Openmp thread divergence? - multithreading

The term thread divergence is used in CUDA; from my understanding it's a situation where different threads are assigned to do different tasks and this results in a big performance hit.
I was wondering, is there a similar penalty for doing this in openmp? For example, say I have a 6 core processor and a program with 6 threads. If I have a conditional that makes 3 threads perform a certain task, and then have the other three threads perform a completely different task, will there be a big performance hit? I guess in essence it's sort of using openmp to do MIMD.
Basically, I'm writing a program with openmp and CUDA. I want two threads to run a CUDA kernel while the other left over threads run C code. Thanks.

No, there is no performance hit for diverging threads using OpenMP. It is a problem in CUDA because of the way instructions are broadcast simultaneously to a set of cores. When an OpenMP thread targets a CPU core, each CPU core has its own independent set of instructions to follow, and it runs just like any other single-threaded program would.
You may see some of your cores being underutilized if you have synchronization barriers following thread divergence, because that would force faster threads to wait for the slower threads to catch up.

When speaking about CPU parallelism, there's no intrinsic performance hit from using a certain threading design pattern. Not at the theoretical level at least.
The only problem I see is that since the threads are doing different things which may have varying completion times, some of the threads may sit idle after finishing their work, waiting for the others to finish a longer task.

The term thread divergence in CUDA refers to the situation when not all threads of a bock evaluate a conditional with the same outcome. Such threads are said to diverge. If diverging threads are in the same warp then such threads may perform work serially which leads to performance loss.
I am not sure that OpenMP has the same issue, though. When different threads perform different work then load balancing may be used by the runtime perhaps, but it doesn't lead to work serialization necessarily.

there is no this kind of problem in openmp because every openmp thread has its own PC.

Related

Performance of multi-threading exceeding cores

If I have a process that starts X amount of threads, will there ever be a performance gain having X higher than the number of CPU cores (assuming all the threads are working synchronously without async calls to storage/network)?
E.G. If I have a two cores CPU, will I just slow down the application starting 3+ constantly working threads?
It really depends on what your code does. it is too broad.
Having more threads than cores might speed up the program for example if some of the threads sleep or try to block on a lock. in this case, the OS scheduler can wake different thread and that thread will work while the other thread is sleeping.
Having more threads than the number of cores may also decrease the program execution time because the OS scheduler has to do more work to switch between the threads execution and that scheduling might be a heavy operation.
As always, benchmarking your application with different amount of threads is the best way to achieve maximum performance. there are also algorithms (like Hill-Climbing) which may help the application fine tune the best number of threads on runtime.
It is possible that such a thing happens.
Both Intel and AMD currently implement forms of SMT in their CPUs. This means that, in general, one single thread of execution may not be able to exploit 100% of the computing resources.
This happens because modern CPUs execute instructions in multiple pipelined steps, so that the clock frequency can be increased (less stuff gets done in every cycle, so you can do more cycles). The downside of this approach is that, if you have two consecutive instructions A and B, with the latter depending on the result of the former, you may have to wait some clock cycles without doing anything, just waiting for instruction A to complete. So, they came up with SMT, which allows the CPU to interleave instructions from two different threads/processes on the same pipeline, in order to fill such gaps.
Note: it is not exactly like this, CPUs don't just wait. They try to guess the result of the first operation and execute the second assuming that result. If their guess is wrong, they cancel the pending instructions and start over. Also, they have some feedback circuits that allow tighter execution of interdependent instructions. And nowadays branch predictors are surprisingly good. Things get better for the pipeline if you can just fill gaps with instructions from some other process, rather than going with a guess, but this potentially halves the amount of cache each executing thread can use.
It makes sense to run more threads if your threads make read/write/send/recv syscalls or similar, or sleep on locks, etc.
If your threads are pure computation threads, adding more of them will slow down system because of context switches.
If you still need more threads by design, you might want to look into the cooperative multitasking. Both Windows and Linux have API for that and that will work faster than the context switches. In Windows it called fibers:
https://msdn.microsoft.com/en-us/library/windows/desktop/ms682661(v=vs.85).aspx
In Linux it is a set of functions make/get/swapcontext():
http://man7.org/linux/man-pages/man3/makecontext.3.html
This question: Optimal number of threads per core might help you.
In the thread I wrote an answer describing a scenario when having higher number of threads than the available number of cores boosts performance.

