Software devs still not doing their jobs right!

[aidanjt]

yes indeed, however practically every C/C++ compiler allows access to assembly routines to allow you to use SIMD instruction sets, and there is plenty of threading libraries around to deal with that issue as well. Library abstraction might be good in a sense that you don't balls up the system, but that is what debugging is for. The tools exist. Unfortunatly they just aren't being commonly used.

Note: thanks for the edit, I didn't intend to rant at users by starting this thread.

[Nein]

Muti-processing is defined when there are two or more cores doing processing, regardless of packaging.. Weither the work is done in a distributed (parallel) or symmetric (serial) mannor is completely irrelivent in the scope of mutli-threaded software applications.

No wonder why programmers can not do their job properly.

Ethernet is a serial signaling communication medium btw, not parallel.

It is a "Scalable Link Interface" which originated from Distributed Processing (Distributed Bandwidth). Unlike "Bus Interface" which is Symmetric Processing (Time-Share/Time-Division) design.

Bus Interface = one data stream at a time design, time division based, maximum data clock rate in one huge time slice is indicative of performance (Single-Tasking).... Hence their rating is the same as all Symmetric Processing designs which is the clock frequency.

Link Interface = multiple data streams design, bandwidth based hence usually rated by a designation standing for the bandwidth throughput rather than clock frequency for example T1, T2, T3, 10/100 Ethernet, PCI-Express - 1x 2x 8x 16x, etc...

[aidanjt]

If multi-processing architecture governs how a software application should be coded, give us one single example.

I think you're getting network transmission mediums confused with toptology, you can scale up a network with extra hosts, but you can't scale extra Cat5 cables to a NIC, Ethernet doesn't 'stream' data, a NIC fragments IP packets and sends them sequentally within Ethernet frames one at a time, thus making it serial. It makes the apearance of sending multipule 'streams' to the computer because it transmits squentually, but this is all hardware abstraction performed by drivers.
Ethernet does not make 2 transmissions at the same time, this is impossible as Ethernet transmits on a set eletrical voltage in binary, it does not trasmit on different frequencies, my telephone/adsl line does, but ethernet does not..

Just as bus signaling can only do one thing at a time.. even AMD's CPUs can only pass one bit at a time between end points.. you can't change the laws of physics.

[Nein]

If multi-processing architecture governs how a software application should be coded, give us one single example.

Do you know the difference of Symmetric Processing from Symmetric Multi-Processing? One used a single processor while the other used more than one.

Do you know the difference of Distributed Processing from Distributed Multi-Processing? One used a single processor while the other used more than one.

I think you're getting network transmission mediums confused with toptology, you can scale up a network with extra hosts, but you can't scale extra Cat5 cables to a NIC, Ethernet doesn't 'stream' data, a NIC fragments IP packets and sends them sequentally within Ethernet frames one at a time, thus making it serial. It makes the apearance of sending multipule 'streams' to the computer because it transmits squentually, but this is all hardware abstraction performed by drivers.
Ethernet does not make 2 transmissions at the same time, this is impossible as Ethernet transmits on a set eletrical frequency in binary, it does not trasmit on different frequencies, my telephone/adsl line does, but ethernet does not..

Please stop. You can't even comprehend your hubs and your switches clearly.

Just as bus signaling can only do one thing at a time.. even AMD's CPUs can only pass one bit at a time between end points.. you can't change the laws of physics.

All processing can be done in simple linear time domain within a period of time (Symmetric Processing - Single-Thread optimum), or that time duration period can be converted to a spectrum of channels (MAC = Media Access Channel) for processing within that time period (Distributed Processing - Multi-Threaded optimum).

You should learn some Calculus or if you did then please learn to use and to apply it properly.

[Moby-Dick]

Nein , I dont expect to have to warn people a 2nd time.

You are not impressing anyone with your posting style , particularly me.

stop it.


now.

