I'm aware that User Level threads are created on the User Mode( no privileges) and Kernel threads are created in the Kernel Mode( privileged).
I am also aware that Processor threads are hardware threads that operate on Kernel Threads( I hope I am correct by putting it in this way)
Here is my confusion:-
User Level threads are not recognized by the OS as they are created, maintained and destroyed on the User Level. The OS doesn't see a multithreaded process from the User Mode as being multithreaded. It treats it as a single threaded process. Therefore, this program cannot take advantage of Multiprocessing, I guess it cannot take advantage of hyperthreading as well since it appears as single threaded in the OS.
So what's the use of Multithreading in this case? I mean the computation time will still be the same🤷♂️.
The last question is, do POSIX thread API and OPenMP create user level threads or Kernel threads?
I know what both libraries are, please don't explain about that.
If none creates Kernel threads then how do we create a multithreaded program that takes advantage of multiprocessing?
...what's the use of Multithreading in this case?
Multithreading is older than multiprocessing. Multithreading is one model of concurrent computing. That is to say, it's a way to write a computer program in which different activities are allowed to happen independently from each other. A classic example is a multi-user network server that creates a new thread for each connected client. Each thread can talk to its own client in a simple, synchronous way even though there may be no synchrony between what the different clients want to do. You don't need to have multiple CPUs for that.
When multi-CPU computers were invented, using multiple threads to exploit them for parallel processing was a natural and obvious choice.
I mean the computation time [for a green-threaded program that cannot exploit multiple CPUs] will still be the same.
That is true, but depending on what the different activities are that the program performs concurrently, the multi-threaded version of it may be easier to read and understand* than a program that's built around a different model of concurrency.
The reason is, we all were taught to write single-threaded, synchronous code when we were beginners. We understood that we were writing instructions that "the computer" would follow. We now say "a thread" instead of saying "the computer," but otherwise, the code executed by each thread can be mostly similar to the style of code that we wrote as beginners.
Part of what makes it so simple is, that the state of each of the concurrent activities can be mostly implicit in the contexts and the local variables (i.e., the stacks) of the different threads. If you choose a different model of concurrency (e.g., an event driven model) then you may have to explicitly represent more of that state with (maybe complex) data structures.
* Easier to read but not necessarily easier to write without making subtle mistakes. But, when I started working with large teams of software developers, they taught me that I'd be reading about ten lines of code for every one line that I wrote, so "easier to read but harder to write" turns out to be a win in the long run.
Pure user level threads are (as you pointed out) not a lot of use because they don't allow you to exploit the processing capability of multiple cores within a process.
The flip-side is that pure kernel level threads will typically incur substantial overheads when switch between threads. (There are things that you can do to deal with that, but ... that's another topic.) But the upshot is that the overheads make it inefficient to preform small tasks (units of work) using kernel level threads.
Another alternative to both is a hybrid of user level and kernel level threads. For example, suppose:
each process has one kernel level thread for each physical core,
each kernel level thread can switch between a bunch of user level threads and,
switching between a user level threads is handled by a scheduler in user space.
The Java Loom project is developing a new threading model (roughly) along those lines. Classic Java threads are still kernel level threads. New virtual threads are user level threads. A Java program gets to choose whether it uses classic or virtual threads ... or both.
There is a lot of material on Loom on the web; e.g.
https://blogs.oracle.com/javamagazine/post/java-loom-virtual-threads-platform-threads
https://www.infoq.com/news/2022/05/virtual-threads-for-jdk19/
https://wiki.openjdk.org/display/loom/Main
Loom is likely to be part of the next Java release: Java 19.
I'm pretty sure that (C / C++) POSIX threads are kernel level. I don't know about OpenMPI threads, but I'd expect they are kernel level too. (They wouldn't be fit for purpose as pure user level threads.)
I have heard of hybrid threading models for C / C++, though I don't know about actual implementations. Look for articles, etcetera that talk about Threads vs Fibres.