How to do the same calculations faster on 4-core CPU: 4 threads or 50 threads?

Lets assume we have fixed amount of calculation work, without blocking, sleeping, i/o-waiting. The work can be parallelized very well - it consists of 100M small and independent calculation tasks.
What is faster for 4-core CPU - to run 4 threads or... lets say 50? Why second variant should be slover and how much slover?
As i assume: when you run 4 heavy threads on 4-core CPU without another CPU-consuming processes/threads, scheduler is allowed to not move the threads between cores at all; it has no reason to do that in this situation. Core0 (main CPU) will be responsible for executing interruption handler for hardware timer 250 times per second (basic Linux configuration) and other hardware interruption handlers, but another cores may not feel any worries.
What is the cost of context switching? The time for store and restore CPU registers for different context? What about caches, pipelines and various code-prediction things inside CPU? Can we say that each time we switch context, we hurt caches, pipelines and some code-decoding facilities in CPU? So more threads executing on a single core, less work they can do together in comparison to their serial execution?
Question about caches and another hardware optimization in multithreading environment is the interesting question for me now.
As #Baile mentions in the comments, this is highly application, system, environment-specific.
And as such, I'm not going to take the hard-line approach of mentioning exactly 1 thread for each core. (or 2 threads/core in the case of Hyperthreading)
As an experienced shared-memory programmer, I have seen from my experience that the optimal # of threads (for a 4 core machine) can range anywhere from 1 to 64+.
Now I will enumerate the situations that can cause this range:
Optimal Threads < # of Cores
In certain tasks that are very fine-grained paralleled (such as small FFTs), the overhead of threading is the dominant performance factor. In some cases, it's it not helpful to parallelize at all. In some cases, you get speedup with 2 threads, but backwards scaling at 4 threads.
Another issue is resource contention. Even if you have a highly parallelizable task that can easily split across 4 cores/threads, you may be bottlenecked by memory bandwidth and cache effects. So often, you find that 2 threads will be just as fast as 4 threads. (as if often the case with very large FFTs)
Optimal Threads = # of Cores
This is the optimal case. No need to explain here - one thread per core. Most embarrassingly parallel applications that are not memory or I/O bound fit right here.
Optimal Threads > # of Cores
This is where it gets interesting... very interesting. Have you heard about load-imbalance? How about over-decomposition and work-stealing?
Many parallelizable applications are irregular - meaning that the tasks do not split into sub-tasks of equal size. So if you may end up splitting a large task into 4 unequal sizes, assign them to 4 threads and run them on 4 cores... the result? Poor parallel performance because 1 thread happened to get 10x more work than the other threads.
A common solution here is to over-decompose the task into many sub-tasks. You can either create threads for each one of them (so now you get threads >> cores). Or you can use some sort of task-scheduler with a fixed number of threads. Not all tasks are suited for both, so quite often, the approach of over-decomposing a task to 8 or 16 threads for a 4-core machine gives optimal results.
Although spawning more threads can lead to better load-balance, the overhead builds up. So there's typically an optimal point somewhere. I've seen as high as 64 threads on 4 cores. But as mentioned, it's highly application specific. And you need to experiment.
EDIT : Expanding answer to more directly answer the question...
What is the cost of context switching? The time for store and restore
CPU registers for different context?
This is very dependent on the environment - and is somewhat difficult to measure directly. Short answer: Very Expensive This might be a good read.
What about caches, pipelines and various code-prediction things inside
CPU? Can we say that each time we switch context, we hurt caches,
pipelines and some code-decoding facilities in CPU?
Short answer: Yes When you context switch out, you likely flush your pipeline and mess up all the predictors. Same with caches. The new thread is likely to replace the cache with new data.
There's a catch though. In some applications where the threads share the same data, it's possible that one thread could potentially "warm" the cache for another incoming thread or another thread on a different core sharing the same cache. (Although rare, I've seen this happen before on one of my NUMA machines - superlinear speedup: 17.6x across 16 cores!?!?!)
So more threads executing on a single core, less work they can do
together in comparison to their serial execution?
Depends, depends... Hyperthreading aside, there will definitely be overhead. But I've read a paper where someone used a second thread to prefetch for the main thread... Yes it's crazy...
Creating 50 threads will actually hurt performance, not improve it. It just doesn't make any sense.
Ideally you should make the 4 threads, not more, not less. There will be some overhead because of context switching, but that is unavoidable. The OS/services/other applications threads should too be executed. But nowadays with so powerful and lighting-fast CPUs this is of no concern since those OS threads will only take less that 2 % of the CPU's time. Almost all of them will be in blocked state while your program is running.
You might think that, since performance is of critical importance, you should code those small critical areas in low-level assembly language. Modern programming languages allow this.
But seriously... compilers and, in case of Java, the JVM, will optimize those portions so well that it just isn't worth it (unless you actually want to exercise something like this). Instead of your calculations finishing in 100 seconds, they'll finish in 97 or 98. The question you must ask yourself is: is it worth all those hours of coding and debugging ?
You asked about the time cost of context switching. These days, these are extremely low. Look at modern day dual-core CPUs that run Windows 7 for example. If you start an Apache web server on that machine and a MySQL database server, you will easily go over 800 threads. The machine just doesn't feel it. To see how low this cost is, read here: How to estimate the thread context switching overhead? . To spare you the searching/reading part: context switching can be done hundreds of thousands of times per second.
4 threads are faster if you can program your 40 tasks switching better than Operating System does.
If you can use 4 threads, use them. There's no way 50 will go faster than 4 on a 4-core machine. All you get is more overhead.
Of course, you're describing an ideal non-real-world situation, so whatever you are actually building, you'll need to measure in order to understand how the performance is affected.
There is Hyperthreading technology which can handle more that one thread per CPU, but it is hardly dependent on type of calculation you want to do. Consider using of GPU or very low assembly language to achieve maximum power.

