You know what!! This is better than hours of lectures. So crisp and to the damn point! 👏👏👏👏 Now I can start reading other references. Thanks for the video.
You Sir are an angel, I've been stuck trying to understand how to use threads in C for SOOOO LONG and this tutorial makes so much sense, i was actually able to code along and it works and i get it ! AAAAH I'm so relieved, it finally makes sense, THANK YOU SO MUCH !
I've been playing around with OpenMP in VS and it works great for optimizing loops for parallel execution, where the threads seem to execute in random order each time the program is run. However, after looking at the documentation a little more attentively, there are a number of command parameters associated with the pragmas that would make life easier rather than trying to play around delay loops, ordering and placements. Great intro to threading.
Cool, thanks for sharing that, I'm glad you enjoyed it! :-) I think I'm due to cover threading some more but there are so many topics I want to cover haha...
You can also get the return value, you just have to pass a pointer to the place where the return value of the function will be put at (the result is also statically a void pointer!) as the second parameter to pthread_join(): void* add(void* param) { long* result = (long*)malloc(sizeof(long)); *result = *(long*)param + 1; return result; } int main() { pthread_t thread1; long value = 10; pthread_create(&thread1, NULL, &add, &value); long* result; pthread_join(thread1, (void**)&result); // output pointer to a `long` pointer! printf("result: %ld ", *result); // 11 free(result); return 0; }
Why are you passing address as argument while calling function in pthread_create ? Why can't you directly pass the value ? Why (void*) & value1 ? Why not (void*) value1 ? Near 7:17
Good question Bharadwaj! Ultimately, we do that because it's the way that the pthread library was created. :-) One thing to keep in mind is that we don't call the function in pthread_create. What we're doing is calling pthread_create, and we give it as arguments the function to call, and a pointer to the value we want to use as an argument. pthread_create() will at some point later "call our function in a new thread", and at that time the function is supplied with that pointer value as an argument. Because the pointer is a void pointer, we can use it to essentially "pass anything we want" to the function.... e.g. a pointer to something like a struct or array that contains all the data needed by the function. Hopefully this helps! :-)
Hi. Just watched this video. Very very cool and interesting. I started learning C due to my interest in numerical methods. In the video you mention that certain hardware and software specs must be met, in order to run threads in C? can you elaborate on this please? it seems a regular PC won't be able to run multi-threads thank you !!
A regular PC would most likely be able to run multiple threads, as the 'pthread_t' type doesn't reference the actual CPU thread directly, but instead references OS threads, which are created on demand by the OS itself depending on things such as available execution resources such as RAM and CPU among other things. Most modern OS's can make up to tens of thousands of threads (depending on the hardware itself of course, on average hardware you would be having around 12), so not meeting hardware requirements shouldn't be a problem. But yeah, as long as the amount of OS threads created is less than or equal to the amount of physical threads you have, you should be able to run anything you want in parallel.
Hello, great tutorial! I just wanted to ask that as you explained difference between the concepts of concurrent and parallel execution at the beginning in the video, i suppose, to that explanation, at 12:12, you should have said that "running parallel" instead of "running concurrently". In my opninon, if they were to run concurrently, it could even take more than 4.5-5sec(run-time for single thread) because of context switch. Of course, i assume that your machine has multiple CPUs and multithreading-supporting software, as most modern PCs have today.
No, the description is accurate as it is, I said that they were running concurrently and added it looks like they were running in parallel too (i.e. due to the execution time): "We can see that the multithreaded program is running roughly twice as fast as the single threaded program, and that makes sense because we have two threads now running concurrently, and in this case it looks like they're running in parallel as well". All threads run concurrently, and in this case, on my machine, they are also running in parallel based on the execution time. Parallelism implies concurrency, it's a "subset" of concurrency, but concurrency does not imply parallelism. Maybe I'll make a video on concurrency vs parallelism though, it's an interesting topic. Concurrency won't help in a situation like this with a computation occurring in each threads, it could even be worse due to the context switch that you mention. But it can improve performance drastically if the threads are working on I/O bound operations for example, which is something I think a lot of people don't realize.