Related
Suppose we have a process with multiple threads in a uniprocessor.
Now I know that if we have several processes, only one of them will be processed at a time in a uniprocessor and hence the processes are not concurrent.
If my understanding is correct, similarly each thread will be processed at a time and not concurrent in a uniprocessor. Is this statement true? If so then does multithreading mean having more than one thread in a process and does not mean running multiple threads at a time? And does that mean there's no benefit of creating user threads in a uniprocessor environment?
TL;DR: threads are switching more often than processes and in real time we have an effect of concurrency because it is happens really fast.
when you wrote:
each thread will be processed at a time and not concurrent in a uni processor
Notice the word "concurrent", there is no real concurrency in uni processor, there is only effect of that thanks to the multiple number of context switches between processes.
Let's clarify something here, the single core of the CPU can handle one thread at a given time, each process has a main thread and (if needed) more threads running together. If a process A is now running and it has 3 threads: A1(main thread), A2, A3 all three will be running as long as process A is being processed by the CPU core. When a context switch occur process A is no longer running and now process B will run with his threads.
About this statement:
there's no benefit of creating user threads in a uni processor environment
That is not true. there is a benefit in creating threads, they are easier to create ("spawn" as in the books) and shearing the process heap memory. Creating a sub process ("child" as in the books) is a overhead comparing to a thread because a process need to have his own memory. For example each google chrome tab is a process not a thread, but this tab has multiple threads running concurrency with little responsibility.
If you are still somehow running a computer with just one, single-core, CPU, then you would be correct to observe that only one thread can be physically executing at one time. But that does not negate the value of breaking up the application into multiple threads and/or processes.
The essential benefit is concurrency. When one thread is waiting (e.g. for an input/output operation to complete), there is something else for the CPU to be doing in the meantime: it can be running a different thread that isn't waiting. With a carefully designed application, you can get much better utilization of every part of the hardware, more parallelism, and thus, more throughput.
My favorite go-to example is a fast food restaurant. About a dozen workers, each one doing different things, cooperate to bring your order to you. Even if one of them (say, "the fry guy") is standing around, someone else always has something to do. Several orders are in-process at once. This overlap, this "concurrency," is what you are shooting for – regardless of how many CPUs you have.
Multithreading is also commonly used with GUI applications that also need to do some kind of "heavy lifting." One thread handles the GUI interaction (and has no other real responsibilities) while other threads, with a slightly inferior priority (or "niceness") do the lifting. When a GUI event comes in, the GUI thread pre-empts the others and responds to it immediately, then of course goes right back to sleep again. But in this way the GUI always remains very responsive – even though the other threads are doing "heavy lifting" things, GUI messages are still handled very promptly. (I scooped-up about a 25% performance improvement by re-tooling an older application to use this approach, because the application was no longer "polling" for GUI events.)
The first question I ask about any thread is, "what does it wait for?" To me, a thread is defined by what event it waits for and what it does when that event happens.
Threads were in wide-spread use for at least a decade before multi-processor computers became commercially available. They are useful when you want to write a program that has to respond to un-synchronized events that come from multiple different sources. There's a few different ways to model a program like that. One way is to have a different thread to wait on each different event source. The next most popular is an event driven architecture in which there's a main loop that waits for all events and calls different event handler functions for each of the different kinds of event.
The multi-threaded style of program often is easier to read* because there's usually different activities going on inside the program, and the state of each activity can be implicit in the context (i.e., registers and call stack) of the thread that's driving it, while in the event-driven model, each activity's state must be explicitly encoded in some object.
The implicit-in-the-context way of keeping the state is much closer to the procedural style of coding a single activity that we learn as beginners.
*Easier to read does not mean that the code is easy to write without making bad and non-obvious mistakes!!
