Monday, September 6, 1993

Today:
	Sequential processes
		terminology and introduction to implementation
------------------------------------------------------------------------
First some related terminology:

Multiprogramming: intuitively, more than one process (activity) can be
	ready to run.  CPU switches among them
Multiprocessor systems: multiple CPUs.  Generally MIMD.
	Symmetric: each processor runs identical copy of OS.  Copies
	communicate as necessary
		tightly coupled: share main memory
		loosely coupled: connected via communication links
	assymmetric: each processor has specific task (e.g.,
		master/slave; channels, etc.)
Multitasking (timesharing): multiprogramming with fast switching so it
	appears to user that all programs are running simultaneously.

Opposite terms:  multiprogramming and uniprogramming
	multiprocessor and uniprocessor
Orthogonal terms: multiprogramming and multiprocessor
------------------------------------------------------------------------
Process:
	First cut at definition:  sequential process (or process) is a
	program in execution.
Program---passive entity (static---code).  Process---active entity (dynamic)
but note that "sequential process" does not exactly correspond to
program in execution: one program can spawn multiple processes (fork);
multiple processes can implement a single program, either in sequence
(e.g., C compiler) or simultaneously as above.

Sequential process characteristics:
- sequential: at any point in time *at most* one instruction executed on
behalf of process
- formed from a running piece of code plus environment [all the things
it affects and all the things affected by it]
- from point of view of CPU, environment encoded in program counter
(next instruction to execute); process stack (return address, subroutine
paramenters, local data, temporary variables); global data section
(global variables)

execution stream: sequence of instructions performed by a process+environment

Aside: notion of process tied in strongly with decomposition
Aside: object and process can be related (e.g., ADT).  Process as
	logical construct.
------------------------------------------------------------------------
Process state (finite state diagram)/System's view of a process
States:
	Ready: The process is waiting to be assigned to a processor
	Running: Instructions are being executed
	Blocked (waiting): The process is waiting for some event to
		occur (e.g., I/O completion)
State transitions (operation: old state -> new state)
	new:	-> ready
	dispatch:ready -> running		invoked by OS
	timeslice over: running -> ready	invoked by OS
	exit: running ->
	block: running-> blocked		invoked by process
	wakeup: blocked -> ready		invoked by OS
---
In uniprocessor at most *one* process can be running at any instant
Many can be ready or blocked
Dispatcher (or scheduler) figures out which process is to be moved from
	ready to running states
Timer causes process to move from running to ready states when time
	slice (quantum) expires
Process requests block by for example invoking i/o system call
wakeup occurs when request satisfied (perhaps by interrupt handler)
------------------------------------------------------------------------
Process control block (PCB) also called process descriptor
- Represents process to operating system
- Located at fixed location (but perhaps in virtual memory)
	How many processes can be active at a given instant? Number of
	PCBs (SYSGEN parameter)
- Stores that information that varies from process to process
- OS designer specifies fields as appropriate (keeping memory use in
mind).  For example:
	- unique identifier for process
	- current state of process (new, ready, running, waiting, halted)
	- pointer to process' parent
	- pointer(s) to children processes if any
	- space to save/restore needed values
		- program counter
		- CPU registers
		- current addressing mode (user/supervisor)
	- CPU scheduling information (priority, scheduler data structures)
	- memory management information (e.g., limit register, page
		tables [for location of process' memory])
	- pointers to allocated resources and i/o status information
		(e.g., devices, list of open files, etc.)
	- accounting information (CPU time used, wall clock time used,
		time limits, account numbers, etc.)
- Additional information as required by configuration can also be
included in design, e.g., identity of processor process is running on if
a multiprocessor system.