Greg Freemyer wrote:

On Tue, Feb 10, 2009 at 11:34 AM, Heinz Diehl <htd@fancy-poultry.org> wrote:

...

Hi,

just want to ask if there's a list what programs in opensuse 11.0/11.1 are multithreaded, meaning "can use multiple cpu cores" . Example: "make -j". Tried to find out by myself but wasn't successful. Can somebody point me to a side or source of information where I can read more (if it exists)?

As you research this you need to understand "multithreaded" does not equate to "can use multiple cpu cores".

Linux supports several threading models. I think only "Native POSIX Linux Threading" can leverage multiple cpus.

OTOH, lots of user space apps utilize multiple processes without the use of threading and thus they "can use multiple cpu cores".

The make command in your example I believe is an example of a app the "can use multiple cpu cores", but is not multithreaded.

Apache is another.

In general linux uses lots of small apps doing specific things. That is an ideal implementation that "can use multiple cpu cores".

Greg

Add to your list: -almost all java apps more sophisticated than "hello world" -all your various web and file browsers -almost anything with a gui will use a MVC architecture where different threads are responsible for different things Lots of multithreading going on, but not always lots of GOOD multithreading. Poorly written apps routinely synchronize threads badly (ever have a gui hang and become unresponsive? almost certianly blocked waiting for a shared resource). make -j is an example of multiprocessing rather than multithreading as make synchronizes among many different processes. In my experience it doesn't always work well anyway. Because linux is a multi processing/multithreaded system you can be sure that for cpu's or cores, the more the better up to a limit. And that limit is rising over time... no idea where it is now, but 20+ seems likely. wcn -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Tuesday February 10 2009, David Haller wrote:

Hello,

On Tue, 10 Feb 2009, Heinz Diehl wrote:

...

just want to ask if there's a list what programs in opensuse 11.0/11.1 are multithreaded, meaning "can use multiple cpu cores". Example: "make -j".

make -j just starts multiple programs.

Which is one way to be free of the need to explicitly code for concurrent execution using threads. If there are multiple processors / cores, then this approach will cause parallel execution of the processes that make initiates. Since make understands dependencies (that's what it's all about), it knows what tasks have no interdependencies and can thus be executed concurrently. So use of "make -j" will indeed exploit a multi-CPU / multi-core system.

Real multihreading programs link to a threading library. E.g.:

rpm -q --whatrequires libpthread.so.0

gives you a list of packages linking against the "classic" pthread library.

But that will not pick up programs such as make that do exploit multiple CPUs or cores. To wit: % rpm -q --whatprovides $(type -p make) make-3.81-128.9 % rpm -q --whatrequires libpthread.so.0 |egrep -i make [none]

HTH, -dnh

Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On 10.02.2009, Randall R Schulz wrote:

% rpm -q --whatrequires libpthread.so.0 |wc -l 410

# Library pthreads dependencies: %rpm -q --whatrequires libpthread.so.0 |egrep lib |wc -l 138

# Non-library pthreads dependencies: % rpm -q --whatrequires libpthread.so.0 |egrep -v lib |wc -l 272

This is my opensuse 11.0: liesel:~ # rpm -q --whatrequires libpthread.so.0 |wc -l 42 liesel:~ # rpm -q --whatrequires libpthread.so.0 |egrep lib |wc -l 16 liesel:~ # rpm -q --whatrequires libpthread.so.0 |egrep -v lib |wc -l 26 It seems my quadcore cpu has not enough workload to even exploit one core, and the other three are just there to consume (a lot of) power. I'll downgrade to an energy efficient 45W Athlon soon. Thanks to all who has answered! -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Tue, 2009-02-10 at 22:12 +0100, Heinz Diehl wrote:

On 10.02.2009, Randall R Schulz wrote:

...

% rpm -q --whatrequires libpthread.so.0 |wc -l 410

...

# Library pthreads dependencies: %rpm -q --whatrequires libpthread.so.0 |egrep lib |wc -l 138

...

# Non-library pthreads dependencies: % rpm -q --whatrequires libpthread.so.0 |egrep -v lib |wc -l 272

This is my opensuse 11.0:

liesel:~ # rpm -q --whatrequires libpthread.so.0 |wc -l 42

liesel:~ # rpm -q --whatrequires libpthread.so.0 |egrep lib |wc -l 16

liesel:~ # rpm -q --whatrequires libpthread.so.0 |egrep -v lib |wc -l 26

The numbers above and your comment below are really unrelated.

It seems my quadcore cpu has not enough workload to even exploit one core, and the other three are just there to consume (a lot of) power. I'll downgrade to an energy efficient 45W Athlon soon.