It is not clear to me why when pthread_join commands were together the program didn't stop in the first one until the computarion is finished and after that it went to the second pthread_join like shown if you put them immediately after the pthread_creates
That's a great question Subtain, I think I'm going to make a video on exactly this question. :-) We can use a struct to pass multiple parameters by using multiple members of the struct, like this example: www.cse.cuhk.edu.hk/~ericlo/teaching/os/lab/9-PThread/Pass.html.
12:15 - I wonder how is it possible for the program to run faster when it is executing code concurrently with only one core, the physics behind it doesn't make sense... Does code do indeed run faster simply due to concurrency with one core?
This is a really good question. :-) It's a big topic though, I definitely want to do a video on exactly this topic one day. For now though I can just point to some resources online such as these ones to explain things: stackoverflow.com/a/8514943 stackoverflow.com/questions/11662781/when-is-clone-and-fork-better-than-pthreads
Hi ! I just started this subject when I was reading about the "Dining philosophers problem" and how to avoid the deadlock , starvation in this case .... you'll learn about mutex, semaphore/signal , the synchronization process, a huge among of information to learn...the call system fonctions are very tricky when you think about them and at the same time it's a very exiting subject to be familiar with because it's full of suspense!!! .. Thanks Kevin and as always your videos are very amazing!!! @@PortfolioCourses
Anyone know how to use a C file that uses pthreads to work with Python's ctypes library? Anything that has the line "pthread_create" Python says it can't find it. Something to do with dependency...I don't get it.
Unfortunately I don't have a video on this topic. Hopefully someone else is able to help you out. Eventually I'd love to cover ctypes in Python. Sometimes putting the error message into Google along with the text "stackoverflow" or "reddit" can lead to helpful conversations online that help to resolve the issue.
@@PortfolioCourses that was a good idea to go to stack overflow, they pointed me in the right direction and I got it solved. I had to specify a path in Python to the dependencies in mingw. First time I had to do that, I had thought I was doing something wrong with gcc. I've used ctypes a fair bit as a way to speed up my python code and learn C at the same time. I've just passed arrays to it but haven't got into any of the more hairy stuff like making python objects into ctypes structs or w/e though. On stack overflow I came across a library I hadn't heard of called omp which does multiprocessing in C also. I've only been programming less than a year so I'd be interested in hearing from more experienced people the pros/cons between pthreads and omp.
@@PortfolioCourses I was playing with omp a little bit and it looks like instead of passing arguments in, you use the thread ID to differentiate what the threads do, and that in pthreads the ID just continues to count up and up each time you loop through the code that splits the program, but in omp it restarts the count each time. Not trying to talk you to death, just thought it might be interesting if you haven't messed with it.
I am a little bit confused now. Ur example runs concurrently and not parallel right? If yes, how I have to code that it runs parallel then or the os decides what to do and I can not control that? Great video anyways!
Yes it runs concurrently but not in parallel. To get parallel execution you would use another solution, for example: curc.readthedocs.io/en/latest/programming/OpenMP-C.html. :-)
I'm glad you enjoyed it John! :-) So prior to C++11 we just used the pthreads library in C++, but now there is the thread library we can use instead: www.geeksforgeeks.org/multithreading-in-cpp/.
How do i prevent memory leaks in Pthreads? The first time i execute the program, answer is right but if i add a while loop, it is providing incorrect values.Even after adding pthread_join, same thing is happening.
If we dynamically allocate memory using a pthread, either as part of creating the thread or in the function called by the thread, we need to use free() to free the dynamically allocated memory the same as usual. 🙂 In this example, we don't use any dynamically allocated memory. A while loop could produce incorrect results for a number of reasons, such as variables being re-used, etc.
