NotesDiagram
The Storage Hierarchy of a Computing system can be seen as:
Properties of the Hierarchy
The Price per bite (Price/byte) goes down as you move up in the hierarchy:
Speed
L1 access time is 1 or 2 CPU cycles L2 access time is in the 10s of CPU cycles
Main memory access time is in the 10s of nanoseconds (ns) (typically is 60 or more)
Sizes
L1 cache has a small amount of storage, typically 8 to 64K bytes
L2 is smaller, usually ranging between 256K and 2M bytes
Programmer’s View of the Storage Hierarchy
Programmers assume that the memory is just an array of bytes. They assume that it is addressable in a linear fashion. Memory cannot hold values between program runs.
Relocatable Programs
Because address spaces can be different each run time for the same program, it is important to make sure that the program works with memory changes.
These are programs that:
- Do not assume anything about where they will be loaded and do not assume anything about where the data is placed
- A linker can prepare them to run anywhere
Implementations
There’s a few ways you can implement relocation:
- Relocating linker
- Relocating loader
- Relocation register
Relocating Linker
The linker knows where the program will run and reads the program binary and translates addresses. If the location of the program needs to be changed, the program has to be relinked.
Relocating Loader
The same as a relocating linker, except that it resolves the references at load time. This is more flexible, though, because it does not need to know where the program will run beforehand. However, this has more overhead because it has to be done on every program run.
Relocation Register
We can just make all the addressing relative to a base register. The program can be relocated by just changing the base register.
Of course, because this is meant to also protect memory between programs, we have to ensure that the programs can’t access the base register themselves.
On modern systems, we need to have a