If 256 threads give better performance than 8 have I likely got the wrong approach?

I've just started programming with POSIX threads on dual-core x86_64 Linux system. It seems that 256 threads is about the optimum for performance with the way I've done it. I'm wondering how this could be? And if it could mean that my approach is wrong and a better approach would require far fewer threads and be just as fast or faster?
For further background (the program in question is a skeleton for a multi-threaded M-set image generator) see the following questions I've asked already:
Using threads, how should I deal with something which ideally should happen in sequential order?
How can my threaded image generating app get it’s data to the gui?
Perhaps I should mention that the skeleton (in which I've reproduced minimal functionality for testing and comparison) is now displaying the image, and the actual calculations are done almost twice as fast as the non-threaded program.
So if 256 threads running faster than 8 threads is not indicative of a poor approach to threading, how come 256 threads does outperform 8 threads?
The speed test case is a portion of the Mandelbrot Set located at:
xmin -0.76243636067708333333333328
xmax -0.7624335575810185185185186
ymax 0.077996663411458333333333929
calculated to a maximum of 30000 iterations.
On the non-threaded version rendering time on my system is around 15 seconds. On the threaded version, averages speed for 8 threads is 7.8 seconds, while 256 threads is 7.6 seconds.
Well, probably yes, you're doing something wrong.
However, there are circumstances where 256 threads would run better than 8 without you necessarily having a bad threading model. One must remember that having 8 threads does not mean all 8 threads are actually running all the time. Anytime one thread makes a blocking syscall to the operating system, the thread will stop running and wait for the result. In the meantime, another thread can often do work.
There's this myth that one can't usefully use more threads than contexts on the CPU, but that's just not true. If your threads block on a syscall, it can be critical to have another thread available to do more work. (In practice when threads block there tends to be less work to do, but this is not always the case.)
It's all very dependent on work-load and there's no one right number of threads for any particular application. Generally you never want less threads available than the OS will run, and that's the only true rule. (Unfortunately this can be very hard to find out and so people tend to just fire up as many threads as contexts and then use non-blocking syscalls where possible.)
Could it be your app is io bound? How is the image data generated?
A performance improvement gained by allocating more threads than cores suggests that the CPU is not the bottleneck. If I/O access such as disk, memory or even network access are involved your results make perfect sense.
You are probably benefitting from Simultaneous Multithreading (SMT). Your operating system schedules more threads than cores available, and will swap in and out the threads that are not stalled waiting for resources (such as a memory load). This can very effectively hide the latencies of your memory system from your program and is the technique used to great effect for massive parallelization in CUDA for general purpose GPU programming.
If you are seeing a performance increase with the jump to 256 threads, then what you are probably dealing with is a resource bottleneck. At some point, your code is waiting for some slow device (a hard disk or a network connection, for example) in order to continue. With multiple threads, waiting on this slow device isn't a problem because instead of sitting idle and twiddling its electronic thumbs, the CPU can process another thread while the first thread is waiting on the slow device. The more parallel threads that are running, the more work the CPU can do while it is waiting on something else.
If you are seeing performance improve all the way up to 256 threads, I am tempted to say that you have a major performance bottleneck somewhere and it's not the CPU. To test this, try to see if you can measure the idle time of individual threads. I suspect that you will see your threads are stuck in a "blocked" or "waiting" state for a longer portion of their lifetime than they spend in the "running" or "active" state. Some debuggers or function profiling tools will let you do this, and I think there are also Linux tools to do this on the command line.