Is it necessary to always "join the threads" or is it just for when you need to be sure a thread has finished before moving on? For instance, if i want s thread to continuously run until the program dies, would I ever really need to join it?
in fact , its gonna take more time compared to the single threaded version , create pthread is big overhead for this single program , user time and sys time has actually increased
The way user and sys time works in multithreaded programs isn't straightforward: stackoverflow.com/a/556411. In particular: "On a multi-processor machine, a multi-threaded process or a process forking children could have an elapsed time smaller than the total CPU time - as different threads or processes may run in parallel. ". So yes user time can go up, but the elapsed time can go down.
Here we again.. no satisfied with Async/await keywords, i need to know how the assembly make sense of a thread... the curiosity is very time consuming lmao
Oh you need to know how to implement pthreads, from scratch? I don't know how to do that sorry, that's getting into some lower level programming that I don't really know enough about to talk about off the top of my head. :-)
why my result turn to be single thread is faster than multi thread? running in WSL ubuntu ❯ gcc -o s main.c ❯ time ./s Add: 1 Add: 2 Add: 3 Add: 4 Add: 5 Add: 6 ./s 55.02s user 0.01s system 99% cpu 55.032 total ❯ gcc -o m main.c ❯ time ./m Add: 5 Add: 4 Add: 3 Add: 6 Add: 2 Add: 1 ./m 68.42s user 0.01s system 596% cpu 11.464 total
I just learned about threads about 4 days ago. You can’t believe how much better a 13+ min video is than my 1 hour lecture in explaining threads.
That's awesome feedback thank you, I'm glad you enjoyed the video Leo! :-)
Yes... yes I can...
university lectures are always overenginneered.
I believe it.
@@happygofishingunderengineered too as one might say.
We're studying this at university right now and I have to say this is very much useful. It's almost clear after the first watch only.
There are very few tutorials on this topic. Thank you for your work!
You're welcome Diana! 🙂
You know what!! This is better than hours of lectures. So crisp and to the damn point! 👏👏👏👏
Now I can start reading other references.
Thanks for the video.
You're welcome Safar, I'm glad you enjoyed the video! :-)
42 Student here, you helped me greatly to understand the concept of threads!
Just wanted to say that I found your channel today and you have done some amazing work on your channel
Thank you very much Jack! I'm glad you enjoy the content. :-D
I am listing this channel as a reference in my school project because I have learned a ton of c from here. Thank you!
You are better than my professor in terms of explaining things. Thank you !!!
You Sir are an angel, I've been stuck trying to understand how to use threads in C for SOOOO LONG and this tutorial makes so much sense, i was actually able to code along and it works and i get it ! AAAAH I'm so relieved, it finally makes sense, THANK YOU SO MUCH !
something that I particularly liked is .. the comparison you do at the end, this shows how useful multi-threading is
They way you explains with examples...Marvelous sir
Thanks for the kind words Junaid! I'm glad you enjoyed the video. 🙂
best ASCII presentation I've seen so far.
Thank you Peter! :-)
13mins video is more valuable than 2 hours lecture on my university :D
wanted to say that you've done some amazing work on your channel
Thank you very much! 😀
Defined greatly n precisely ✨ best single shortest video to learn whole fkn threads topic …
Thank you very much for the kind positive feedback Emmad! I'm glad you enjoyed the video. :-)
I've been playing around with OpenMP in VS and it works great for optimizing loops for parallel execution, where the threads seem to execute in random order each time the program is run. However, after looking at the documentation a little more attentively, there are a number of command parameters associated with the pragmas that would make life easier rather than trying to play around delay loops, ordering and placements. Great intro to threading.
Cool, thanks for sharing that, I'm glad you enjoyed it! :-) I think I'm due to cover threading some more but there are so many topics I want to cover haha...
you teach so well man, great job! :p
Thank you! :-)
I think i have clear understanding about pthreads post watching this video....thanks for the free tuts.
You're welcome Aditya, I'm glad the video was helpful for you! :-)
best explanation. needed for my exam.
cs288?