The main impetus for developing threads was Ada compliance. Prior to that, different operating systems had their own ways of handing multiple things at once. In eunuchs, the way to do more than one thing was to spin off a new process. In VMS, software interrupts (aka Asynchronous System Traps or Asynchronous Procedure Calls in Windoze). In those days (1970's) multiprocessor systems were rare.
One of the goals of Ada was to have a system independent way of doing things. It adopted the "task" which is effectively a thread. In order to support Ada, compiler developers had to include task (thread) libraries.
With the rise of multiprocessors, operating systems started to make threads (rather than processes) the basic schedulable unit in a system.
Threads then give a way for programs to handle multiple things simultaneously, even if there is only one processor. Sadly, support for threads in programming languages has been woefully lacking. Ada is the only major language I can think of that has real support for threads (tasks). Thread support in Java, for example, is a complete, sick joke. The result is threads are not as effective in practice as they could be.
I've been playing with the Linux kernel recently and diving back into the days of OS courses from college.
Just like back then, I'm playing around with threads and the like. All this time I had been assuming that threads were automatically running concurrently on multiple cores but I've recently discovered that you actually have to explicitly code for handling multiple cores.
So what's the point of multi-threading on a single core? The only example I can think of is from college when writing a client/server program but that seems like a weak point.
All this time I had been assuming that threads were automatically
running concurrently on multiple cores but I've recently discovered
that you actually have to explicitly code for handling multiple cores.
The above is incorrect for any widely used, modern OS. All of Linux's schedulers, for example, will automatically schedule threads on different cores and even automatically move threads from one core to another when necessary to maximize core usage. There are some APIs that allow you to modify the schedulers' behavior, but these APIs are generally used to disable automatic thread-to-core scheduling, not to enable it.
So what's the point of multi-threading on a single core?
Imagine you have a GUI program whose purpose is to execute an expensive computation (for example, render a 3D image or a Mandelbrot set) and then display the result. Let's say this computation takes 30 seconds to complete on this particular CPU. If you implement that program the obvious way, and use only a single thread, then the user's GUI controls will be unresponsive for 30 seconds while the calculation is executing -- the user will be unable to do anything with your program, and possibly unable to do anything with his computer at all. Since users expect GUI controls to be responsive at all times, that would be a poor user experience.
If you implement that program with two threads (one GUI thread and one rendering thread), on the other hand, the user will be able to click buttons, resize the window, quit the program, choose menu items, etc, even while the computation is executing, because the OS is able to wake up the GUI thread and allow it to handle mouse/keyboard events when necessary.
Of course, it is possible to write this program with a single thread and keep its GUI responsive, by writing your single thread to do just a few milliseconds worth of computation, then check to see if there are GUI events available to process, handling them, then going back to do a bit more computation, etc. But if you code your app this way, you are essentially writing your own (very primitive) thread scheduler inside your app anyway, so why reinvent the wheel?
The first versions of MacOS were designed to run on a single core, but had no real concept of multithreading. This forced every application developer to correctly implement some manual thread management -- even if their app did not have any extended computations, they had to explicitly indicate when they were done using the CPU, e.g. by calling WaitNextEvent. This lack of multithreading made early (pre-MacOS-X) versions of MacOS famously unreliable at multitasking, since just one poorly written application could bring the whole computer to a grinding halt.
First, a program not only computes, but also waits for input/output and so can be considered as executing on an I/O processor. So even single-core machine is a multi-processor machine, and employing of multi-threading is justified.
Second, a task can be divided in several threads in the sake of modularity.
Multithreading is not only for taking advantage of multiple cores.
You need multiple processes for multitasking. For similar reason you are allowed to have multiple threads, which are lightweight compared with processes.
You probably don't want to spawn processes all the time for things like blocking I/O. That may be overkill.
And there is fiber, which is even more lightweight. So we have process, thread, and fiber for different levels of needs.
Well, when you say multithreading on a single core, there are things you need to consider. For example, the thread API that you are using - is it user level or kernel level. Most probably from you question I believe you are using user level threads.
