We have code that makes use of sched_yield inside a loop. When we do this we seem to get a slower performance of other threads, in particular those involving kernel calls (like IO and mutex/event handling). I'm trying to determine the exact cause of this behaviour.
Can excessive calls to sched_yield lead to a bottleneck in the kernel?
My suspicion is if we keep asking the kernel to check its process list then other threads will suffer as key data structures may be continually locked -- whereas if we didn't call sched_yield those kernel locks would tend to be uncontested. Does this make sense, or should it be totally okay to repeatedly call sched_yield.
Have a look at the sched_yield man page for Linux:
Avoid calling sched_yield()
unnecessarily or inappropriately (e.g., when resources needed by other
schedulable threads are still held by the caller), since doing so will result
in unnecessary context switches, which will degrade system performance.
Calling it in a tight loop will cause problems. Reduce the rate at which you're calling it.
(And check that you need to call it in the first place. The scheduler often does the Right Thing all by itself.)
Other options you could find interesting to investigate if you have a low priority thread:
sched_setscheduler - with SCHED_IDLE or SCHED_BATCH maybe (affects the whole process)
thread_setschedparam - per thread, but might have restrictions on what policies you can use (can't find it right now).
Or the good old nice command of course.
Related
When a thread does something that may cause it to become blocked locally, for example, waiting for another thread in its process to complete some work, it calls a run-time system procedure. This procedure checks to see if the thread must be put into blocked state. If so, it stores the thread's registers in the thread table, looks in the table for a ready thread to run, and reloads the machine registers with the new thread's saved values. As soon as the stack pointer and program counter have been switched, the new thread comes to life again automatically. If the machine happens to have an instruction to store all the registers and another one to load them all, the entire thread switch can be done in just a handful of instructions. Doing thread switching like this is at least an order of magnitude-maybe more-faster than trapping to the kernel and is a strong argument in favor of user-level threads packages.
Source: Modern Operating Systems (Andrew S. Tanenbaum | Herbert Bos)
The above argument is made in favor of user-level threads. The user-level thread implementation is depicted as kernel managing all the processes, where individual processes can have their own run-time (made available by a library package) that manages all the threads in that process.
Of course, merely calling a function in the run-time than trapping to kernel might have a few less instructions to execute but why the difference is so huge?
For example, if threads are implemented in kernel space, every time a thread has to be created the program is required to make a system call. Yes. But the call only involves adding an entry to the thread table with certain attributes (which is also the case in user space threads). When a thread switch has to happen, kernel can simply do what the run-time (at user-space) would do. The only real difference I can see here is that the kernel is being involved in all this. How can the performance difference be so significant?
Threads implemented as a library package in user space perform significantly better. Why?
They're not.
The fact is that most task switches are caused by threads blocking (having to wait for IO from disk or network, or from user, or for time to pass, or for some kind of semaphore/mutex shared with a different process, or some kind of pipe/message/packet from a different process) or caused by threads unblocking (because whatever they were waiting for happened); and most reasons to block and unblock involve the kernel in some way (e.g. device drivers, networking stack, ...); so doing task switches in kernel when you're already in the kernel is faster (because it avoids the overhead of switching to user-space and back for no sane reason).
Where user-space task switching "works" is when kernel isn't involved at all. This mostly only happens when someone failed to do threads properly (e.g. they've got thousands of threads and coarse-grained locking and are constantly switching between threads due to lock contention, instead of something sensible like a "worker thread pool"). It also only works when all threads are the same priority - you don't want a situation where very important threads belonging to one process don't get CPU time because very unimportant threads belonging to a different process are hogging the CPU (but that's exactly what happens with user-space threading because one process has no idea about threads belonging to a different process).
Mostly; user-space threading is a silly broken mess. It's not faster or "significantly better"; it's worse.
When a thread does something that may cause it to become blocked locally, for example, waiting for another thread in its process to complete some work, it calls a run-time system procedure. This procedure checks to see if the thread must be put into blocked state. If so, it stores the thread's registers in the thread table, looks in the table for a ready thread to run, and reloads the machine registers with the new thread's saved values. As soon as the stack pointer and program counter have been switched, the new thread comes to life again automatically. If the machine happens to have an instruction to store all the registers and another one to load them all, the entire thread switch can be done in just a handful of instructions. Doing thread switching like this is at least an order of magnitude-maybe more-faster than trapping to the kernel and is a strong argument in favor of user-level threads packages.