You can also get the return value, you just have to pass a pointer to the place where the return value of the function will be put at (the result is also statically a void pointer!) as the second parameter to pthread_join():
void* add(void* param) {
long* result = (long*)malloc(sizeof(long));
*result = *(long*)param + 1;
return result;
}
int main() {
pthread_t thread1;
long value = 10;
pthread_create(&thread1, NULL, &add, &value);
long* result;
pthread_join(thread1, (void**)&result); // output pointer to a `long` pointer!
printf("result: %ld
", *result); // 11
free(result);
return 0;
}
Best video ever !! on threads topic
Why are you passing address as argument while calling function in pthread_create ? Why can't you directly pass the value ?
Why (void*) & value1 ? Why not (void*) value1 ? Near 7:17
Good question Bharadwaj! Ultimately, we do that because it's the way that the pthread library was created. :-) One thing to keep in mind is that we don't call the function in pthread_create. What we're doing is calling pthread_create, and we give it as arguments the function to call, and a pointer to the value we want to use as an argument. pthread_create() will at some point later "call our function in a new thread", and at that time the function is supplied with that pointer value as an argument. Because the pointer is a void pointer, we can use it to essentially "pass anything we want" to the function.... e.g. a pointer to something like a struct or array that contains all the data needed by the function. Hopefully this helps! :-)
Multithreading it’s not about executing multiple instructions at the same time. It’s about sharing resource beetween them.
This video is a gem, thanks man!
You’re welcome Pietro, I’m glad you enjoyed it! :-)
man i love you thanks so much for this vid helped A LOT
You're very welcome, I'm so glad to hear the video helped you out! :-)
Really thank you for this great explanation
great video, so clear!
I'm glad you enjoyed it Julianna! :-)
nice explanation with easy example🙂
I’m glad you enjoyed it Manish! :-)
great and easy to follow explanation
Awesome, I'm glad you enjoyed it! :-)
Hi. Just watched this video. Very very cool and interesting. I started learning C due to my interest in numerical methods.
In the video you mention that certain hardware and software specs must be met, in order to run threads in C?
can you elaborate on this please?
it seems a regular PC won't be able to run multi-threads
thank you !!
A regular PC would most likely be able to run multiple threads, as the 'pthread_t' type doesn't reference the actual CPU thread directly, but instead references OS threads, which are created on demand by the OS itself depending on things such as available execution resources such as RAM and CPU among other things. Most modern OS's can make up to tens of thousands of threads (depending on the hardware itself of course, on average hardware you would be having around 12), so not meeting hardware requirements shouldn't be a problem.
But yeah, as long as the amount of OS threads created is less than or equal to the amount of physical threads you have, you should be able to run anything you want in parallel.
Did you make a video about semaphores also? My teacher is covering that right now and it might be helpful if there was a video about that also.
very well explained sir;
Hello, great tutorial! I just wanted to ask that as you explained difference between the concepts of concurrent and parallel execution at the beginning in the video, i suppose, to that explanation, at 12:12, you should have said that "running parallel" instead of "running concurrently". In my opninon, if they were to run concurrently, it could even take more than 4.5-5sec(run-time for single thread) because of context switch. Of course, i assume that your machine has multiple CPUs and multithreading-supporting software, as most modern PCs have today.
No, the description is accurate as it is, I said that they were running concurrently and added it looks like they were running in parallel too (i.e. due to the execution time): "We can see that the multithreaded program is running roughly twice as fast as the single threaded program, and that makes sense because we have two threads now running concurrently, and in this case it looks like they're running in parallel as well". All threads run concurrently, and in this case, on my machine, they are also running in parallel based on the execution time. Parallelism implies concurrency, it's a "subset" of concurrency, but concurrency does not imply parallelism. Maybe I'll make a video on concurrency vs parallelism though, it's an interesting topic. Concurrency won't help in a situation like this with a computation occurring in each threads, it could even be worse due to the context switch that you mention. But it can improve performance drastically if the threads are working on I/O bound operations for example, which is something I think a lot of people don't realize.