Threads & Processes Vs MultiThreading & Multi-Core/MultiProcessor : How they are mapped?

I was very confused but the following thread cleared my doubts:
Multiprocessing, Multithreading,HyperThreading, Multi-core
But it addresses the queries from the hardware point of view. I want to know how these hardware features are mapped to software?
One thing that is obvious is that there is no difference between MultiProcessor(=Mutlicpu) and MultiCore other than that in multicore all cpus reside on one chip(die) where as in Multiprocessor all cpus are on their own chips & connected together.
So, mutlicore/multiprocessor systems are capable of executing multiple processes (firefox,mediaplayer,googletalk) at the "sametime" (unlike context switching these processes on a single processor system) Right?
If it correct. I'm clear so far. But the confusion arises when multithreading comes into picture.
MultiThreading "is for" parallel processing. right?
What are elements that are involved in multithreading inside cpu? diagram? For me to exploit the power of parallel processing of two independent tasks, what should be the requriements of CPU?
When people say context switching of threads. I don't really get it. because if its context switching of threads then its not parallel processing. the threads must be executed "scrictly simultaneously". right?
My notion of multithreading is that:
Considering a system with single cpu. when process is context switched to firefox. (suppose) each tab of firefox is a thread and all the threads are executing strictly at the same time. Not like one thread has executed for sometime then again another thread has taken until the context switch time is arrived.
What happens if I run a multithreaded software on a processor which can't handle threads? I mean how does the cpu handle such software?
If everything is good so far, now question is HOW MANY THREADS? It must be limited by hardware, I guess? If hardware can support only 2 threads and I start 10 threads in my process. How would cpu handle it? Pros/Cons? From software engineering point of view, while developing a software that will be used by the users in wide variety of systems, Then how would I decide should I go for multithreading? if so, how many threads?
First, try to understand the concept of 'process' and 'thread'. A thread is a basic unit for execution: a thread is scheduled by operating system and executed by CPU. A process is a sort of container that holds multiple threads.
Yes, either multi-processing or multi-threading is for parallel processing. More precisely, to exploit thread-level parallelism.
Okay, multi-threading could mean hardware multi-threading (one example is HyperThreading). But, I assume that you just say multithreading in software. In this sense, CPU should support context switching.
Context switching is needed to implement multi-tasking even in a physically single core by time division.
Say there are two physical cores and four very busy threads. In this case, two threads are just waiting until they will get the chance to use CPU. Read some articles related to preemptive OS scheduling.
The number of thread that can physically run in concurrent is just identical to # of logical processors. You are asking a general thread scheduling problem in OS literature such as round-robin..
I strongly suggest you to study basics of operating system first. Then move on multithreading issues. It seems like you're still unclear for the key concepts such as context switching and scheduling. It will take a couple of month, but if you really want to be an expert in computer software, then you should know such very basic concepts. Please take whatever OS books and lecture slides.
Threads running on the same core are not technically parallel. They only appear to be executed in parallel, as the CPU switches between them very fast (for us, humans). This switch is what is called context switch.
Now, threads executing on different cores are executed in parallel.
Most modern CPUs have a number of cores, however, most modern OSes (windows, linux and friends) usually execute much larger number of threads, which still causes context switches.
Even if no user program is executed, still OS itself performs context switches for maintanance work.
This should answer 1-3.
About 4: basically, every processor can work with threads. it is much more a characteristic of operating system. Thread is basically: memory (optional), stack and registers, once those are replaced you are in another thread.
5: the number of threads is pretty high and is limited by OS. Usually it is higher than regular programmer can successfully handle :)
The number of threads is dictated by your program:
is it IO bound?
can the task be divided into a number of smaller tasks?
how small is the task? the task can be too small to make it worth to spawn threads at all.
synchronization: if extensive synhronization is required, the penalty might be too heavy and you should reduce the number of threads.
