[QUOTE=Ziks;49584372][vid]http://files.facepunch.com/ziks/2016/January/22/2016-01-22-1620-28.mp4[/vid]
Redesigned the world menu, although it's still a little placeholdery. You can now take a preview screenshot using it too.[/QUOTE]
That's a pretty cool way to take a screenshot.
[QUOTE=DrDevil;49584396]That's a pretty cool way to take a screenshot.[/QUOTE]
Ziks' whole project is extremely cool. I'm going to go as far as saying that his project is the best VR related project that I've seen to date.
[QUOTE=Ziks;49584372][vid]http://files.facepunch.com/ziks/2016/January/22/2016-01-22-1620-28.mp4[/vid]
Redesigned the world menu, although it's still a little placeholdery. You can now take a preview screenshot using it too.[/QUOTE]
I bet that screenshot taking method is patentable (and evil).
I made an extremely simple falling sand shader, I'm learning you need to use a lot of ifs and gpus don't like ifs.
[IMG]http://i.imgur.com/981p9pm.gif[/IMG]
Should I github this shi?
[QUOTE=polkm;49585208]I made an extremely simple falling sand shader, I'm learning you need to use a lot of ifs and gpus don't like ifs.
[url]http://i.imgur.com/981p9pm.gif[/url]
Should I github this shi?[/QUOTE]
Put it on shadertoy!
[QUOTE=polkm;49585208]I made an extremely simple falling sand shader, I'm learning you need to use a lot of ifs and gpus don't like ifs.
[IMG]http://i.imgur.com/981p9pm.gif[/IMG]
Should I github this shi?[/QUOTE]
Yeah, shadertoy it, I wanna play with it a bit.
^Could you briefly describe how that works on a high level please?
OK I ported it to shadertoy, find iritating cause I lost progress like three times because I thought it saved when it didn't, but yeah. It's slower on shadertoy because I can't think of a good way to force multiple pases every frame like in love.
[url]https://www.shadertoy.com/view/XsV3zW[/url]
[QUOTE=DoctorSalt;49585511]^Could you briefly describe how that works on a high level please?[/QUOTE]
Shaders can store data in a texture to be used across multiple frames. In this case you just have to store whether a pixel contains a grain of sand or not. Then, as you iterate over all the pixels, you just check if the pixel below the current pixel is free. If so, move down. If not, check if the pixel left or right is free, and move there if it is so.
[QUOTE=DoctorSalt;49585511]^Could you briefly describe how that works on a high level please?[/QUOTE]
At the top level the shader is recursive in that the last frame is fed into the next. All empty pixels check above them to see if they need to change into a sand pixel. Each sand pixel checks below them if than can turn into an empty pixel. Making what looks like movement. Also when falling diagonally left or right I check to make sure there isn't another sand falling in the same place at that frame.
i don't know what i did
[media]https://www.youtube.com/watch?v=qV2ufHHfKBw[/media]
How do people browse the shadertoy website, it ranges from 20+ second freezes to display driver crashes when you press the browse button in all major browsers.
so im playing around now with MATLAB and by the end of this quarter sometime I'm going to try to start simulating fluids equations for MHD fusion applications. The equation is a nonlinear, three dimensional, time dependent equation and many models take hundreds of wall-clock hours to simulate and thousands of CPU hours. My goal isn't to break any ground, but rather go through the MIT MHD fusion course and learn to model the various types/geometries of tokomaks, soooooo I'm thinking of getting something like a parallela for offloading computing on so I can possibly do some fun simulations in reasonable time without completely destroying accuracy.
Would that be of any use to me, or should I look for something else? Where can I find more data about this? There's a ton of resources on parallel processing, but not on the computational requirements of tokomak simulations (weirdly, how to program this stuff is super easy to find just not how resource intensive it is). I'm trying to do this so that I can work in a fusion lab this summer, too, so regardless of how accurate this is I really just want to apply my scientific computing class and learn more about parallel computing in research work. That and I'm really having a blast with scientific computing and data analysis, so I'd probably try and use this for fun astronomy data analysis as well.
Implemented saving and loading. I am saving each node and all connections associated with the node. Even though just saving either the output-links or the input-links for each node would be sufficient to reconstruct the whole graph, I save both. That way I have some redundance in case some node fails to load.
