OSDev.org

Anyone using operating system as the research topic in the postgraduate program? If yes, may I know what is your topic?

Yes, me)

what topic you are working on?

Yes. Well, sort of.

If you want a topic, I think something that would benefit the world greatly is a dissertation on the current state of affairs wrt OS design and research, i.e. what has been developed since the beginning of time, the aim of each OS and whether that OS reached its goals. Current research topics and the influencing phenomenon would round it off nicely.

Does sound like an awful lot of work, but then again, who said Post grad is supposed to be little work?

My topic is something like "(almost) single-tasking operating system for efficient computation on modern multi-core and multiprocessor systems"
It's about minimizing task switching using event-driven scheduling, etc...

JamesM wrote:Yes. Well, sort of.

JamesM doesn't want to say more about his super-secret project

Hi all, I want to implement a network-enable-libc in my os. Everything function in my libc may call a subroutine on other machine.

That's really more of an IPC thing than a libc thing. Also, RPC within a machine is much easier than RPC across multiple machines - you need a network stack (and probably a TCP implementation) to do that first. I'm pretty sure RPC has been done before too, so it's not much of a research topic.

hi , my idea is to add network function to libc, so all porogram without recompile can enjoy the performance benefit of cluster. The need to rewrite the libc and the program loader and part of the kernel.

I need to find a research topic to apply a research postgradeuate in hong kong. So the topic cant be too kindergarten, athought we dont have much great computer scientist in the university.

Thanks

First of all you can't just add any functions to libc...
At least they will not by standartized, IMHO.
You'd better implement another library that provides such interfaces.
And, keeping in mind that we are on osdev,
You may include this functions to your kernel's API...

But there are some similar widely used libraries.. (Like MPI, etc...)
Anyway, you have to think about what benefits people can extract from jour project.

Good luck, and don't develop a bicycle)))
P.S. At least after developing a bicycle, develop an improved one

Libc's functions are mostly small and compact, so I would imagine the overhead of calling many functions would outweigh the overhead of doing everything locally.

On the other hand, something like OpenGL, where the textures, vertex/index buffers, and shaders get uploaded to a server, and render calls batch together and operate remotely returning only the final rendered image seems much more useful.

because I want to make every program can have performance benefit in cluster environment. I need a topic to apply a research postgraduate degree. So the topic need to sounds good and not kindergarten.
thanks
from Peter

MessiahAndrw wrote:On the other hand, something like OpenGL, where the textures, vertex/index buffers, and shaders get uploaded to a server, and render calls batch together and operate remotely returning only the final rendered image seems much more useful.

Except that this is about huge amounts of data, many times per second. And in-order, as well. Doesn't seem too efficient to me doing it over a network...

JAL

jal wrote:

MessiahAndrw wrote:On the other hand, something like OpenGL, where the textures, vertex/index buffers, and shaders get uploaded to a server, and render calls batch together and operate remotely returning only the final rendered image seems much more useful.

Except that this is about huge amounts of data, many times per second. And in-order, as well. Doesn't seem too efficient to me doing it over a network...

Whilst that may be true, OpenGL's API is designed with a client/server concept in mind. That's why functions such as glFlush exist. It's kind of like TCP and the PUSH flag - the local client buffers and transmits where it deems necessary, and glFlush is used by programmers when they absolutely need some command to run.

A proper C library where every command can be run over a network cannot buffer commands (if it did, it'd have to break POSIX and add a way to flush said buffer, and therefore break compatibility with existing applications in some circumstances). The performance would also be horrible. OpenGL at least can wait until something needs to be actually drawn before it utilises the network