This is talking about a situation where the CPU itself does the actual task switch (and either the kernel or a user-space library tells the CPU when to do a task switch to what). This has some relatively interesting history behind it...
In the 1980s Intel designed a CPU ("iAPX" - see https://en.wikipedia.org/wiki/Intel_iAPX_432 ) for "secure object oriented programming"; where each object has its own isolated memory segments and its own privilege level, and can transfer control directly to other objects. The general idea being that you'd have a single-tasking system consisting of global objects using cooperating flow control. This failed for multiple reasons, partly because all the protection checks ruined performance, and partly because the majority of software at the time was designed for "multi-process preemptive time sharing, with procedural programming".
When Intel designed protected mode (80286, 80386) they still had hopes for "single-tasking system consisting of global objects using cooperating flow control". They included hardware task/object switching, local descriptor table (so each task/object can have its own isolated segments), call gates (so tasks/objects can transfer control to each other directly), and modified a few control flow instructions (call far and jmp far) to support the new control flow. Of course this failed for the same reason iAPX failed; and (as far as I know) nobody has ever used these things for the "global objects using cooperative flow control" they were originally designed for. Some people (e.g. very early Linux) did try to use the hardware task switching for more traditional "multi-process preemptive time sharing, with procedural programming" systems; but found that it was slow because the hardware task switch did too many protection checks that could be avoided by software task switching and saved/reloaded too much state that could be avoided by a software task switching;p and didn't do any of the other stuff needed for a task switch (e.g. keeping statistics of CPU time used, saving/restoring debug registers, etc).
Now.. Andrew S. Tanenbaum is a micro-kernel advocate. His ideal system consists of isolated pieces in user-space (processes, services, drivers, ...) communicating via. synchronous messaging. In practice (ignoring superficial differences in terminology) this "isolated pieces in user-space communicating via. synchronous messaging" is almost entirely identical to Intel's twice failed "global objects using cooperative flow control".
Mostly; in theory (if you ignore all the practical problems, like CPU not saving all of the state, and wanting to do extra work on task switches like tracking statistics), for a specific type of OS that Andrew S. Tanenbaum prefers (micro-kernel with synchronous message passing, without any thread priorities), it's plausible that the paragraph quoted above is more than just wishful thinking.
I think the answer to this can use a lot of OS and parallel distributive computing knowledge (And I am not sure about the answer but I will try my best)
So if you think about it. The library package will have a greater amount of performance than you write in the kernel itself. In the package thing, interrupt given by this code will be held at once and al the execution will be done. While when you write in kernel different other interrupts can come before. Plus accessing threads again and again is harsh on the kernel since everytime there will be an interrupt. I hope it will be a better view.
it's not correct to say the user-space threads are better that the kernel-space threads since each one has its own pros and cons.
in terms of user-space threads, as the application is responsible for managing thread, its easier to implement such threads and that kind of threads have not much reliance on OS. however, you are not able to use the advantages of multi processing.
In contrary, the kernel space modules are handled by OS, so you need to implement them according to the OS that you use, and it would be a more complicated task. However, you have more control over your threads.
for more comprehensive tutorial, take a look here.
Can mlock() be called safely from independently executing OpenMP or Posix threads, given that each thread is operating on a different region of virtual memory? Does it create a systemwide synchronization barrier or force all threads to stall in some way?
I apologize if this is a duplicate; I was surprised when google searches for "mlock openmp"/"mlock thread safety" did not turn up the answer immediately. Closest I could find was the second answer of Non-blocking mlock(), which seems to indicate that mlock() CAN be called from separate threads and does not enforce or require any synchronization barriers.
mlock() is safe to call from multiple threads at once.
As to whether it synchronises against other calls to mlock(), it's a quality-of-implementation issue - in principle any system call could synchronise against any other, there's no text in POSIX that disallows it. In practice you will often find that system calls that work on the process's memory map tend to contend with each other (so mlock() might not just contend with other mlock() calls but also mmap()). You will need to test to see if contention is a actually a problem in your use case.