[herulach]

This is a bit OT, but does anyone have any good links on writing software for multiprocessor systems? Im gonna be doing a fair bit of intensive simulation for my masters project, and if i can take advantage of my housemates X2 to get it done, then all the better

[aidanjt]

*sigh* that didn't answer the question, and MAC stands for Media Address Control, every NIC has an individual MAC code, its a hexidecmal number, it has nothing to do with transmission. I will no longer reply anymore incoherent and invalid statements.

PrivatePyle.. yes, you're absolutely correct, RAD is needed to prevent the cost of development from getting out of hand. But my point is this, imagine you're working on an application that works on huge ammounts of data, say a video editing program. You're not going to code it in Visual Basic are you?
No, of course not, you'd use C or C++, further more, most people who edit video have SMP systems, if you don't mutli-thread your application you wont use their other CPU cores. If you fail to use SIMD instructions scanning through video frames will be slow and tiresome. Furthermore, if you make your own video codec and don't use threading and SIMD instructions, encoding the video stream will take several hours longer.

Developers need to use the right tools for the job, not stick exclusively with one particular development enviroment they are acustomed with.

[aidanjt]

herulach - this gives you a basic overview of what you need to do:
http://www.cswl.com/whiteppr/white/multithreading.html

If you're going to use VC++/MFC to do your project search for CWinThread on MSDN.

[aidanjt]

hahaha, my deepest appologies.

[Nein]

Oh I see. Well I can't really argue the point that software people are keeping up with hardware. They aren't. In fact, it seems that the faster computers get the more inept programmers become, simply because they can. They are dealing with virtually unlimited storage and memory, and incredible processor and graphics card performance. They simply do not need to be as good as programmers used to be.

And just about everyone knows why servers use Intel more often than AMD. It's pretty much the same reason more people use Celeron than A64. Nothing complicated, just marketing.

The problem with idiot-proofing to make things easier and simpler is that it also made for even more idiots. All you can get usually are wannabe experts. Can't get many real programmers anymore, there are few of them sprinkled in a pile of wannabes.

The new X2 will stomp most Xeon dualies at almost everything. The dual-core Opterons are so much more powerful than anything Intel has to offer. Does it matter? No, because most IT Networking types know squat about hardware.

Ha ha.. I met quite a few IT Networking "experts" myself, it is funny that a graphic hardware guy like me had to tell them what "crossbar" was for nearly a year, even more hilarious with "Scalable Link Interface".

The whole Internet is a humongous Distributed Processing design, supporting point-2-point, non-uniform access, simultaneous multiple data streams from multiple sources to multiple destinations.

Where the hell did all the internet experts go? Do they wanted the internet to be a symmetric Processing design? All users should wait for acess in their own processing time slice duration? All servers running software which knew little differences of Single-tasking from Multi-Tasking.

[herulach]

herulach - this gives you a basic overview of what you need to do:
http://www.cswl.com/whiteppr/white/multithreading.html

If you're going to use VC++/MFC to do your project search for CWinThread on MSDN.

Itll be fortran probably ( i know, i know, but its physics, you cant get away from iit) But ill be doing it in VS .NET, so i can write the main bits in whatever i like, just fortran dlls

In the context of .NET and windows XP as a target system, do i need to do anything other than multithread to take advantage of more than one processor core?

[aidanjt]

In the context of .NET and windows XP as a target system, do i need to do anything other than multithread to take advantage of more than one processor core?

I don't have any fortran experience myself. But no, so long as the program impiments threads the kernel will take care of allocating threads between the two cores. You'll learn more as you go along testing code.

[Nein]

Itll be fortran probably ( i know, i know, but its physics, you cant get away from iit) But ill be doing it in VS .NET, so i can write the main bits in whatever i like, just fortran dlls

In the context of .NET and windows XP as a target system, do i need to do anything other than multithread to take advantage of more than one processor core?

Would you accept optimization advice from a hardware guy (not a programmer)?

[Moby-Dick]

Nein as you can't keep you ego to yourself I think you should take 24 hours to think about how not to irritate people so much.

[Moby-Dick]

Sorry I've had to close the thread folks. It just seems that inspite of warnings , people can't discuss things without making snide comments to each other.