[media]https://www.youtube.com/watch?v=1cpAzRJxGVs[/media]
[QUOTE=paindoc;49586086]so im playing around now with MATLAB and by the end of this quarter sometime I'm going to try to start simulating fluids equations for MHD fusion applications. The equation is a nonlinear, three dimensional, time dependent equation and many models take hundreds of wall-clock hours to simulate and thousands of CPU hours. My goal isn't to break any ground, but rather go through the MIT MHD fusion course and learn to model the various types/geometries of tokomaks, soooooo I'm thinking of getting something like a parallela for offloading computing on so I can possibly do some fun simulations in reasonable time without completely destroying accuracy.
Would that be of any use to me, or should I look for something else? Where can I find more data about this? There's a ton of resources on parallel processing, but not on the computational requirements of tokomak simulations (weirdly, how to program this stuff is super easy to find just not how resource intensive it is). I'm trying to do this so that I can work in a fusion lab this summer, too, so regardless of how accurate this is I really just want to apply my scientific computing class and learn more about parallel computing in research work. That and I'm really having a blast with scientific computing and data analysis, so I'd probably try and use this for fun astronomy data analysis as well.[/QUOTE]
GPGPU is your best bet, beyond that point its HPC. Parallela is interesting because it doesn't have some of the restrictions a GPU has and has a decent performance/wat, but its also much slower then a modern GPU.
You could find a way to cluster your workload and run it on EC2/Compute Engine, but generally that's hard with fluid equations, especially if you don't want to cut corners and lose out on accuracy.
[QUOTE=paindoc;49586086]so im playing around now with MATLAB and by the end of this quarter sometime I'm going to try to start simulating fluids equations for MHD fusion applications. The equation is a nonlinear, three dimensional, time dependent equation and many models take hundreds of wall-clock hours to simulate and thousands of CPU hours. My goal isn't to break any ground, but rather go through the MIT MHD fusion course and learn to model the various types/geometries of tokomaks, soooooo I'm thinking of getting something like a parallela for offloading computing on so I can possibly do some fun simulations in reasonable time without completely destroying accuracy.
Would that be of any use to me, or should I look for something else? Where can I find more data about this? There's a ton of resources on parallel processing, but not on the computational requirements of tokomak simulations (weirdly, how to program this stuff is super easy to find just not how resource intensive it is). I'm trying to do this so that I can work in a fusion lab this summer, too, so regardless of how accurate this is I really just want to apply my scientific computing class and learn more about parallel computing in research work. That and I'm really having a blast with scientific computing and data analysis, so I'd probably try and use this for fun astronomy data analysis as well.[/QUOTE]
You can do octree stuff do speed up greatly fluid simulations.
[QUOTE=Handsome Matt;49586594][t]http://i.imgur.com/pZvrq2q.jpg[/t][t]http://i.imgur.com/xwT4epl.jpg[/t]
started learning ue4 yesterday ish, spent most the time writing fbx exporters and trying to get a heightmap to play nice - scales, materials and some sub meshes are fucked but I'm glad I've managed to get something rendered to an extent[/QUOTE]
fuck YES
Well, it only took me two hours :v:
[t]http://i.imgur.com/tuyMPRn.png[/t]
[QUOTE=Handsome Matt;49586594][t]http://i.imgur.com/pZvrq2q.jpg[/t][t]http://i.imgur.com/xwT4epl.jpg[/t]
started learning ue4 yesterday ish, spent most the time writing fbx exporters and trying to get a heightmap to play nice - scales, materials and some sub meshes are fucked but I'm glad I've managed to get something rendered to an extent[/QUOTE]
Good job dude!
Our first multiplayer test went rather well, considering...
[img]http://i.imgur.com/gbmmP1Z.png[/img]
[img]http://i.imgur.com/Z5wlcHF.png[/img]
I just spent the best part of the last 5 hours refactoring over 150 unit tests because they werent being run properly and as such were reporting as false positives... kill me now :nope:
[QUOTE=Handsome Matt;49586594][t]http://i.imgur.com/pZvrq2q.jpg[/t][t]http://i.imgur.com/xwT4epl.jpg[/t]
started learning ue4 yesterday ish, spent most the time writing fbx exporters and trying to get a heightmap to play nice - scales, materials and some sub meshes are fucked but I'm glad I've managed to get something rendered to an extent[/QUOTE]
Is this stuff called "Handsome Matt" flesh of god or something.