I have time-critical piece of code in my app. I made the thread which runs it Highest priority - that's the most I could do.
Is there any suggestions on how to make part of the code run in this thread to be interrupted as few times as possible (less context switch occurred) ?
The code is not complicated. I replaced all the method calls with inlined code and I don't use anything of high level (like no LINQ). The most of operations are arithmetic. There is only one comparison of strings (I am thinking of ways how to rid of it). Half of maths are with ints and half with doubles.
The code is x86 .NET 4 C#. Runs on single Xenon X3450 W2008R2. Single app server.
(Unfortunately data is coming from 3d party API which doesn't support x64 (hate it!))
I'd appreciate grown-up discussion with experienced developers.
P.S. The server has no paging file so hard page faults wont happen either (no unwanted IO operations).
The only thing you need to worry about in terms of context switches, is blocking your thread. So there should be no problem with using LINQ (that is, LINQ-to-objects, obviously LINQ-to-SQL or whatever would involve blocking!). Any sort of arithmetic or calling methods and so on will also not block the thread and so have no impact on context switches.
The other thing that affects context switching is, as you noted, priority. But not just thread priority, also your process's priority. You can use SetPriorityClass to increase your process's priority to ABOVE_NORMAL_PRIORITY_CLASS (I wouldn't bother putting it higher than that) and then set your thread's priority to Above Normal as well.
However, in general, priorities are really only useful when it's a matter of timing (that is, making sure your process responds to external input (network, user input, disk I/O) as fast as possible). It will actually have very little impact on your thread's actual throughput, unless you have other processes that are also CPU-bound running at the same time. But if that's the case, then fiddling with priorities is not going to be a viable long-term solution anyway. This is because you'll find that by setting one of the processes to a higher priority, it'll completely starve the other processes and they'll never run.
So anyway, I would carefully consider things before adjusting thread and process priorities. And, as always, test, test, test!
If you make that unmanaged WINAPI code instead, the SetThreadPriority function also supports a THREAD_PRIORITY_TIME_CRITICAL (higher than THREAD_PRIORITY_HIGHEST).
It's also worth boosting the priority of the process in which the thread is running (actual priority depends on a combination of thread and process priority).
You should also avoid making I/O calls on the thread (which could block). Taking it to a perhaps-ridiculous extreme you could also avoid making I/O calls on other threads (which could temporarily boost the priority of those threads).
I am new to linux and linux threads. I have spent some time googling to try to understand the differences between all the functions available for thread synchronization. I still have some questions.
I have found all of these different types of synchronizations, each with a number of functions for locking, unlocking, testing the lock, etc.
gcc atomic operations
futexes
mutexes
spinlocks
seqlocks
rculocks
conditions
semaphores
My current (but probably flawed) understanding is this:
semaphores are process wide, involve the filesystem (virtually I assume), and are probably the slowest.
Futexes might be the base locking mechanism used by mutexes, spinlocks, seqlocks, and rculocks. Futexes might be faster than the locking mechanisms that are based on them.
Spinlocks dont block and thus avoid context swtiches. However they avoid the context switch at the expense of consuming all the cycles on a CPU until the lock is released (spinning). They should only should be used on multi processor systems for obvious reasons. Never sleep in a spinlock.
The seq lock just tells you when you finished your work if a writer changed the data the work was based on. You have to go back and repeat the work in this case.
Atomic operations are the fastest synch call, and probably are used in all the above locking mechanisms. You do not want to use atomic operations on all the fields in your shared data. You want to use a lock (mutex, futex, spin, seq, rcu) or a single atomic opertation on a lock flag when you are accessing multiple data fields.
My questions go like this:
Am I right so far with my assumptions?
Does anyone know the cpu cycle cost of the various options? I am adding parallelism to the app so we can get better wall time response at the expense of running fewer app instances per box. Performances is the utmost consideration. I don't want to consume cpu with context switching, spinning, or lots of extra cpu cycles to read and write shared memory. I am absolutely concerned with number of cpu cycles consumed.
