cache
Download as PPT, PDF1 like859 views
This document discusses memory management techniques in computer architecture, including: 1) Memory is divided into partitions for the operating system and active processes in a uni-program system, while in a multi-program system memory is further subdivided and shared among active processes. 2) Swapping allows processes to exceed main memory size by storing inactive processes on disk and swapping them back into memory as needed, reducing idle CPU time from I/O waits. 3) Paging maps processes and memory into uniform-sized pages and page frames, using a page table to track mappings and allowing non-contiguous allocation of pages to processes in memory.
cache
- 1. Computer Architecture Topic: Memory Management SATHISH.V VELAMMAL INSTITUTE OF TECHNOLOGY ECE 2010-2014 CSCI 4717 – Computer Architecture Memory Management – Page 1 of 49
- 2. Memory Management • Uni-program – memory split into two parts – One for Operating System (monitor) – One for currently executing program • Multi-program – Non-O/S part is sub-divided and shared among active processes • Remember segment registers in the 8086 architecture – Hardware designed to meet needs of O/S – Base Address = segment address CSCI 4717 – Computer Architecture Memory Management – Page 2 of 49
- 3. Swapping • Problem: I/O (Printing, Network, Keyboard, etc.) is so slow compared with CPU that even in multi-programming system, CPU can be idle most of the time • Solutions: – Increase main memory • Expensive • Programmers will eventually use all of this memory for a single process – Swapping CSCI 4717 – Computer Architecture Memory Management – Page 3 of 49
- 4. What is Swapping? • Long term queue of processes stored on disk • Processes “swapped” in as space becomes available • As a process completes it is moved out of main memory • If none of the processes in memory are ready (i.e. all I/O blocked) – Swap out a blocked process to intermediate queue – Swap in a ready process or a new process • But swapping is an I/O process! – It could make the situation worse – Disk I/O is typically fastest of all, so it still is an improvement CSCI 4717 – Computer Architecture Memory Management – Page 4 of 49
- 5. Partitioning • Splitting memory into sections to allocate to processes (including Operating System) • Two types – Fixed-sized partitions – Variable-sized partitions CSCI 4717 – Computer Architecture Memory Management – Page 5 of 49
- 6. Fixed-Sized Partitions (continued) • Equal size or Unequal size partitions • Process is fitted into smallest hole that will take it (best fit) • Some wasted memory due to each block having a hole of unused memory at the end of its partition • Leads to variable sized partitions CSCI 4717 – Computer Architecture Memory Management – Page 6 of 49
- 7. Fixedsized partitions CSCI 4717 – Computer Architecture Memory Management – Page 7 of 49
- 8. Variable-Sized Partitions • Allocate exactly the required memory to a process • This leads to a hole at the end of memory, too small to use – Only one small hole - less waste • When all processes are blocked, swap out a process and bring in another • New process may be smaller than swapped out process • Reloaded process not likely to return to same place in memory it started in • Another hole • Eventually have lots of holes (fragmentation) CSCI 4717 – Computer Architecture Memory Management – Page 8 of 49
- 9. Variable-Sized Partitions CSCI 4717 – Computer Architecture Memory Management – Page 9 of 49
- 10. Solutions to Holes in VariableSized Partitions • Coalesce - Join adjacent holes into a single large hole • Compaction - From time to time go through memory and move all holes into one free block (c.f. disk de-fragmentation) CSCI 4717 – Computer Architecture Memory Management – Page 10 of 49
- 11. Overlays • In the early days of computing, programmers had a small amount of memory to squeeze programs into • First "stabs" at memory management were overlays: – Programmer divided single application into smaller independent programs called overlays – When program first loaded, load first overlay into memory – When new overlay was required, next overlay is read from drive and loaded in place of previous one CSCI 4717 – Computer Architecture Memory Management – Page 11 of 49
- 12. Relocation • No guarantee that process will load into the same place in memory • Instructions contain addresses – Locations of data – Addresses for instructions (branching) • Logical address – relative to beginning of program • Physical address – actual location in memory (this time) • Base Address – start of program or block of data • Automatic conversion using base address CSCI 4717 – Computer Architecture Memory Management – Page 12 of 49
- 13. Paging (continued) • Split memory into equal sized, small chunks -page frames • Split programs (processes) into equal sized small chunks – pages • Allocate the required number page frames to a process • Operating System maintains list of free frames • A process does not require contiguous page frames CSCI 4717 – Computer Architecture Memory Management – Page 13 of 49
- 14. Paging (continued) • Use page table to keep track of how the process is distributed through the pages in memory • Now addressing becomes page number:relative address within page which is mapped to frame number:relative address within frame. CSCI 4717 – Computer Architecture Memory Management – Page 14 of 49
