[Tutorial] Processes -- Introduction and setpriority

[Tony]

Processes Revised

HoltSoft wrote:

A process declaration is much like a procedure declaration, but is activated by a fork statement rather than by a call. The fork statement starts concurrent (parallel) execution of the process while the statements following the fork continue to execute.

Such allows us to execute multiple lines of code in parallel rather then sequentially one after one. An important thing to realize is that commands take turns to execute, and two processes will not run nicely next to each other. Note that it is individual commands that take turns, not the entire line statement as one might assume at first.

Turing:

process foo     loop         put "foo", "foo", "foo"     end loop end foo process bar     loop         put "bar", "bar", "bar"     end loop end bar fork foo fork bar

What kind of output would you expect?



Not quite what we wanted, is it?
Reducing the number of commands per line decreases the chance of corruption, but a) that tells us to stop using chain functions, and b) glitches will still occur.

Turing:

put "foo"

Is two statement. First outputs string "foo", second outputs the newline character. Try it out.

Turing:

process foo     loop         put "foo"     end loop end foo process bar     loop         put "bar"     end loop end bar fork foo fork bar

delay (1) after each put fixes the late \n glitch, but a) it slows down the execution time as delays accumulate, and b) this simply stops working as we go back to our put "foo", "foo", "foo" example. Glitches reemerge even at delay(100) as processes simply have a lot of picks to make. Delay becomes just another command in the mix.

Although processes come across looking quite badly at this point, it is only because they are unmanaged.

setpriority ( p : nat )

HoltSoft wrote:

The setpriority procedure is used to set the priority of a process in a concurrent program. This priority cannot be counted on to guarantee critical access to shared variables. A smaller value of p means increased speed. The argument to setpriority may be limited to the range 0 to 2**15 - 1.

That is not an entirely accurate description. Smaller value is not an increase in speed, it is an order the process takes in the query.

By default everything is set to a priority of 1000. You can find out the priority of any given function using getpriority (returning function's nat priority value). When processes start piling up in a query, by default one to chosen at random. You would need to tell the processes the order you want them to take. setpriority could (and should) be used as triggers for executing processes. When a block of code needs to be executed together (such as our "foo", "foo", "foo" output without breaking up), it should be wrapped around in a priority shift.

Turing:

process foo     loop         setpriority (999) %increase priority         put "foo", "foo", "foo"         setpriority (1000) %return to normal     end loop end foo process bar     loop         put "bar"     end loop end bar fork bar fork foo

Since put "foo", "foo", "foo" was given a higher priority (smaller number == higher priority) than put "bar"'s default 1000, the former was executed in full before it's priority was set back to 1000 (equivalent to that of bar) and other processes were given a chance to execute. Such technique is similar to Visual Basic's DoEvents functionality.

[jamonathin]

ver noice, very noice

[Cervantes]

Uh oh, I'm not getting the nice output.

I am still getting things like barfoofoofoo and black lines.

[Martin]

So it's uhh...just as random with different odds?

Processes still suck.