A threaded app is one where a single app can possibly run in different cores/CPUs, perhaps even concurrently. The presence of this library indicates that the app is perhaps creating threads directly, and at least considering this possible concurrency. For threading to have any effect, the OS must support multiple cores/CPUs. So threading builds on and requires multiprocessing. A multiprocessor system, on the other hand, is managed by the OS. It can put processes to run in different core/CPU. The applications themselves need know nothing about this. They do not need to do any threading or call anything in a thread library. So, vi could run in one core at the same time as some unrelated process runs in another. And vi need not do anything to allow this unrelated process to do so. No libpthread is needed. It is the OS that manages this. If vi wanted to get fancy and do more than one thing at the same time, it would involve libpthread. So, the presence of libpthread does not indicate how busy your CPU cores are. Try 'ps -efLmP' to list all threads. Those listed with the same PID are all threads of that PID. The PSR column tells which CPU the process is assigned to. I think that is the column you are interested in. You will note that processes with one thread still run in different CPUs. Hope this helps. top also has some thread viewing options. I have a compute-intensive multithreaded app that tracks the time spent in each thread. On a 2 CPU processor, I am always happy to see it report that in 30 minutes it used almost 60 minutes of CPU time. Luckily, this app can be divided into two threads. Adding more CPUs would probably not help it much. So, YMWV. Of course, if the system is not busy, there is always the address below. Anyone finding themselves in possession of a needlessly fast computer can ship them there. No questions asked. :) -- Roger Oberholtzer OPQ Systems / Ramböll RST Ramböll Sverige AB Krukmakargatan 21 P.O. Box 17009 SE-104 62 Stockholm, Sweden Office: Int +46 8-615 60 20 Mobile: Int +46 70-815 1696 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Wednesday February 11 2009, Larry Stotler wrote:

On Tue, Feb 10, 2009 at 4:12 PM, Heinz Diehl <htd@fancy-poultry.org> wrote:

...

It seems my quadcore cpu has not enough workload to even exploit one core, and the other three are just there to consume (a lot of) power. I'll downgrade to an energy efficient 45W Athlon soon.

One thing to keep in mind is that almost all processors are idle most of the time in a desktop environment. ...

That's what SETI@home, Folding@home, BOINC in general and other distributed collaborative CPU-warmers are for...

...

Also, for things like doing movie re-encodes, I have found that you get better performance by running 1 thread per cpu core versus trying to run 1 encode over multiple cores because the system has to split the data up and put it back together. On a quad core system, you can get 4 movies done faster that way than you can doing 4 one at a time.

That's to be expected. Parallelizing an arbitrary computation can yield anything from no speed-up to near N:1 where N is the number of independent processors available. First of all, we must exclude from consideration purely single-threaded computations which can never exhibit speed-up 'cause they are, by definition, not parallelizable. (Note that, as mentioned earlier, Java programs will almost always exhibit some degree of multi-threaded behavior because at a minimum JVM garbage collection is handled by a dedicated thread. Presumably other languages or run-times with GC will do the same.) For other computations, the degree of benefit obtained by parallelizing the computation all comes down to a matter of how much the parallel threads of execution must coordinate with each other and how much they contend even when not interacting. The former comes when results from one thread are needed by another. The latter happens when multiple threads contend for a shared resource, most commonly main store (RAM). So parallelizing a computation with a great deal of interaction between the parallel threads will yield much less than N:1 speed-up. Parallelizing a computation with little or no such interaction and a very high degree of locality (meaning a high ratio of instructions executed to cache misses) can exhibit nearly N:1 speed-up.

Just my 2 cents.

Randall Schulz -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

On Wed, Feb 11, 2009 at 1:12 PM, M Harris <harrismh777@earthlink.net> wrote:

It is my experience that a well tuned system with one processor (and one core) runs more efficiently than an smp system with multiple cores that has not been optimized. The entire system must be optimized for multi-cores, and the application (whatever that means) must be written specifically with smp in mind. Throwing multiple cores at an existing system (including apps) without tuning and optimization can actually result in "slower" performance (depending)... again, the truth of the paragraph at the top of the page reigns supreme.

However, I have found that, at least under linux, that an SMP system can be faster than a single core system that's 50% faster. However, it just depends on what you are doing. If you are watching a tv show and surfing the web, then dual core makes much more sense because the system is more responsive. So, I don't recommend any single core systems anymore. Even the Celerons are almost all DC. Only the Sempron is single core anymore. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

I think that speaks to the fact that there's a good bit of latent parallelizability at the overall system operation level for what constitutes typical mixes of activities, both those overtly and explicitly commanded by the human user and the plethora of behind-the-scenes processes that interact and coorindate to provide the user's overall computing experience.

Linux will show good SMP. With win2k, I only saw about a 10% increase over 1 proc. With XP it's a little better. Haven't really played with Vista or 7 yet. I hear the BSDs(free I think) have introduced some really kickin SMP code that's supposed to be wicked fast. Haven't had time to play with that either tho... -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org