parallalization and taskset

[Marcofras]

is it possible have parallalization with taskset?

[Hu]

Why are you using taskset at all? If you want four instances, run four instances:

[Marcofras]

With your method i have no guarante its parallelized in different core with taskset instead i have guarante. Or not?

[Hu]

With taskset, you guarantee that each task either runs on the specified core or, if that core is not available, does not run at all. Without taskset, you let the scheduler pick which core to run each one on, and it will try to run as many things in parallel as you have cores available. Why do you think you need to force particular core assignments for particular processes?

[Marcofras]

Are you sure that its sufficient & for have parallalization?
I read somewhere that its not sure this

[Hu]

& tells the shell to place the command in the background. Where did you read that this would not work? Have you tried the command I gave you? What went wrong?

[Marcofras]

Im under c++ and that command no works, i use a series of system(./desc &) or something similar is it correct? Will it works?

[Hu]

If you are using system from C++, why is your opening post expressed as shell? Rather than asking if something will work after I told you it will, why not try it and report whether it works?

I think you should provide a Minimum Reproducible Example demonstrating your problem.

[Marcofras]

Indeed i was wrong im tryng in shell before c++, dont you think its the same? I dont make c++ code yet, do you think system(./desc &) will work and save time with parallelization i wish?

[Hu]

I gave you a loop that should work. I later reiterated that I expect it to work. Yes, system should work fine, since, according to its manual page, that is delegating back to the shell, which is almost always the wrong thing to do in a C++ program. If you need further help, please demonstrate a specific problem.

[Marcofras]

I will make the code c++ in days.
Why its wrong thing which is the correct form? Maybe thread or something similar?

[GDH-gentoo]

Because a C++ program can do things like spawning processes by using the libc directly (or through some convenience library) instead of invoking the shell.
_________________

[szatox]

Marcofras, you're bouncing around too much. Pause for a minute and focus.
What are you _actually_ trying to do?
_________________
Make Computing Fun Again

[Marcofras]

parallalization, because i had no success with thread (in the past) im opting to divide the program in several subprograms and run parallelized, it seems more easy.

[szatox]

Eh... This is clearly a case of https://en.wikipedia.org/wiki/XY_problem.
Regarding your original question:
> is it possible have parallalization with taskset?
The only correct answer is: "Yes but no". Unfortunately, on top of being correct, it is also completely useless.

Parallelization where? Why did you start with shell and then you're suddenly going to cpp? This part is confusing.
More importantly: What is the thing you're trying to write there? What is the problem you're trying to solve? What is standing in your way?

> I think you should provide a Minimum Reproducible Example demonstrating your problem.
Yeah, I second that. Although this technique is primarily used for reporting bugs, it does help with communication in general.
Overkill is still a kill.
_________________
Make Computing Fun Again

[Hu]

Yes, a Minimum Reproducible Example might be overkill here. I asked for it because OP's questions implied to me that he had directly experienced shell background not to exhibit the desired parallelism that I had stated would work. If he was experiencing that lack, we could try to understand why. If he has not tried the solutions given to him, I would like for him to do so or explain why not (particularly since this should be easy to test, due to being independent of the content of ./desc), before we put more work into assuring him again that his doubts are misplaced.

[Marcofras]

the original problem is i need parallalization for save time, i tried thread in desc that parallelize the function shap but it doesn' t work and freeze in ruuntime

[Hu]

That does not appear to be a MRE. None of those identifiers are defined, and it is not legal to use a statement at global scope to call join. A Minimal Reproducible Example does not need to be your full and secret code, and by its very definition, generally should not be. We want it to be Minimal. Sharing your full secret code is very likely not minimal.

What I told you is that using system from C++ is "almost always the wrong thing to do." GDH-gentoo then elaborated my point for you.

Putting all that together, I will again ask you to show us a MRE that exhibits your problem. As a counter to your claim that this hangs, I offer this MRE that is not defective. It exactly mirrors the fragment you showed, and fills in the missing bits as needed to make it compile.

[Marcofras]

i made the same thing of you but it freeze

[szatox]

Works for me.
Adding a 1 sec sleep shows the parallel execution even better. If we delay print, we get to see the order randomized by scheduler due to system load:

[Marcofras]

the order is not important in my case i want use system() and shell its more easy for me.
suggestions?

[GDH-gentoo]

[Genone]

Have you looked into just using fork() for your issue?

Using system() is almost always a bad idea.

[Marcofras]

i dont knew fork i read it can call a program from the filesystem, will it be parallellized with time saving?
i dont understand well how to use:

[pingtoo]