- 15. Paging (continued) CSCI 4717 – Computer Architecture Memory Management – Page 15 of 49
- 16. Paging Example – Before Free frame list 13 14 15 18 13 14 15 20 16 Process A Page 0 Page 1 Page 2 Page 3 In use 17 In use 18 19 20 21 CSCI 4717 – Computer Architecture In use In use Memory Management – Page 16 of 49
- 17. Paging Example – After Free frame list 13 Process A Page 0 13 Page 1 14 Page 2 15 Page 3 18 Page 1 of A 15 Process A page table Page 0 of A 14 20 Page 2 of A 16 In use 17 In use 18 Page 3 of A 19 In use 20 21 CSCI 4717 – Computer Architecture In use Memory Management – Page 17 of 49
- 18. Virtual Memory • Remember the Principle of Locality which states that “active” code tends to cluster together, and if a memory item is used once, it will most likely be used again. • Demand paging – Do not require all pages of a process in memory – Bring in pages as required CSCI 4717 – Computer Architecture Memory Management – Page 18 of 49
- 19. Page Fault in Virtual Memory • Required page is not in memory • Operating System must swap in required page • May need to swap out a page to make space • Select page to throw out based on recent history CSCI 4717 – Computer Architecture Memory Management – Page 19 of 49
- 20. Virtual Memory Bonus • We do not need all of a process in memory for it to run • We can swap in pages as required • So - we can now run processes that are bigger than total memory available! • Main memory is called real memory • User/programmer sees much bigger memory - virtual memory CSCI 4717 – Computer Architecture Memory Management – Page 20 of 49
- 21. Thrashing • Too many processes in too little memory • Operating System spends all its time swapping • Little or no real work is done • Disk light is on all the time • Solutions – Better page replacement algorithms – Reduce number of processes running – Get more memory CSCI 4717 – Computer Architecture Memory Management – Page 21 of 49
- 22. Page Table Structure • VAX architecture – each process may be allocated up to 231 = 2 GBytes of virtual memory broken in to 29=512 byte pages. • Therefore, each process may have a page table with 2(31-9)=222=4 Meg entries. • This uses a bunch of memory! CSCI 4717 – Computer Architecture Memory Management – Page 22 of 49
- 23. Pages of Page Table • Some processors solve this with a page directory that points to page tables, each table of which is limited to a page and treated as such • Another approach is the inverted page table structure CSCI 4717 – Computer Architecture Memory Management – Page 23 of 49
- 24. Inverted Page Table • Page tables based on logical (program's) address space can be huge • Alternatively, restrict page table entries to real memory, not virtual memory • Problem: – Simple page table says each line of table maps to logical page – Inverted Page Table need to have mapping algorithm because there isn't a one-to-one mapping of logical to virtual pages CSCI 4717 – Computer Architecture Memory Management – Page 24 of 49
- 25. Page of Page Table (continued) CSCI 4717 – Computer Architecture Memory Management – Page 25 of 49
- 26. Translation Lookaside Buffer • Every virtual memory reference causes two physical memory access • Fetch page table entry • Fetch data • Use special cache for page table – TLB CSCI 4717 – Computer Architecture Memory Management – Page 26 of 49
- 27. Translation Lookaside Buffer (continued) CSCI 4717 – Computer Architecture Memory Management – Page 27 of 49
- 28. Translation Lookaside Buffer (continued) • Complexity! Virtual address translated to a physical address • Reference to page table – might be in TLB, main memory, or disk • Referenced word may be in cache, main memory, or disk • If referenced word is on disk, it must be copied to main memory • If in main memory or on disk, block must be loaded to cache and cache table must be updated CSCI 4717 – Computer Architecture Memory Management – Page 28 of 49
- 29. TLB and Cache Operation CSCI 4717 – Computer Architecture Memory Management – Page 29 of 49
- 30. Multi-Level Page Tables Source: Rusling, D., "Linux Page Tables," The Linux Knowledge Base and Tutorial, Online: http://www.linux-tutorial.info/modules.php?name=MContent&pageid=307 CSCI 4717 – Computer Architecture Memory Management – Page 30 of 49
- 31. Segmentation • Paging is not (usually) visible to the programmer • Segmentation is visible to the programmer • Usually different segments allocated to program and data • There may be a number of program and data segments • Segmentation partitions memory CSCI 4717 – Computer Architecture Memory Management – Page 31 of 49
- 32. Advantages of Segmentation • Simplifies handling of growing data structures – O/S will expand or contract the segment as needed • Allows programs to be altered and recompiled independently, without re-linking and re-loading • Lends itself to sharing among processes • Lends itself to protection since O/S can specify certain privileges on a segment-by-segment basis • Some systems combine segmentation with paging CSCI 4717 – Computer Architecture Memory Management – Page 32 of 49
- 33. In-Class Discussion • The TLB is basically a cache for page tables. A TLB "miss" is a request for a page that isn't in the TLB. Name some ways that we can reduce the chances of a TLB miss. • Using paging with N processes and a page size of P, what is the most memory that is wasted? • What problem is caused by small pages? • What problem is caused by large pages? CSCI 4717 – Computer Architecture Memory Management – Page 33 of 49
