Main Memory

Main Memory

1. Background

   • A program must be loaded into memory to be executed
   • CPU have direct access only to registers and main memory
   • What does memory unit see?
     • Stream of
       • Address + read request
       • Address + write request
   • Time to access
     • Registers
       • ~1 clock cycle
     • Main Memory
       • Multiple clock cycles
         • May generate stalls
   • Cache
     • Managed by hardware
     • Memory between CPU and main memory
     • Faster than main memory, slower than registers
   • Protection of memory
     • It ensures that each program only accesses segments of main memory that are allowed to them.

2. Memory Protection

   • Mechanism to make sure that each process will access only the segments of memory to which they are allowed
     • Pair of registers that define the logical address space of a process: BASE and LIMIT
   • Algorithm
     • OBS.: loading base and limit register requires high privilege
     • CPU checks if the memory access generated in user space is valid.

3. Address binding

   • Source code
     • Define symbols
       • Ex.: name of a variable , name of a function
     • Accessible at compile time
   • Compiled code
     • Bind symbols from source code to Relocatable addresses
       • Ex.: N bytes from the beginning of this module
     • Accessible at load time
   • Linked and loaded code
     • Bind relocatable addresses to absolute logical (?) addresses
       • Ex. : <base_address> + N
     • Accessible at execution time

4. Address space

   • Physical
     • Seen by the memory unit
   • Logical (virtual address)
     • Seen by the CPU
   • Memory Management Unit
     • Translates logical address, from CPU, to physical address for the physical memory
     • Ex.
     • Not all systems have MMU. Less complex systems, like embedded systems, do not have MMU.

5. Dynamic Loading

   • Load the routines of the program only when they are called
   • Routines on disk in relocatable load format

6. Linking

   • Dynamic
     • Linking happens in execution time
       • Stub
         • Small piece of code used to locate the appropriate memory-resident library routine. It replaces itself with the address of the routine and executes the routine.
   • Static
     • system libs + program code = binary program image

7. Partitions

   • Operating Systems maintain information about: (a) - allocated partitions and (b) - free partitions.
   • Dynamic Storage-allocation problem: satisty a request of size N from a list of free holes
     • First-fit
       • Allocate the first hole that is big enough
     • Best-fit
       • Allocate the smallest hole the is big enough
     • Worst-fit
       • Allocate the largest hole.
     • The first-fit and best-fit strategies are usually better in terms of speed and storage than worst-fit
   • Fragmentation
     • External
       • Total memory is sufficient to satisfy a request but it is not contiguous.
     • Internal
       • Allocated memory is larger than what is actually used.

8. Paging

   • It is a technique in which the physical memory is divided into fixed-size blocks (frames) of 512 B to 2 MB and logical memory into fixed-size blocks (pages).
     • Make it possible for the program to be loaded even if no contiguous physical memory space is available but the total space is available.
       • Ex. A program of size N pages needs N free frames to be loaded.
   • Address Translation
     • Page Table
       • Translates logical addresses to physical addresses.
       • It may is stored in main memory, but it is inefficient. It is better to be assisted by a cache (TLB).
       • Each process will have its own page table
         • It allows using random addresses to make it harder to find the final physical address, something good for security.
       • Implementation
         • Registers
           • Page-table Base Register
             • Points to the table
           • Page-table length register
             • Size of the table
         • Cache
           • Translation Look-aside buffer (TLB) or (Associative Memory)
             • TLB entry
               • Uniquely identify each process to provide address space protection to that process
             • Greater TLBs are better because they avoid miss.
       • Structure
         • Strategies
           • Hierarchical Paging
             • The idea consists of using multiple levels of page table.
               • OBS.: Caches should be used to avoid multiple main memory accesses
             • Ex. 1
               • Ex. 2:
           • Hashed Page Table
           • Inverted Page Tables
     • Logical Address
       • Page Number
         • It is an index for a base physical address of each page in page table.
       • Page Offset
         • With the base address, it defines the physical memory
     • Ex. 1
       • Ex. 2
         • Ex. 3 (with TLB)

9. Effective Access Time (EAT)

   • Hit Ration
     • % of times that a page number was found in TLB.
   • ETA = <Hit Ratio> x <Mean time to access memory> + <Miss Ratio> x 2 x <Mean time to access memory>
     • OBS.: When a miss happens, it is necessary to access the memory twice. One to get information from page table and other to read/write from the memory

10. Protection

    • A protection bit is associated with each frame, indicating read/write or other kind of operation is allowed. A violation results in a trap to the kernel.
    • Valid-invalid bit
      • VALID: Indicate that the associated page is in the process' logical address space
      • INVALID: indicate that the associated page is not in the process' logical address space

11. Swapping

    • It is the process to move a process/page from main memory to a secondary memory temporarily.
    • Ex. of process swapping:

Images and files in the map