What's the cost/overhead of thread context switch?
So far as I know, there are direct costs:
Saving and restoring the context(generally,it includes general
purpose registers and program counter)
Costs of thread scheduling(deciding which thread is the next to run)
And there maybe some indirect costs, such as:
If a thread switched out was arranged to run in another CPU later, then this might re-load variables from main memory(or actually from other CPU via coherence protocol), that is a cache read miss might occur.
Is there any other indirect costs?
Related
Copy pasted from this link:
Thread switching does not require Kernel mode privileges.
User level threads are fast to create and manage.
Kernel threads are generally slower to create and manage than the user threads.
Transfer of control from one thread to another within the same process requires a mode switch to the Kernel.
I never came across these points while reading standard operating systems reference books. Though these points sound logical, I wanted to know how they reflect in Linux. To be precise :
Can someone give detailed steps involved in context switching between user threads and kernel threads, so that I can find the step difference between the two.
Can someone explain the difference with actual context switch example or code. May be system calls involved (in case of context switching between kernel threads) and thread library calls involved (in case of context switching between user threads).
Can someone link me to Linux source code line (say on github) handling context switch.
I also doubt why context switch between kernel threads requires changing to kernel mode. Aren't we already in kernel mode for first thread?
Can someone give detailed steps involved in context switching between user threads and kernel threads, so that I can find the step difference between the two.
Let's imagine a thread needs to read data from a file, but the file isn't cached in memory and disk drives are slow so the thread has to wait; and for simplicity let's also assume that the kernel is monolithic.
For kernel threading:
thread calls a "read()" function in a library or something; which must cause at least a switch to kernel code (because it's going to involve device drivers).
the kernel adds the IO request to the disk driver's "queue of possibly many pending requests"; realizes the thread will need to wait until the request completes, sets the thread to "blocked waiting for IO" and switches to a different thread (that may belong to a completely different process, depending on global thread priorities). The kernel returns to the user-space of whatever thread it switch to.
later; the disk hardware causes an IRQ which causes a switch back to the IRQ handler in kernel code. The disk driver finishes up the work it had to do the for (currently blocked) thread and unblocks that thread. At this point the kernel might decide to switch to the "now unblocked" thread; and the kernel returns to the user-space of the "now unblocked" thread.
For user threading:
thread calls a "read()" function in a library or something; which must cause at least a switch to kernel code (because it's going to involve device drivers).
the kernel adds the IO request to the disk driver's "queue of possibly many pending requests"; realizes the thread will need to wait until the request completes but can't take care of that because some fool decided to make everything worse by doing thread switching in user space, so the kernel returns to user-space with "IO request has been queued" status.
after the pointless extra overhead of switching back to user-space; the user-space scheduler does the thread switch that the kernel could have done. At this point the user-space scheduler will either tell kernel it has nothing to do and you'll have more pointless extra overhead switching back to kernel; or user-space scheduler will do a thread switch to another thread in the same process (which may be the wrong thread because a thread in a different process is higher priority).
later; the disk hardware causes an IRQ which causes a switch back to the IRQ handler in kernel code. The disk driver finishes up the work it had to do for the (currently blocked) thread; but the kernel isn't able to do the thread switch to unblock the thread because some fool decided to make everything worse by doing thread switching in user space. Now we've got a problem - how does kernel inform the user-space scheduler that the IO has finished? To solve this (without any "user-space scheduler running zero threads constantly polls kernel" insanity) you have to have some kind of "kernel puts notification of IO completion on some kind of queue and (if the process was idle) wakes the process up" which (on its own) will be more expensive than just doing the thread switch in the kernel. Of course if the process wasn't idle then code in user-space is going to have to poll its notification queue to find out if/when the "notification of IO completion" arrives, and that's going to increase latency and overhead. In any case, after lots of stupid pointless and avoidable overhead; the user-space scheduler can do the thread switch.
Can someone explain the difference with actual context switch example or code. May be system calls involved (in case of context switching between kernel threads) and thread library calls involved (in case of context switching between user threads).
The actual low-level context switch code typically begins with something like:
save whichever registers are "caller preserved" according to the calling conventions on the stack
save the current stack top in some kind of "thread info structure" belonging to the old thread
load a new stack top from some kind of "thread info structure" belonging to the new thread
pop whichever registers are "caller preserved" according to the calling conventions
return
However:
usually (for modern CPUs) there's a relatively large amount of "SIMD register state" (e.g. for 80x86 with support for AVX-512 I think it's over 4 KiB of of stuff). CPU manufacturers often have mechanisms to avoid saving parts of that state if it wasn't changed, and to (optionally) postpone the loading of (pieces of) that state until its actually used (and avoid it completely if its not actually used). All of that requires kernel.
if it's a task switch and not just used for thread switches you might need some kind of "if virtual address space needs to change { change virtual address space }" on top of that
normally you want to keep track of statistics, like how much CPU time a thread has used. This requires some kind of "thread_info.time_used += now() - time_at_last_thread_switch;"; which gets difficulty/ugly when "process switching" is separated from "thread switching".