Mm 🎂🎂 mm mmm 🎂🎂🎂🎂 mm🎂mm9
🔥🔥🔥m9 🔥🔥🔥 mm 🔥m 🔥mm 🔥 okk 🔥🔥🔥🔥
99 😍 mom 😍 mm 😍 mm 😍mm 😍😍m 😍😍 mm 😍😍 mm. 9
🎉🎉 mo 🎉 mmmm 🎉m9 🎉🎉🎉 mmmm 🎉mm 🎉
thank you for the explanation I was only wondering why my single thread execution is faster in the example?
Thank you, super digestible!
It is not clear to me why when pthread_join commands were together the program didn't stop in the first one until the computarion is finished and after that it went to the second pthread_join like shown if you put them immediately after the pthread_creates
great explanation! Thanks
Good introduction! Thanks :)
You're welcome Audun! :-)
Amazing video mate
I’m glad you liked it! :-)
For some of us, we also have to add a flag to the linker: -lpthread.
What if we have to pass multiple parameters in a function?
That's a great question Subtain, I think I'm going to make a video on exactly this question. :-) We can use a struct to pass multiple parameters by using multiple members of the struct, like this example: www.cse.cuhk.edu.hk/~ericlo/teaching/os/lab/9-PThread/Pass.html.
@@PortfolioCourses thank you sir !
@@subtainmushtaq3237 You're welcome! :-D
12:15 - I wonder how is it possible for the program to run faster when it is executing code concurrently with only one core, the physics behind it doesn't make sense... Does code do indeed run faster simply due to concurrency with one core?
The code in the video is running in parallel in addition to running concurrently, that’s why it’s faster. :-)
Very well explained!
I'm glad you enjoyed it! :-)
What are the main differences between threads and the fork() system call? And why would you choose one over the other?
This is a really good question. :-) It's a big topic though, I definitely want to do a video on exactly this topic one day. For now though I can just point to some resources online such as these ones to explain things:
stackoverflow.com/a/8514943
stackoverflow.com/questions/11662781/when-is-clone-and-fork-better-than-pthreads
Hi ! I just started this subject when I was reading about the "Dining philosophers problem" and how to avoid the deadlock , starvation in this case .... you'll learn about mutex, semaphore/signal , the synchronization process, a huge among of information to learn...the call system fonctions are very tricky when you think about them and at the same time it's a very exiting subject to be familiar with because it's full of suspense!!! .. Thanks Kevin and as always your videos are very amazing!!! @@PortfolioCourses
Anyone know how to use a C file that uses pthreads to work with Python's ctypes library? Anything that has the line "pthread_create" Python says it can't find it. Something to do with dependency...I don't get it.
Unfortunately I don't have a video on this topic. Hopefully someone else is able to help you out. Eventually I'd love to cover ctypes in Python. Sometimes putting the error message into Google along with the text "stackoverflow" or "reddit" can lead to helpful conversations online that help to resolve the issue.
@@PortfolioCourses that was a good idea to go to stack overflow, they pointed me in the right direction and I got it solved. I had to specify a path in Python to the dependencies in mingw. First time I had to do that, I had thought I was doing something wrong with gcc.
I've used ctypes a fair bit as a way to speed up my python code and learn C at the same time. I've just passed arrays to it but haven't got into any of the more hairy stuff like making python objects into ctypes structs or w/e though.
On stack overflow I came across a library I hadn't heard of called omp which does multiprocessing in C also. I've only been programming less than a year so I'd be interested in hearing from more experienced people the pros/cons between pthreads and omp.
@@undeadpresident I'm glad you got that figured out. 🙂 Maybe someone else viewing the video can address your other question.
@@PortfolioCourses I was playing with omp a little bit and it looks like instead of passing arguments in, you use the thread ID to differentiate what the threads do, and that in pthreads the ID just continues to count up and up each time you loop through the code that splits the program, but in omp it restarts the count each time.