Which (if any) of the locks prevent interruption of a thread by the scheduler or interrupt...or am I just an idiot and all synchonization mechanisms do this. What kinds of interruption are prevented? Can I block all threads or threads just on the locking thread's CPU? This question stems from my fear of interrupting a thread holding a lock for a very commonly used function. I expect that the scheduler might schedule any number of other workers who will likely run into this function and then block because it was locked. A lot of context switching would be wasted until the thread with the lock gets rescheduled and finishes. I can re-write this function to minimize lock time, but still it is so commonly called I would like to use a lock that prevents interruption...across all processors.
I am writing user code...so I get software interrupts, not hardware ones...right? I should stay away from any functions (spin/seq locks) that have the word "irq" in them.
Which locks are for writing kernel or driver code and which are meant for user mode?
Does anyone think using an atomic operation to have multiple threads move through a linked list is nuts? I am thinking to atomicly change the current item pointer to the next item in the list. If the attempt works, then the thread can safely use the data the current item pointed to before it was moved. Other threads would now be moved along the list.
futexes? Any reason to use them instead of mutexes?
Is there a better way than using a condition to sleep a thread when there is no work?
When using gcc atomic ops, specifically the test_and_set, can I get a performance increase by doing a non atomic test first and then using test_and_set to confirm? I know this will be case specific, so here is the case. There is a large collection of work items, say thousands. Each work item has a flag that is initialized to 0. When a thread has exclusive access to the work item, the flag will be one. There will be lots of worker threads. Any time a thread is looking for work, they can non atomicly test for 1. If they read a 1, we know for certain that the work is unavailable. If they read a zero, they need to perform the atomic test_and_set to confirm. So if the atomic test_and_set is 500 cpu cycles because it is disabling pipelining, causes cpu's to communicate and L2 caches to flush/fill .... and a simple test is 1 cycle .... then as long as I had a better ratio of 500 to 1 when it came to stumbling upon already completed work items....this would be a win.
I hope to use mutexes or spinlocks to sparilngly protect sections of code that I want only one thread on the SYSTEM (not jsut the CPU) to access at a time. I hope to sparingly use gcc atomic ops to select work and minimize use of mutexes and spinlocks. For instance: a flag in a work item can be checked to see if a thread has worked it (0=no, 1=yes or in progress). A simple test_and_set tells the thread if it has work or needs to move on. I hope to use conditions to wake up threads when there is work.
Thanks!
Application code should probably use posix thread functions. I assume you have man pages so type
man pthread_mutex_init
man pthread_rwlock_init
man pthread_spin_init
Read up on them and the functions that operate on them to figure out what you need.
If you're doing kernel mode programming then it's a different story. You'll need to have a feel for what you are doing, how long it takes, and what context it gets called in to have any idea what you need to use.
Thanks to all who answered. We resorted to using gcc atomic operations to synchronize all of our threads. The atomic ops were about 2x slower than setting a value without synchronization, but magnitudes faster than locking a mutex, changeing the value, and then unlocking the mutex (this becomes super slow when you start having threads bang into the locks...) We only use pthread_create, attr, cancel, and kill. We use pthread_kill to signal threads to wake up that we put to sleep. This method is 40x faster than cond_wait. So basicly....use pthreads_mutexes if you have time to waste.
in addtion you should check the nexts books
Pthreads Programming: A POSIX
Standard for Better Multiprocessing
and
Programming with POSIX(R) Threads
regarding question # 8
Is there a better way than using a condition to sleep a thread when there is no work?
yes i think that the best aproach instead of using sleep
is using function like sem_post() and sem_wait of "semaphore.h"
regards
A note on futexes - they are more descriptively called fast userspace mutexes. With a futex, the kernel is involved only when arbitration is required, which is what provides the speed up and savings.
Implementing a futex can be extremely tricky (PDF), debugging them can lead to madness. Unless you really, really, really need the speed, its usually best to use the pthread mutex implementation.
Synchronization is never exactly easy, but trying to implement your own in userspace makes it inordinately difficult.
Whilst learning the "assembler language" (in linux on a x86 architecture using the GNU as assembler), one of the aha moments was the possibility of using system calls. These system calls come in very handy and are sometimes even necessary as your program runs in user-space.
However system calls are rather expensive in terms of performance as they require an interrupt (and of course a system call) which means that a context switch must be made from your current active program in user-space to the system running in kernel-space.
