Scope lisp, apl, snobol, spitbol, scheme main(){ float r; float r; r :=.25; r :=.25; show; small; show; small; fun show{ fun show{ printf(r); printf(r); } fun small{ fun small{ fun r; fun r; r := 0.125; r := 0.125; show; show; }}  What is printed?

30. Parameter Passing  Call by value  Call by reference (address)  Call by copy restore  Call by name

31. Call by Value  Only the value is passed  Storage is in the A.R. of called function  Caller evaluates and places the value in callee A.R.  Operations do not effect original value

32. Call by Reference  Caller passes pointer to callee  if a+b is passed, the address of a temporary is used  Consider swap(i,a[i]) –Does indeed swap –Addresses are bound at time of call

33. Call by Copy-Restore  Value of argument is given to callee  Upon completion value is copied back to caller  swap(i, a[i]) works correctly

34. Copy-Restore/Reference Example int a = 1; main() { unsafe(a); unsafe(a); print(a); print(a);} fun unsafe( int x) { x = 2; x = 2; a = 0; a = 0;}

35. Copy-Restore/Reference Example Nested Functions main() { int a = 1; int a = 1; unsafe(a); unsafe(a); print(a); print(a); fun unsafe( int x) { fun unsafe( int x) { x = 2; x = 2; a = 0; a = 0; }}

36. Call by Name  A macro  Body of the function replaces the call  Local values are protected  swap(i, a[i]) does not work since i will have changed value

37. Call by Name Example  #include  #include  int temp;  #define swap(x,y) { temp = x, x = y, y = temp; }   main(){  int i;  int a[5] ={1,2,3,4,5};  for(i = 0; i < 5; i++)  printf("a[%d]=%d ",i,a[i]);  i = 3;  swap(i,a[i]);  for(i = 0; i < 5; i++ )  printf("a[%d]=%d ",i,a[i]);  }  Prints a[0]=1 a[1]=2 a[2]=3 a[3]=4 a[4]=3