Now, user level threads, depending upon the host OS or the API itself may map to single kernel thread or multiple. Many relations are possible like 1-1,many-1 or many-many.
Now, if there is a single core, your OS can still provide you several Kernel level threads which may behave as multiple processes to the CPU. In which case, OS will give you a time-slicing (and multi-programming) on the kernel threads leading to superfast context switch and via the user level API - you/your code will seem to have multithreaded features.
Also note that eventhough your processor is a single core, depending on the make, it can be hyperthreaded and have super deep pipelines allowing the concurrent running of Kernel threads with very low overhead.
For references: Check Intel/AMD architecture and how various OS provide Kernel threads.
What are the main advantages of using a model for concurrency based on processes over one
based on threads and in what contexts is the latter appropriate?
Fault-tolerance and scalability are the main advantages of using Processes vs. Threads.
A system that relies on shared memory or some other kind of technology that is only available when using threads, will be useless when you want to run the system on multiple machines. Sooner or later you will need to communicate between different processes.
When using processes you are forced to deal with communication via messages, for example, this is the way Erlang handles communication. Data is not shared, so there is no risk of data corruption.
Another advantage of processes is that they can crash and you can feel relatively safe in the knowledge that you can just restart them (even across network hosts). However, if a thread crashes, it may crash the entire process, which may bring down your entire application. To illustrate: If an Erlang process crashes, you will only lose that phone call, or that webrequest, etc. Not the whole application.
In saying all this, OS processes also have many drawbacks that can make them harder to use, like the fact that it takes forever to spawn a new process. However, Erlang has it's own notion of processes, which are extremely lightweight.
With that said, this discussion is really a topic of research. If you want to get into more of the details, you can give Joe Armstrong's paper on fault-tolerant systems]1 a read, it explains a lot about Erlang and the philosophy that drives it.
The disadvantage of using a process-based model is that it will be slower. You will have to copy data between the concurrent parts of your program.
The disadvantage of using a thread-based model is that you will probably get it wrong. It may sound mean, but it's true-- show me code based on threads and I'll show you a bug. I've found bugs in threaded code that has run "correctly" for 10 years.
The advantages of using a process-based model are numerous. The separation forces you to think in terms of protocols and formal communication patterns, which means its far more likely that you will get it right. Processes communicating with each other are easier to scale out across multiple machines. Multiple concurrent processes allows one process to crash without necessarily crashing the others.
The advantage of using a thread-based model is that it is fast.
It may be obvious which of the two I prefer, but in case it isn't: processes, every day of the week and twice on Sunday. Threads are too hard: I haven't ever met anybody who could write correct multi-threaded code; those that claim to be able to usually don't know enough about the space yet.
In this case Processes are more independent of eachother, while Threads shares some resources e.g. memory. But in a general case Threads are more light-weight than Processes.
Erlang Processes is not the same thing as OS Processes. Erlang Processes are very light-weight and Erlang can have many Erlang Processes within the same OS Thread. See Technically why is processes in Erlang more efficient than OS threads?
First and foremost, processes differ from threads mostly in the way their memory is handled:
Process = n*Thread + memory region (n>=1)
Processes have their own isolated memory.
Processes can have multiple threads.
Processes are isolated from each other on the operating system level.
Threads share their memory with their peers in the process.
(This is often undesirable. There are libraries and methods out there to remedy this, but that is usually an artificial layer over operating system threads.)
The memory thing is the most important discerning factor, as it has certain implications:
Exchanging data between processes is slower than between threads. Breaking the process isolation always requires some involvement of kernel calls and memory remapping.
Threads are more lightweight than processes. The operating system has to allocate resources and do memory management for each process.
Using processes gives you memory isolation and synchronization. Common problems with access to memory shared between threads do not concern you. Since you have to make a special effort to share data between processes, you will most likely sync automatically with that.