The point I want to make is this: I'm currently implementing a compiler (for a university project) and one of the extra features I wanted to add is the support for multi-threaded code in order to enhance the performance of the compiled program. Because some of the multi-threaded code will be automatically generated by the compiler itself, this will almost guarantee that there will be really tiny bits of multi-threaded code in it as well. In order to gain a performance win, I must be sure that using threads will make this happen.
My fear however is that, in order to use threading, I must make system calls and the necessary interrupts. The tiny little (auto-generated) threads will therefore be highly affected by the time it takes to make these system calls, which could even lead to a performance loss...
my question is therefore twofold (with an extra bonus question underneath it):
Is it possible to write assembler
code which can run multiple threads
simultaneously on multiple cores at
once, without the need of system
calls?
Will I get a performance gain if I have really tiny threads (tiny as in the total execution time of the thread), performance loss, or isn't it worth the effort at all?
My guess is that multithreaded assembler code is not possible without system calls. Even if this is the case, do you have a suggestion (or even better: some real code) for implementing threads as efficient as possible?
The short answer is that you can't. When you write assembly code it runs sequentially (or with branches) on one and only one logical (i.e. hardware) thread. If you want some of the code to execute on another logical thread (whether on the same core, on a different core on the same CPU or even on a different CPU), you need to have the OS set up the other thread's instruction pointer (CS:EIP) to point to the code you want to run. This implies using system calls to get the OS to do what you want.
User threads won't give you the threading support that you want, because they all run on the same hardware thread.
Edit: Incorporating Ira Baxter's answer with Parlanse. If you ensure that your program has a thread running in each logical thread to begin with, then you can build your own scheduler without relying on the OS. Either way, you need a scheduler to handle hopping from one thread to another. Between calls to the scheduler, there are no special assembly instructions to handle multi-threading. The scheduler itself can't rely on any special assembly, but rather on conventions between parts of the scheduler in each thread.
Either way, whether or not you use the OS, you still have to rely on some scheduler to handle cross-thread execution.
"Doctor, doctor, it hurts when I do this". Doctor: "Don't do that".
The short answer is you can do multithreaded programming without
calling expensive OS task management primitives. Simply ignore the OS for thread
scheduling operations. This means you have to write your own thread
scheduler, and simply never pass control back to the OS.
(And you have to be cleverer somehow about your thread overhead
than the pretty smart OS guys).
We chose this approach precisely because windows process/thread/
fiber calls were all too expensive to support computation
grains of a few hundred instructions.
Our PARLANSE programming langauge is a parallel programming language:
See http://www.semdesigns.com/Products/Parlanse/index.html
PARLANSE runs under Windows, offers parallel "grains" as the abstract parallelism
construct, and schedules such grains by a combination of a highly
tuned hand-written scheduler and scheduling code generated by the
PARLANSE compiler that takes into account the context of grain
to minimimze scheduling overhead. For instance, the compiler
ensures that the registers of a grain contain no information at the point
where scheduling (e.g., "wait") might be required, and thus
the scheduler code only has to save the PC and SP. In fact,
quite often the scheduler code doesnt get control at all;
a forked grain simply stores the forking PC and SP,
switches to compiler-preallocated stack and jumps to the grain
code. Completion of the grain will restart the forker.
Normally there's an interlock to synchronize grains, implemented
by the compiler using native LOCK DEC instructions that implement
what amounts to counting semaphores. Applications
can fork logically millions of grains; the scheduler limits
parent grains from generating more work if the work queues
are long enough so more work won't be helpful. The scheduler
implements work-stealing to allow work-starved CPUs to grab
ready grains form neighboring CPU work queues. This has
been implemented to handle up to 32 CPUs; but we're a bit worried
that the x86 vendors may actually swamp use with more than
that in the next few years!
PARLANSE is a mature langauge; we've been using it since 1997,
and have implemented a several-million line parallel application in it.
Implement user-mode threading.
Historically, threading models are generalised as N:M, which is to say N user-mode threads running on M kernel-model threads. Modern useage is 1:1, but it wasn't always like that and it doesn't have to be like that.
You are free to maintain in a single kernel thread an arbitrary number of user-mode threads. It's just that it's your responsibility to switch between them sufficiently often that it all looks concurrent. Your threads are of course co-operative rather than pre-emptive; you basically scatted yield() calls throughout your own code to ensure regular switching occurs.
