Menu
I've been reading a lot about how parallelism is different than concurrency, but I'm having a lot of questions like how can I achieve a parallelism with a programming language like Java?
What I understood so far is with a VM like Java we can only achieve concurrency and parallelism is not in scope of JVM, it's threads mapped to ( I'm confused here ) OS threads or Kernel threads?
To understand the multi-threading one needs to understand how the threads of a VM ( JVM ) or a process are actually mapped down to the CPU hardware. Please help here.
Correct me if I'm wrong, here are few things I think I understand right:
1. CPU threads ( Intel hyper-threading ): Hardware threads provided by adding more registers in CPU having same ALU, so do not actually behave as a dual-CPU ( if 2 threads are there in CPU ) and a "just"-concurrent use of ALU can just mask ( hide by an interleaved MUXing ) a principal inability to deliver indeed a true parallel code-execution.
2. Kernel threads: Started by Kernel and these can be more in number than CPU threads but are a bit expensive to start.
3. OS threads: I don't understand it. I'm guessing this is what you mean when you call native threads and probably same as previous #2 ie Kernel threads.
4. User threads: Just like new Thread() in JVM or .NET. I think they are not recognized as threads by OS so nothing much can be done here from the parallelism perspective.
I need to understand how threads created in #4 are mapped to #1, may be with examples to Windows/Linux/MacOS.
Guys, I understand how wide this question can be but I can live with brief info and link to any detailed documentation. Thanks in advance.
533
Threads are individual sequences of instructions. Old/small CPUs only do one instruction at a time hence support “single threaded” execution, newer multicore CPUs can handle multiple threads executing at the same time.
Difference between User-Level & Kernel-Level Thread User-level threads are faster to create and manage. Kernel-level threads are slower to create and manage. Implementation is by a thread library at the user level. Operating system supports creation of Kernel threads.
A kernel thread is a kernel entity, like processes and interrupt handlers; it is the entity handled by the system scheduler. A kernel thread runs within a process, but can be referenced by any other thread in the system.
There are two types of threads to be managed in a modern system: User threads and kernel threads.
1. Intel Hyperthreading - It is a technology developed by Intel to improve the efficiency of one core, because in a lot of its execution time, the CPU waits for retrieving data & code from RAM. The operating system views hyperthreading-cores as two separate logical processors, and each contains separate registers.
2. Kernel Threads - This are managed by the OS-kernel and are scheduled in kernel-mode only. They are attached to registers & EIP along with some attributes. These are created and managed by the process by using some "defined" interfaces by the kernel.
3. User Threads - These are created by the userspace process are executed on a kernel thread (because only a kernel one can be scheduled). But multiple user-threads can take turns for executing on a kernel thread, and giving higher performance for 'temporary' threads that go & come, due to time saved by not doing a system call.
4. OS Threads - OS threads don't have a proper definition in my knowledge. They can be threads used by the OS, or just kernel threads.
There are multiple models by which user-threads execute on kernel-threads - one-to-one, m-to-n & all-to-one, you can google these.
156
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With
