Virtual Memory
- CPU accesses main memory by generating a virtual address which is converted to the appropriate physical address before being sent to the memory.
- address translation: convert the virtual address to an appropriate physical address
- Virtual memory is partitioned into fixed-sized blocks called virtual pages. Similarly , physical memory is partitioned into physical pages.
- The set of virtual pages is partitioned into three disjoint subset:
unallocated: Pages that have not yet been allocatedcached: allocated pages that are currently cached in physical memoryuncached: allocated pages that are not cached in physical memory
- Page table
- Page table maps virtual pages to physical pages
- Operating system provides a separate page table for each process
- Every process on a given linux system starts at virtual address 0x400000(64bit)
- Address translation is a mapping between the elements of an N-element virtual address space and an M-element physical address space
- A control register in the CPU points to the current page table
- The n-bit virtual address has two components: a
p-bitvirtual page offset and ann-pbit virtual page number(VPN). The MMU uses the VPN to select the appropriate PTE. - The corresponding physical address is the concatenation of the physical page number from the page table entry and the VPO from the virtual address.
- Translation lookaside buffer (TLB): a small virtually addressed cache where each line holds a block consisting a single PTE.
- Important capabilities for virtual memory system:
- automatically caches recently used contents of virtual address space stored on disk in main memory
- simplifies memory management
- simplifies memory protection by incorporating protection bits into every page table entry
- most programs rely on a dynamic memory allocator such as malloc, which manages memory in an area of the virtual address space called the heap.