If you want to gain performance, you'll have to leverage kernel threads. Only the kernel can help you get code running simultaneously on more than one CPU core. Unless your program is I/O bound (or performing other blocking operations), performing user-mode cooperative multithreading (also known as fibers) is not going to gain you any performance. You'll just be performing extra context switches, but the one CPU that your real thread is running will still be running at 100% either way.
System calls have gotten faster. Modern CPUs have support for the sysenter instruction, which is significantly faster than the old int instruction. See also this article for how Linux does system calls in the fastest way possible.
Make sure that the automatically-generated multithreading has the threads run for long enough that you gain performance. Don't try to parallelize short pieces of code, you'll just waste time spawning and joining threads. Also be wary of memory effects (although these are harder to measure and predict) -- if multiple threads are accessing independent data sets, they will run much faster than if they were accessing the same data repeatedly due to the cache coherency problem.
Quite a bit late now, but I was interested in this kind of topic myself.
In fact, there's nothing all that special about threads that specifically requires the kernel to intervene EXCEPT for parallelization/performance.
Obligatory BLUF:
Q1: No. At least initial system calls are necessary to create multiple kernel threads across the various CPU cores/hyper-threads.
Q2: It depends. If you create/destroy threads that perform tiny operations then you're wasting resources (the thread creation process would greatly exceed the time used by the tread before it exits). If you create N threads (where N is ~# of cores/hyper-threads on the system) and re-task them then the answer COULD be yes depending on your implementation.
Q3: You COULD optimize operation if you KNEW ahead of time a precise method of ordering operations. Specifically, you could create what amounts to a ROP-chain (or a forward call chain, but this may actually end up being more complex to implement). This ROP-chain (as executed by a thread) would continuously execute 'ret' instructions (to its own stack) where that stack is continuously prepended (or appended in the case where it rolls over to the beginning). In such a (weird!) model the scheduler keeps a pointer to each thread's 'ROP-chain end' and writes new values to it whereby the code circles through memory executing function code that ultimately results in a ret instruction. Again, this is a weird model, but is intriguing nonetheless.
Onto my 2-cents worth of content.
I recently created what effectively operate as threads in pure assembly by managing various stack regions (created via mmap) and maintaining a dedicated area to store the control/individualization information for the "threads". It is possible, although I didn't design it this way, to create a single large block of memory via mmap that I subdivide into each thread's 'private' area. Thus only a single syscall would be required (although guard pages between would be smart these would require additional syscalls).
This implementation uses only the base kernel thread created when the process spawns and there is only a single usermode thread throughout the entire execution of the program. The program updates its own state and schedules itself via an internal control structure. I/O and such are handled via blocking options when possible (to reduce complexity), but this isn't strictly required. Of course I made use of mutexes and semaphores.
To implement this system (entirely in userspace and also via non-root access if desired) the following were required:
A notion of what threads boil down to:
A stack for stack operations (kinda self explaining and obvious)
A set of instructions to execute (also obvious)
A small block of memory to hold individual register contents
What a scheduler boils down to:
A manager for a series of threads (note that processes never actually execute, just their thread(s) do) in a scheduler-specified ordered list (usually priority).
A thread context switcher:
A MACRO injected into various parts of code (I usually put these at the end of heavy-duty functions) that equates roughly to 'thread yield', which saves the thread's state and loads another thread's state.
So, it is indeed possible to (entirely in assembly and without system calls other than initial mmap and mprotect) to create usermode thread-like constructs in a non-root process.
I only added this answer because you specifically mention x86 assembly and this answer was entirely derived via a self-contained program written entirely in x86 assembly that achieves the goals (minus multi-core capabilities) of minimizing system calls and also minimizes system-side thread overhead.
System calls are not that slow now, with syscall or sysenter instead of int. Still, there will only be an overhead when you create or destroy the threads. Once they are running, there are no system calls. User mode threads will not really help you, since they only run on one core.
First you should learn how to use threads in C (pthreads, POSIX theads). On GNU/Linux you will probably want to use POSIX threads or GLib threads.
Then you can simply call the C from assembly code.
Here are some pointers:
Posix threads: link text
A tutorial where you will learn how to call C functions from assembly: link text
Butenhof's book on POSIX threads link text