Not trying to talk you to death, just thought it might be interesting if you haven't messed with it.
@@undeadpresident Cool, thanks for sharing, I want to play around with more myself next month sometime.
I am a little bit confused now. Ur example runs concurrently and not parallel right? If yes, how I have to code that it runs parallel then or the os decides what to do and I can not control that?
Great video anyways!
Yes it runs concurrently but not in parallel. To get parallel execution you would use another solution, for example: curc.readthedocs.io/en/latest/programming/OpenMP-C.html. :-)
great example, very usefull ty
Another awesome video! I'm getting into threads now. Is multi-threading much different in C++? I'm sure they are very similar...
I'm glad you enjoyed it John! :-) So prior to C++11 we just used the pthreads library in C++, but now there is the thread library we can use instead: www.geeksforgeeks.org/multithreading-in-cpp/.
@@PortfolioCourses That's great info thanks!
@@lawniczakjohn you're welcome! 🙂
How do i prevent memory leaks in Pthreads? The first time i execute the program, answer is right but if i add a while loop, it is providing incorrect values.Even after adding pthread_join, same thing is happening.
If we dynamically allocate memory using a pthread, either as part of creating the thread or in the function called by the thread, we need to use free() to free the dynamically allocated memory the same as usual. 🙂 In this example, we don't use any dynamically allocated memory. A while loop could produce incorrect results for a number of reasons, such as variables being re-used, etc.
Is it necessary to always "join the threads" or is it just for when you need to be sure a thread has finished before moving on?
For instance, if i want s thread to continuously run until the program dies, would I ever really need to join it?
in fact , its gonna take more time compared to the single threaded version , create pthread is big overhead for this single program , user time and sys time has actually increased
The way user and sys time works in multithreaded programs isn't straightforward: stackoverflow.com/a/556411. In particular: "On a multi-processor machine, a multi-threaded process or a process forking children could have an elapsed time smaller than the total CPU time - as different threads or processes may run in parallel. ". So yes user time can go up, but the elapsed time can go down.
why dont you show how to get a return value from the thread function?
I think I may make more videos covering more aspects of pthreads as these videos have become popular. :-)
Very helpful, thank you
Hello , Thanks for the video. It is possible to pass 2 arguments to a thread?
Great question Moad! :-) We can if we use a struct as a way to pass both arguments: www.cse.cuhk.edu.hk/~ericlo/teaching/os/lab/9-PThread/Pass.html.
@@PortfolioCourses thanks for answering :)
Sir I respect you so much
Thank you for leaving this kind feedback! :-)
Wow . Thanks ❤
Thank you man.
Very useful.
I’m really glad you found it useful Nabanita! :-)
Thank you sir
You’re welcome Chaitanya! :-)
Hahn great material but I’m blind. How bout dark mode
Here we again.. no satisfied with Async/await keywords, i need to know how the assembly make sense of a thread... the curiosity is very time consuming lmao
there are so many ads in a 13 min video :(
Thank you for the feedback!
Bro pls use dark theme
please use dark mode !!! my eyes screaming haha
36236 Bashirian Summit
I hate it 😂 I need focus in library to understand how it implemented
Oh you need to know how to implement pthreads, from scratch? I don't know how to do that sorry, that's getting into some lower level programming that I don't really know enough about to talk about off the top of my head. :-)
@@PortfolioCourses yeah I believe it is system call so I downloaded the gun c library reference manual to figure out
Good luck! :-)
Hickle Mount
why my result turn to be single thread is faster than multi thread? running in WSL ubuntu
❯ gcc -o s main.c
❯ time ./s
Add: 1
Add: 2
Add: 3
Add: 4
Add: 5
Add: 6
./s 55.02s user 0.01s system 99% cpu 55.032 total
❯ gcc -o m main.c
❯ time ./m
Add: 5
Add: 4
Add: 3
Add: 6
Add: 2
Add: 1
./m 68.42s user 0.01s system 596% cpu 11.464 total
Thank you sir
You're welcome! :-)