Multiple threads are separate 'chains' of commands within one process. From CPU point of view threads are more or less like processes. Each thread has its own set of registers and its own stack.
The reason why you can have more threads than CPUs is that most threads don't need CPU all the time. Thread can be waiting for user input, downloading something from the web or writing to disk. While it is doing that, it does not need CPU, so CPU is free to execute other threads.
In your example, each tab of Firefox probably can even have several threads. Or they can share some threads. You need one for downloading, one for rendering, one for message loop (user input), and perhaps one to run Javascript. You cannot easily combine them because while you download you still need to react to user's input. However, download thread is sleeping most of the time, and even when it's downloading it needs CPU only occasionally, and message loop thread only wakes up when you press a button.
If you go to task manager you'll see that despite all these threads your CPU use is still quite low.
Of course if all your threads do some number-crunching tasks, then you shouldn't create too many of them as you get no performance benefit (though there may be architectural benefits!).
However, if they are mainly I/O bound then create as many threads as your architecture dictates. It's hard to give advice without knowing your particular task.
Broadly speaking, yeah, but "parallel" can mean different things.
It depends what tasks you want to run in parallel.
Not necessarily. Some (indeed most) threads spend a lot of time doing nothing. Might as well switch away from them to a thread that wants to do something.
The OS handles thread switching. It will delegate to different cores if it wants to. If there's only one core it'll divide time between the different threads and processes.
The number of threads is limited by software and hardware. Threads consume processor and memory in varying degrees depending on what they're doing. The thread management software may impose its own limits as well.
The key thing to remember is the separation between logical/virtual parallelism and real/hardware parallelism. With your average OS, a system call is performed to spawn a new thread. What actually happens (whether it is mapped to a different core, a different hardware thread on the same core, or queued into the pool of software threads) is up to the OS.
Parallel processing uses all the methods not just multi-threading.
Generally speaking, if you want to have real parallel processing, you need to perform it in hardware. Take the example of the Niagara, it has up to 8-cores each capable of executing 4-threads in hardware.
Context switching is needed when there are more threads than is capable of being executed in parallel in hardware. Even then, when executed in series (switching between one thread to the next), they are considered concurrent because there is no guarantee on the order of switching. So, it may go T0, T1, T2, T1, T3, T0, T2 and so on. For all intents and purposes, the threads are parallel.
Time slicing.
That would be up to the OS.
Multithreading is the execution of more than one thread at a time. It can happen both on single core processors and the multicore processor systems. For single processor systems, context switching effects it. Look!Context switching in this computational environment refers to time slicing by the operating system. Therefore do not get confused. The operating system is the one that controls the execution of other programs. It allows one program to execute in the CPU at a time. But the frequency at which the threads are switched in and out of the CPU determines the transparency of parallelism exhibited by the system.
For multicore environment,multithreading occurs when each core executes a thread.Though,in multicore again,context switching can occur in the individual cores.
I think answers so far are pretty much to the point and give you a good basic context. In essence, say you have quad core processor, but each core is capable of executing 2 simultaneous threads.
Note, that there is only slight (or no) increase of speed if you are running 2 simultaneous threads on 1 core versus you run 1st thread and then 2nd thread vertically. However, each physical core adds speed to your general workflow.
Now, say you have a process running on your OS that has multiple threads (i.e. needs to run multiple things in "parallel") and has some kind of stack of tasks in a queue (or some other system with priority rules). Then software sends tasks to a queue and your processor attempts to execute them as fast as it can. Now you have 2 cases:
If a software supports multiprocessing, then tasks will be sent to any available processor (that is not doing anything or simply finished doing some other job and job send from your software is 1st in a queue).
If your software does not support multiprocessing, then all of your jobs will be done in a similar manner, but only by one of your cores.
I suggest reading Wikipedia page on thread. Very first picture there already gives you a nice insight. :)