normally there's other state (e.g. pointer to thread local storage, special registers for performance monitoring and/or debugging, ...) that may need to be saved/loaded during thread switches. Often this state is not directly accessible in user code.
normally you also want to set a timer to expire when the thread has used too much time; either because you're doing some kind of "time multiplexing" (e.g. round-robin scheduler) or because its a cooperating scheduler where you need to have some kind of "terminate this task after 5 seconds of not responding in case it goes into an infinite loop forever" safe-guard.
this is just the low level task/thread switching in isolation. There is almost always higher level code to select a task to switch to, handle "thread used too much CPU time", etc.
Can someone link me to Linux source code line (say on github) handling context switch
Someone probably can't. It's not one line; it's many lines of assembly for each different architecture, plus extra higher-level code (for timers, support routines, the "select a task to switch to" code, for exception handlers to support "lazy SIMD state load", ...); which probably all adds up to something like 10 thousand lines of code spread across 50 files.
I also doubt why context switch between kernel threads requires changing to kernel mode. Aren't we already in kernel mode for first thread?
Yes; often you're already in kernel code when you find out that a thread switch is needed.
Rarely/sometimes (mostly only due to communication between threads belonging to the same process - e.g. 2 or more threads in the same process trying to acquire the same mutex/semaphore at the same time; or threads sending data to each other and waiting for data from each other to arrive) kernel isn't involved; and in some cases (which are almost always massive design failures - e.g. extreme lock contention problems, failure to use "worker thread pools" to limit the number of threads needed, etc) it's possible for this to be the dominant cause of thread switches, and therefore possible that doing thread switches in user space can be beneficial (e.g. as a work-around for the massive design failures).
Don't limit yourself to Linux or even UNIX, they are neither the first nor last word on systems or programming models. The synchronous execution model dates back to the early days of computing, and are not particularly well suited to larger scale concurrent and reactive programming.
Golang, for example, employs a great many lightweight user threads -- goroutines -- and multiplexes them on a smaller set of heavyweight kernel threads to produce a more compelling concurrency paradigm. Some other programming systems take similar approaches.
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.
I have this question related to memory barriers.
In a multi-threaded applications a memory barrier must be used if data is shared between them , because a write in a thread that is runing on one core , may not be seen by another thread on an another core.
From what I read from other explanations of memory barriers, it was said that if you have a single thread working with some data you don't need a memory barrier.
And here is my question: it could be the case that a thread modifies some data on a specific core, and then after some time the scheduler decides to migrate that thread to another core.
Is it possible that this thread will not see its modifications done on the other core?
In principle: Yes, if program execution moves from one core to the next, it might not see all writes that occurred on the previous core.
Keep in mind though that processes don't switch cores by themselves. It is the operating system that preempts execution and moves the thread to a new core. Thus it is also the operating system's responsibility to ensure that memory operations are properly synchronized when performing a context switch.
For you as a programmer this means that, as long as you are not trying work on a level where there is no SMP-aware OS (for instance, when you are trying to write your own OS or when working on an embedded platform without a fully-fledged OS), you do not need to worry about synchronization issues for this case.
The OS is responsible of memory coherency or visibility in additonal to memory ordering after a thread migration. a.k.a, below test always passes:
int a = A
/* migration here */
assert(a == A)
Suppose we have a dual-core machine with a mainstream, modern OS capable to utilize both the cores.
If I have two threads, P1 and Q1 within the same process, and they happen to commence creating child threads, say, P2 and Q2, at approximately the same machine cycle, will OS perform the thread creation concurrently?
I heard thread creation is expensive, so the question came forth...
Thanks in advance.
Any reasonably well designed OS can have multiple processors executing kernel code at the same time. Therefore some of the tasks involved in a thread creation can be happening concurrently. But there will be some necessary serialization to manipulate some shared data structures (e.g. allocating memory, inserting a newly created threat structure into a global list). The processors could contend for the same lock thereby reducing concurrency.
Systems/applications which make new threads so often that the overhead of thread creation actually matters are probably designed wrong (doing too little useful work in a thread relative to the startup time, and not taking advantage of the obvious optimization of reusing short-lived threads from a pool).
It will be sorta-concurrently. There are aspects of thread-creation that cannot proceed in parallel - it would be unfortunate if the kernel memory-manager allocated both threads the same stack!
Thread creation is sufficiently expensive that it's worth while avoiding doing it at all during an app. run, hence the popularity of thread pools. Long-running tasks that block can be threaded off and left for the life of the app - often this means that explicit thread termination, (awkward at best, almost impossible at worst, from user code), is not necessary.
I think developers continually start and stop threads because they like to think of them as 'functions', where you 'pass parameters' in at the start and 'return' results when the thread ends. Ths is not the best way of conceptualizing 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.