Multi-tasking OSes

[pyrojunkie]

So to put it frankly, true multitasking is only capable on multiprocessor systems? I still think multitasking should extend to systems that are capable of balancing the CPU cycles between tasks.

 

[rjain]

Originally posted by: pyrojunkie
So to put it frankly, true multitasking is only capable on multiprocessor systems? I still think multitasking should extend to systems that are capable of balancing the CPU cycles between tasks.

Well, it's even more than that. You'd have to have a separate processor for each task a user of the system will run at the same time. AFAIK, "true" multitasking is only a theoretical extreme, an idea that tells us what the goal is. Multi-tasking does extend to systems that task swtich, but those aren't "true" multi-tasking.

 

[uart]

I think you're getting into semantics a bit here by insisting that "true" multi-tasking requires more than one processor. Normal time slicing on Win9x, WinNT/2K/XP or *nix is multitasking as far as most peoples definition.

The key issues that some people here seem confused about are :

1. Having a CPU executing multiple instructions per cycle (whether by "SIMD or by "superscalar" design or otherwise) is not equivalent to multitasking, is not necessary for multitasking and does not imply multitasking.

2. Multitasking by time slicing is multitasking. It's not concurrent multiprocessing but it's multitasking never the less. The feeling of true concurrency is only an illusion with a single CPU but it can be true multitasking just the same. (Also the "illusion" of concurrency can be quite good due to the high speed of modern cpu's, the fine quantization of the time slices and hopefully the efficiency of prioritization of a good OS).

BTW rjain, I think you made a really good point way back a few posts when you made the distinction between "multi-tasking" and "multi-user". Two completely different things of-course, but I get the feeling that is was precisely these two things that derekblankmccoy was confusing when he made what otherwise appeared to be a very mis-informed post to start this thread.

 

[sao123]

Wrong. You can run OpenGL (GLX) Quake just fine remotely.

Using Sun Solaris 6 running on UltraSparc Machine, with X-client XWin32 for X86 Version 3 (last unix system setup i used, OpenGl didnt work)

Do not have access to newer version to test.

 

[rjain]

As if XWin32 would support that. XFree86 and IRIX support it fine. Dunno about Sun's X. But the existence of XWin32 (or any other X server, for that matter) shows that remote graphics is perfectly possible.