- 34. Recursion • Many complex algorithmic functions can be broken into a repetitive application of a simple algorithm. • The typical recursion function begins with an initial value of n which is decremented with each recursive call until the last call reaches a terminal value of n. • A recursive function contains a call to itself. • "Definition of recursion: See recursion" CSCI 4717 – Computer Architecture Memory Management – Page 34 of 49
- 35. Recursion – Factorial • Non-Recursive Function: int factorial(int n) { int return_val = 1; for (int i = 1; i <= n; i++) return_val = return_val * i; return return_val; } • Recursive Function: int rfactorial(int n) { if ((n == 1) || (n == 0)) return (1); else return (n*rfactorial(n - 1)); } CSCI 4717 – Computer Architecture Memory Management – Page 35 of 49
- 36. Recursion – Fibonacci Numbers "f(i) = f(i–1) + f(i–2)" • Non-Recursive Function: int fibonacci(int n) { int fibval_i = 1; int fibval_i_minus_1 = 0; int fibval_i_minus_2 = 0; if ((n == 0)||(n == 1)) return n; else { for (int i = 2; i <= n; i++) { fibval_i_minus_2 = fibval_i_minus_1; fibval_i_minus_1 = fibval_i; fibval_i = fibval_i_minus_1 + fibval_i_minus_2; } } return fibval_i; } CSCI 4717 – Computer Architecture Memory Management – Page 36 of 49
- 37. Recursion – Fibonacci Numbers (continued) • Recursive Function: int rfibonacci(int n) { if ((n == 0)||(n == 1)) return n; else return rfibonacci(n - 1) + rfibonacci(n - 2); } CSCI 4717 – Computer Architecture Memory Management – Page 37 of 49
- 38. Comparing Recursive and NonRecursive Functions • Non-recursive function has more variables. Where does recursive function store values. • Non-recursive function has more code recursive requires less code and therefore less memory. CSCI 4717 – Computer Architecture Memory Management – Page 38 of 49
- 39. In-Class Exercise • In groups, discuss how recursion might affect an operating system • Compare & contrast iterative vs. recursion algorithms in terms of growth/memory usage CSCI 4717 – Computer Architecture Memory Management – Page 39 of 49
- 40. Pentium II • Hardware for segmentation and paging • Unsegmented unpaged – virtual address = physical address – Low complexity – High performance • Unsegmented paged – Memory viewed as paged linear address space – Protection and management via paging – Berkeley UNIX CSCI 4717 – Computer Architecture Memory Management – Page 40 of 49
- 41. Pentium II (continued) • Segmented unpaged – Collection of local address spaces – Protection to single byte level – Translation table needed is on chip when segment is in memory • Segmented paged – Segmentation used to define logical memory partitions subject to access control – Paging manages allocation of memory within partitions – Unix System V CSCI 4717 – Computer Architecture Memory Management – Page 41 of 49
- 42. Pentium II Segmentation • Each virtual address is 16-bit segment and 32bit offset • 2 bits of segment are protection mechanism • 14 bits specify segment • Unsegmented virtual memory 232 = 4Gbytes • Segmented 246=64 terabytes – Can be larger – depends on which process is active – Half (8K segments of 4Gbytes) is global – Half is local and distinct for each process CSCI 4717 – Computer Architecture Memory Management – Page 42 of 49
- 43. Pentium II Protection Protection bits give 4 levels of privilege • 0 most protected, 3 least • Use of levels software dependent • Usually level 3 is for applications, level 1 for O/S and level 0 for kernel (level 2 not used) • Level 2 may be used for apps that have internal security, e.g., database • Some instructions only work in level 0 CSCI 4717 – Computer Architecture Memory Management – Page 43 of 49
- 44. Pentium II Paging • Segmentation may be disabled in which case linear address space is used • Two level page table lookup • First, page directory – 1024 entries max – Splits 4G linear memory into 1024 page groups of 4Mbyte – Each page table has 1024 entries corresponding to 4Kbyte pages – Can use one page directory for all processes, one per process or mixture Page directory for current process always in memory • Use TLB holding 32 page table entries • Two page sizes available 4k or 4M CSCI 4717 – Computer Architecture Memory Management – Page 44 of 49
- 45. Pentium Virtual Address Breakdown CSCI 4717 – Computer Architecture Memory Management – Page 45 of 49
- 46. Pentium Segment/Paging Operation CSCI 4717 – Computer Architecture Memory Management – Page 46 of 49
- 47. PowerPC Memory Management Hardware • 32 bit – paging with simple segmentation – 64 bit paging with more powerful segmentation • Or, both do block address translation – Map 4 large blocks of instructions & 4 of memory to bypass paging – e.g. OS tables or graphics frame buffers • 32 bit effective address – 12 bit byte selector 4kbyte pages – 16 bit page id 64k pages per segment – 4 bits indicate one of 16 segment registers Segment registers under OS control CSCI 4717 – Computer Architecture Memory Management – Page 47 of 49
- 48. PowerPC 32-bit Memory Management Formats CSCI 4717 – Computer Architecture Memory Management – Page 48 of 49
- 49. PowerPC 32-bit Address Translation CSCI 4717 – Computer Architecture Memory Management – Page 49 of 49