Using processes gives you good (or ultimate) encapsulation. Since inter process communication needs special effort, you will be forced to define a clean interface. It is a good idea to break certain parts of your application out of the main executable. Maybe you can split dependencies like that.
e.g. Process_RobotAi <-> Process_RobotControl
The AI will have vastly different dependencies compared to the control component. The interface might be simple: Process_RobotAI --DriveXY--> Process_RobotControl.
Maybe you change the robot platform. You only have to implement a new RobotControl executable with that simple interface. You don't have to touch or even recompile anything in your AI component.
It will also, for the same reasons, speed up compilation in most cases.
Edit: Just for completeness I will shamelessly add what the others have reminded me of :
A crashing process does not (necessarily) crash your whole application.
In General:
Want to create something highly concurrent or synchronuous, like an algorithm with n>>1 instances running in parallel and sharing data, use threads.
Have a system with multiple components that do not need to share data or algorithms, nor do they exchange data too often, use processes. If you use a RPC library for the inter process communication, you get a network-distributable solution at no extra cost.
1 and 2 are the extreme and no-brainer scenarios, everything in between must be decided individually.
For a good (or awesome) example of a system that uses IPC/RPC heavily, have a look at ros.
I'm reading up on concurrency. I've got a bit over my head with terms that have confusingly similar definitions. Namely:
Processes
Threads
"Green threads"
Protothreads
Fibers
Coroutines
"Goroutines" in the Go language
My impression is that the distinctions rest on (1) whether truly parallel or multiplexed; (2) whether managed at the CPU, at the OS, or in the program; and (3..5) a few other things I can't identify.
Is there a succinct and unambiguous guide to the differences between these approaches to parallelism?
OK, I'm going to do my best. There are caveats everywhere, but I'm going to do my best to give my understanding of these terms and references to something that approximates the definition I've given.
Process: OS-managed (possibly) truly concurrent, at least in the presence of suitable hardware support. Exist within their own address space.
Thread: OS-managed, within the same address space as the parent and all its other threads. Possibly truly concurrent, and multi-tasking is pre-emptive.
Green Thread: These are user-space projections of the same concept as threads, but are not OS-managed. Probably not truly concurrent, except in the sense that there may be multiple worker threads or processes giving them CPU time concurrently, so probably best to consider this as interleaved or multiplexed.
Protothreads: I couldn't really tease a definition out of these. I think they are interleaved and program-managed, but don't take my word for it. My sense was that they are essentially an application-specific implementation of the same kind of "green threads" model, with appropriate modification for the application domain.
Fibers: OS-managed. Exactly threads, except co-operatively multitasking, and hence not truly concurrent.
Coroutines: Exactly fibers, except not OS-managed.
Goroutines: They claim to be unlike anything else, but they seem to be exactly green threads, as in, process-managed in a single address space and multiplexed onto system threads. Perhaps somebody with more knowledge of Go can cut through the marketing material.
It's also worth noting that there are other understandings in concurrency theory of the term "process", in the process calculus sense. This definition is orthogonal to those above, but I just thought it worth mentioning so that no confusion arises should you see process used in that sense somewhere.
Also, be aware of the difference between parallel and concurrent. It's possible you were using the former in your question where I think you meant the latter.
I mostly agree with Gian's answer, but I have different interpretations of a few concurrency primitives. Note that these terms are often used inconsistently by different authors. These are my favorite definitions (hopefully not too far from the modern consensus).
Process:
OS-managed
Each has its own virtual address space
Can be interrupted (preempted) by the system to allow another process to run
Can run in parallel with other processes on different processors
The memory overhead of processes is high (includes virtual memory tables, open file handles, etc)
The time overhead for creating and context switching between processes is relatively high
Threads:
OS-managed
Each is "contained" within some particular process
All threads in the same process share the same virtual address space
Can be interrupted by the system to allow another thread to run
Can run in parallel with other threads on different processors
The memory and time overheads associated with threads are smaller than processes, but still non-trivial
(For example, typically context switching involves entering the kernel and invoking the system scheduler.)
