CS 153 Design of Operating Systems

Transcription

1 CS 153 Design of Operating Systems Spring 18 Lectre 26: Dynamic Memory (2) Instrctor: Chengy Song Slide contribtions from Nael Ab-Ghazaleh, Harsha Madhyvasta and Zhiyn Qian Some slides modified from originals by Dave O hallaron

2 Keeping Track of Free Blocks Method 1: Implicit list sing length links all blocks Method 2: Explicit list among the free blocks sing pointers Method 3: Segregated free list Different free lists for different size classes Method 4: Blocks sorted by size Can se a balanced tree (e.g. Red-Black tree) with pointers within each free block, and the length sed as a key 2

3 Explicit Free Lists Allocated (as before) Free Size a Size a Next Payload and padding Prev Size a Size a Maintain list(s) of free blocks, not all blocks The next free block cold be anywhere» So we need to store forward/back pointers, not jst sizes Still need bondary tags for coalescing Lckily we track only free blocks, so we can se payload area 3

4 Explicit Free Lists Logically: A B C Physically: blocks can be in any order Forward (next) links A B C Back (prev) links CS 153 Lectre 26 Dynamic Memory (2) 4

5 Allocating From Explicit Free Lists Before conceptal graphic After (with splitting) = malloc( ) CS 153 Lectre 26 Dynamic Memory (2) 5

6 Freeing With Explicit Free Lists Insertion policy: Where in the free list do yo pt a newly freed block? LIFO (last-in-first-ot) policy» Insert freed block at the beginning of the free list» Pro: simple and constant time» Con: stdies sggest fragmentation is worse than address ordered Address-ordered policy» Insert freed blocks so that free list blocks are always in address order: addr(prev) < addr(crr) < addr(next)» Con: reqires search» Pro: stdies sggest fragmentation is lower than LIFO CS 153 Lectre 26 Dynamic Memory (2) 6

7 Freeing With a LIFO Policy (Case 1) Before free( ) conceptal graphic Root Insert the freed block at the root of the list After Root CS 153 Lectre 26 Dynamic Memory (2) 7

8 Freeing With a LIFO Policy (Case 2) Before free( ) conceptal graphic Root Splice ot predecessor block, coalesce both memory blocks, and insert the new block at the root of the list After Root 8

9 Freeing With a LIFO Policy (Case 3) Before free( ) conceptal graphic Root Splice ot sccessor block, coalesce both memory blocks and insert the new block at the root of the list After Root 9

10 Freeing With a LIFO Policy (Case 4) Before free( ) conceptal graphic Root Splice ot predecessor and sccessor blocks, coalesce all 3 memory blocks and insert the new block at the root of the list After Root 10

11 Explicit List Smmary Comparison to implicit list: Allocate is linear time in nmber of free blocks instead of all blocks» Mch faster when most of the memory is fll Slightly more complicated allocate and free since needs to splice blocks in and ot of the list Some extra space for the links (2 extra words needed for each block)» Does this increase internal fragmentation? Most common se of linked lists is in conjnction with segregated free lists Keep mltiple linked lists of different size classes, or possibly for different types of objects CS 153 Lectre 26 Dynamic Memory (2) 11

12 Keeping Track of Free Blocks Method 1: Implicit list sing length links all blocks Method 2: Explicit list among the free blocks sing pointers Method 3: Segregated free list Different free lists for different size classes Method 4: Blocks sorted by size Can se a balanced tree (e.g. Red-Black tree) with pointers within each free block, and the length sed as a key 12

13 Segregated List (Seglist) Allocators Each size class of blocks has its own free list inf Often have separate classes for each small size For larger sizes: one class for each two-power size CS 153 Lectre 26 Dynamic Memory (2) 13

14 Seglist Allocator Given an array of free lists, each one for some size class To allocate a block of size n: Search appropriate free list for block of size m > n If an appropriate block is fond:» Split block and place fragment on appropriate list (optional) If no block is fond, try next larger class Repeat ntil block is fond If no block is fond: Reqest additional heap memory from OS (sing sbrk()) Allocate block of n bytes from this new memory Place remainder as a single free block in largest size class. CS 153 Lectre 26 Dynamic Memory (2) 14

15 Seglist Allocator (cont.) To free a block: Coalesce and place on appropriate list (optional) Advantages of seglist allocators Higher throghpt» log time for power-of-two size classes Better memory tilization» First-fit search of segregated free list approximates a best-fit search of entire heap.» Extreme case: Giving each block its own size class is eqivalent to best-fit. CS 153 Lectre 26 Dynamic Memory (2) 15