How many simultaneous threads in an application is a lot?

5, 100, 1000?
I guess, "it depends", but on what?
What is common in applications that run as server daemons / services?
What are hard limits?
Given that the machine can handle the overall workload, how do I determine at how many threads the overhead starts to have an impact on performance?
What are important differences between OS's?
What else should be considered?
I'm asking because I would like to employ threads in an application to organize subcomponents of my application that do not share data and are designed to do their work in parallel. As the application would also use thread pools for parallelizing some tasks, I was wondering at what point I should start to think about the number of threads that's going to run in total.
I know the n+1 rule as a guideline for determining the number of threads that simultaneously work on the same task to gain performance. However, I want to use threads like one might use processes in a larger scope, i. e. to organize independent tasks that should not interfere with each other.
In this related question, some people advise to minimise the number of threads because of the added complexity. To me it seems that threads can also help to keep things sorted more orderly and actually reduce interference. Isn't that correct?
I can't answer your question about "how much is many" but I agree that you should not use threads for every task possible.
The optimal amount of threads for performance of application is (n+1), where n is the amount of processors/cores your computer/claster has.
The more your actual thread amount differs from n+1, the less optimal it gets and gets your system resources wasted on thread calculations.
So usually you use 1 thread for the UI, 1 thread for some generic tasks, and (n+1) threads for some huge-calculation tasks.
Actually Ajmastrean is a little out of date. Quoting from his own link
The thread pool has a default size of
250 worker threads per available
processor, and 1000 I/O completion
threads. The number of threads in the
thread pool can be changed by using
the SetMaxThreads method.
But generally I think 25 is really where the law of diminishing returns (and programmers abilities to keep track of what is going on) starts coming into effect. Although Max is right, as long as all of the threads are performing non-blocking calculations n+1 is the optimal number, in the real world most of the threading tasks I perform tend to be done on stuff with some kind of IO.
Also depends on your architecture. E.g. in NVIDIA GPGPU lib CUDA you can put on an 8 thread multiprocessor 512 threads simoultanously. You may ask why assign each of the scalar processors 64 threads? The answer is easy: If the computation is not compute bound but memory IO bound, you can hide the mem latencies by executing other threads. Similar applies to normal CPUs. I can remember that a recommendation for the parallel option for make "-j" is to use approx 1.5 times the number of cores you got. Many of the compiling tasks are heavy IO burden and if a task has to wait for harddisk, mem ... whatever, CPU could work on a different thread.
Next you have to consider, how expensive a task/thread switch is. E.g. it is comes free, while CPU has to perform some work for a context switch. So in general you have to estimate if the penalty for two task switches is longer than the time the thread would block (which depends heavily on your applications).
Microsoft's ThreadPool class limits you to 25 threads per processor. The limit is based on context switching between threads and the memory consumed by each thread. So, that's a good guideline if you're on the Windows platform.

Resources