Cooperative Threads:
May or may not be OS-managed
Each is "contained" within some particular process
In some implementations, each is "contained" within some particular OS thread
Cannot be interrupted by the system to allow a cooperative peer to run
(The containing process/thread can still be interrupted, of course)
Must invoke a special yield primitive to allow peer cooperative threads to run
Generally cannot be run in parallel with cooperative peers
(Though some people think it's possible: http://ocm.dreamhosters.com/.)
There are lots of variations on the cooperative thread theme that go by different names:
Fibers
Green threads
Protothreads
User-level threads (user-level threads can be interruptable/preemptive, but that's a relatively unusual combination)
Some implementations of cooperative threads use techniques like split/segmented stacks or even individually heap-allocating every call frame to reduce the memory overhead associated with pre-allocating a large chunk of memory for the stack
Depending on the implementation, calling a blocking syscall (like reading from the network or sleeping) will either cause a whole group of cooperative threads to block or implicitly cause the calling thread to yield
Coroutines:
Some people use "coroutine" and "cooperative thread" more or less synonymously
I do not prefer this usage
Some coroutine implementations are actually "shallow" cooperative threads; yield can only be invoked by the "coroutine entry procedure"
The shallow (or semi-coroutine) version is easier to implement than threads, because each coroutine does not need a complete stack (just one frame for the entry procedure)
Often coroutine frameworks have yield primitives that require the invoker to explicitly state which coroutine control should transfer to
Generators:
Restricted (shallow) coroutines
yield can only return control back to whichever code invoked the generator
Goroutines:
An odd hybrid of cooperative and OS threads
Cannot be interrupted (like cooperative threads)
Can run in parallel on a language runtime-managed pool of OS threads
Event handlers:
Procedures/methods that are invoked by an event dispatcher in response to some action happening
Very popular for user interface programming
Require little to no language/system support; can be implemented in a library
At most one event handler can be running at a time; the dispatcher must wait for a handler to finish (return) before starting the next
Makes synchronization relatively simple; different handler executions never overlap in time
Implementing complex tasks with event handlers tends to lead to "inverted control flow"/"stack ripping"
Tasks:
Units of work that are doled out by a manager to a pool of workers
The workers can be threads, processes or machines
Of course the kind of worker a task library uses has a significant impact on how one implements the tasks
In this list of inconsistently and confusingly used terminology, "task" takes the crown. Particularly in the embedded systems community, "task" is sometimes used to mean "process", "thread" or "event handler" (usually called an "interrupt service routine"). It is also sometimes used generically/informally to refer to any kind of unit of computation.
One pet peeve that I can't stop myself from airing: I dislike the use of the phrase "true concurrency" for "processor parallelism". It's quite common, but I think it leads to much confusion.
For most applications, I think task-based frameworks are best for parallelization. Most of the popular ones (Intel's TBB, Apple's GCD, Microsoft's TPL & PPL) use threads as workers. I wish there were some good alternatives that used processes, but I'm not aware of any.
If you're interested in concurrency (as opposed to processor parallelism), event handlers are the safest way to go. Cooperative threads are an interesting alternative, but a bit of a wild west. Please do not use threads for concurrency if you care about the reliability and robustness of your software.
Protothreads are just a switch case implementation that acts like a state machine but makes implementation of the software a whole lot simpler. It is based around idea of saving a and int value before a case label and returning and then getting back to the point after the case by reading back that variable and using switch to figure out where to continue. So protothread are a sequential implementation of a state machine.
Protothreads are great when implementing sequential state machines. Protothreads are not really threads at all, but rather a syntax abstraction that makes it much easier to write a switch/case state machine that has to switch states sequentially (from one to the next etc..).
I have used protothreads to implement asynchronous io: http://martinschroder.se/asynchronous-io-using-protothreads/
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