16 More Info on Allocators D. Knth, The Art of Compter Programming, 2 nd edition, Addison Wesley, 1973 The classic reference on dynamic storage allocation Wilson et al, Dynamic Storage Allocation: A Srvey and Critical Review, Proc Int l Workshop on Memory Management, Kinross, Scotland, Sept, Comprehensive srvey Available from CS:APP stdent site (csapp.cs.cm.ed) CS 153 Lectre 26 Dynamic Memory (2) 16

17 Implicit Memory Management: Garbage Collection Garbage collection: atomatic reclamation of heapallocated storage application never has to free void foo() { int *p = malloc(128); retrn; /* p block is now garbage */ } Common in fnctional langages, scripting langages, and modern object oriented langages: Lisp, ML, Java, Perl, Mathematica Variants ( conservative garbage collectors) exist for C and C++ However, cannot necessarily collect all garbage CS 153 Lectre 26 Dynamic Memory (2) 17

18 Garbage Collection How does the memory manager know when memory can be freed? In general we cannot know what is going to be sed in the ftre since it depends on conditionals Bt we can tell that certain blocks cannot be sed if there are no pointers to them Mst make certain assmptions abot pointers Memory manager can distingish pointers from non-pointers All pointers point to the start of a block Cannot hide pointers (e.g., by coercing them to an int, and then back again) CS 153 Lectre 26 Dynamic Memory (2) 18

19 Classical GC Algorithms Mark-and-sweep collection (McCarthy, 1960) Does not move blocks (nless yo also compact ) Reference conting (Collins, 1960) Does not move blocks (not discssed) Copying collection (Minsky, 1963) Moves blocks (not discssed) Generational Collectors (Lieberman and Hewitt, 1983) Collection based on lifetimes» Most allocations become garbage very soon» So focs reclamation work on zones of memory recently allocated For more information Jones and Lin, Garbage Collection: Algorithms for Atomatic Dynamic Memory, John Wiley & Sons,

20 Memory as a Graph We view memory as a directed graph Each block is a node in the graph Each pointer is an edge in the graph Locations not in the heap that contain pointers into the heap are called root nodes (e.g. registers, locations on the stack, global variables) Root nodes Heap nodes reachable Not-reachable (garbage) A node (block) is reachable if there is a path from any root to that node. Non-reachable nodes are garbage (cannot be needed by the application) 20

21 Mark and Sweep Collecting Can bild on top of malloc/free package Allocate sing malloc ntil yo rn ot of space When ot of space: Before mark Use extra mark bit in the head of each block Mark: Start at roots and set mark bit on each reachable block Sweep: Scan all blocks and free blocks that are not marked root Note: arrows here denote memory refs, not free list ptrs. After mark Mark bit set After sweep free free 21

22 Assmptions For a Simple Implementation Application new(n): retrns pointer to new block with all locations cleared read(b,i): read location i of block b into register write(b,i,v): write v into location i of block b Each block will have a header word addressed as b[-1], for a block b Used for different prposes in different collectors Instrctions sed by the Garbage Collector is_ptr(p): determines whether p is a pointer length(b): retrns the length of block b, not inclding the header get_roots(): retrns all the roots CS 153 Lectre 26 Dynamic Memory (2) 22

23 Mark and Sweep (cont.) Mark sing depth-first traversal of the memory graph ptr mark(ptr p) { if (!is_ptr(p)) retrn; // do nothing if not pointer if (markbitset(p)) retrn; // check if already marked setmarkbit(p); // set the mark bit for (i=0; i < length(p); i++) // call mark on all words mark(p[i]); // in the block retrn; } Sweep sing lengths to find next block ptr sweep(ptr p, ptr end) { while (p < end) { if markbitset(p) clearmarkbit(); else if (allocatebitset(p)) free(p); p += length(p); } 23

24 Conservative Mark & Sweep in C A conservative garbage collector for C programs is_ptr() determines if a word is a pointer by checking if it points to an allocated block of memory Bt, in C pointers can point to the middle of a block So how to find the beginning of the block? Header ptr Can se a balanced binary tree to keep track of all allocated blocks (key is start-of-block) Balanced-tree pointers can be stored in header (se two additional words) Size Left Head Right Data Left: smaller addresses Right: larger addresses 24

25 Next time Final review CS 153 Lectre 26 Dynamic Memory (2) 25