Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Gantolandon]

About multithreading: I don't think it would be even a good idea to apply it to pathfinding. Even if you ensured every thread which calculate its path will be treated equally by the processor, more calculations would still require more time. It means that units with less complicated path would begin (and end) their movements earlier.

Of course, pathfinding needs some tweaking, but I don't think it's multithreading which will solve the problem...

[Tylui]

About multithreading: I don't think it would be even a good idea to apply it to pathfinding. Even if you ensured every thread which calculate its path will be treated equally by the processor, more calculations would still require more time. It means that units with less complicated path would begin (and end) their movements earlier.

Of course, pathfinding needs some tweaking, but I don't think it's multithreading which will solve the problem...

O_O

What?  I didn't understand any of that, but I got that you don't understand multithreading at all, it seems.

Multithreading will improve pathfinding related FPS drop issures, because it will use a different core of the processor(or a different processor) that wasn't being used before.  That means that you'd get two processors calculating things, instead of just one.  That means performance boost.

[Gantolandon]

What?  I didn't understand any of that, but I got that you don't understand multithreading at all, it seems.

Multithreading will improve pathfinding related FPS drop issures, because it will use a different core of the processor(or a different processor) that wasn't being used before.  That means that you'd get two processors calculating things, instead of just one.  That means performance boost.

Multithreading is not about transferring some tasks to another core. It's effectively like creating another process to work on a specific task. Then the processor can switch between threads. The main point of multithreading is that you can avoid locking an entire process until it completes one particular task.

What I thought you propose to do, is splitting pathfinding between several threads. It seems I misunderstood you. Splitting the process between two or more cores would indeed improve efficiency.

[Tylui]

Indeed, there must be two or more types of multithread.  I understand your original post now! 

[Gantolandon]

Indeed, there must be two or more types of multithread.

Indeed, multithreading can be applied in two or more ways. And it was used even before multiple cores became a standard. So why the sarcasm?

[Tylui]

Indeed, there must be two or more types of multithread.

Indeed, multithreading can be applied in two or more ways. And it was used even before multiple cores became a standard. So why the sarcasm?

Sarcasm?  I... wasn't being sarcastic.  I was stating that there must be two kinds of multithreading...  And knowing that, it makes your first post understandable.  I really wasn't being sarcastic. O.o

[Gantolandon]

OK, I'm sorry then. I tend to misunderstand people sometimes.

Multithreading is just creating threads. It can be used to get the program to work with multiple cores. It's just something I didn't think of when I had read about multithreading in DF.

[Draco18s]

I don't have enough time to monitor this community to the point that it irritates me, and if you do something is wrong.

You're missing the point of restricting making new threads: noise. This thread isn't noise in that has a clear title that makes it very easily to dismiss

Only if I know what CUDA is, which I don't (even after this thread I still have no idea what it stands for, nor do I care).

[Tylui]

CUDA ("Compute Unified Device Architecture")

It's right there, good sir.  In the OP. ^_^

It's a language that allows the GPU to take over computing certain portions of code.  Or something like that.  But it only works with nVidia 8 series or above.

[Idiom]

There's almost ten different GPGPU language options. CUDA was just one off the top of my head while I had a couple spare minutes to post. Mainly, CUDA is for physics calculations, which I thought would be particularly useful. It's what happened to PhysX. ATI does have something new that's physics calculations on the video card, but I can't find much. Looks like they dropped the R&D on that. It was planned for the X1000 series is all I can find right now. BrookGPU appears to be a universal, using OpenGL which I'm certain I heard Toady is already using.

It's very similar to stream processing. Computing many different instances in a similar fashion. Sounded useful for DF, and I can understand why they would use it for VR physics.

So would it be possible to port a version that runs solely on the video card? Not going to happen for practical reasons (until Intel finally wins and processors start at $1000. Considering DF won't even be finished for another decade or so), but is it possible for curiosity's sake?

Draco, if you don't even read the first post all the way through, or don't even care what you are bashing, just don't reply.

[Draco18s]

I know exactly what I'm "bashing."  I just don't give a damn about all the acronyms.