[sp] Really nice for 5hs of works, this is what I am saying[/sp]
[video=youtube;QM1iUe6IofM]http://www.youtube.com/watch?v=QM1iUe6IofM[/video]
This guy makes a lot of good points. Lately I've been gravitating more towards procedural code as much as possible and it's nice to see someone explain a lot of the things I've been thinking about as this guy did. The only thing that for my use case specifically OOP still makes a lot of sense is when writing gameplay objects. The relationship here between code and simulation is a 1:1 mapping so it makes no sense to try anything else. But anything OTHER than gameplay objects it doesn't make a lot of sense to wrap in classes.
The more important thing though is that I've been gravitating towards minimizing mental effort spent while coding, and this usually results in things that go against the "best" practices. For instance, I've been copy pasting a lot of my own code around lately instead of properly making it reusable, simply because it's the thing that takes less mental effort in the long run. In the general case doing this is bad, but when you're working alone it totally isn't because the code base is just so small. Even something like 100k lines of code you can still have the general way things are tied together in short term memory.
[QUOTE=adnzzzzZ;49587576]The more important thing though is that I've been gravitating towards minimizing mental effort spent while coding, and this usually results in things that go against the "best" practices. For instance, I've been copy pasting a lot of my own code around lately instead of properly making it reusable, simply because it's the thing that takes less mental effort in the long run. In the general case doing this is bad, but when you're working alone it totally isn't because the code base is just so small. Even something like 100k lines of code you can still have the general way things are tied together in short term memory.[/QUOTE]
What happens if you need to ask someone to take a look at your code? What happens if you decide you want to open-source it and have it be usable? What happens if you ever have to work in a team and you're fighting against your habit of just writing lazy code? "I'm the only one working on it" isn't a new justification for writing sloppy code, in fact it's probably the only one. When you copy and paste code you're basically assuming that however you did it before is optimal and ready to go, what happens when you realize there's a bug in that code that you've copied into 15 files? I mean yeah, it sounds trivial to go back and copy/paste 15 times with the new code, but the point is you're just creating more work for yourself later on when things inevitably go wrong, or you find a better way of doing something.
[editline]23rd January 2016[/editline]
Not trying to be confrontational or anything, just food for thought regarding the whole "write whatever is easiest" thing.
[QUOTE=srobins;49588000]What happens if you need to ask someone to take a look at your code? What happens if you decide you want to open-source it and have it be usable? What happens if you ever have to work in a team and you're fighting against your habit of just writing lazy code? "I'm the only one working on it" isn't a new justification for writing sloppy code, in fact it's probably the only one. When you copy and paste code you're basically assuming that however you did it before is optimal and ready to go, what happens when you realize there's a bug in that code that you've copied into 15 files? I mean yeah, it sounds trivial to go back and copy/paste 15 times with the new code, but the point is you're just creating more work for yourself later on when things inevitably go wrong, or you find a better way of doing something.
[/QUOTE]
Caring about all those things costs more than just doing things the easy way first. Things don't go wrong as often as I'm having to add new things to my code base, so it makes mental economic sense to do the easy thing as often as possible. Now, it's important that "doing the easy thing" has some structure to it. This doesn't mean I'm going to use globals all the time or abandon commenting code forever or whatever. It just means that you have to look at situations on a case by case basis and not be afraid to do the easy and "wrong" thing if it turns out that that thing is the one that makes most sense. For instance, I'm only copy pasting things in a few processes (like adding a new attack) where it makes sense to do that. In other processes it doesn't make sense to do that so I don't.
[QUOTE=adnzzzzZ;49588120]Caring about all those things costs more than just doing things the easy way first.[/QUOTE]
But it pays off in the long run when, as srobins said, you want to change the code that you copy-pasted everwhere.
[QUOTE=adnzzzzZ;49588120]Things don't go wrong as often as I'm having to add new things to my code base,[/QUOTE]
You're often adding new things to your code base... and therefor things don't go wrong? :what:
[QUOTE=adnzzzzZ;49588120]so it makes mental economic sense to do the easy thing as often as possible.[/QUOTE]
...until you come to a point where your fast and loose method falls apart and leaves you with a mess that causes you more of a headache than you would have had if you just took the time and effort to do it in the first place.
[QUOTE=adnzzzzZ;49588120]Now, it's important that "doing the easy thing" has some structure to it. This doesn't mean I'm going to use globals all the time or abandon commenting code forever or whatever. It just means that you have to look at situations on a case by case basis and not be afraid to do the easy and "wrong" thing if it turns out that that thing is the one that makes most sense. For instance, I'm only copy pasting things in a few processes (like adding a new attack) where it makes sense to do that. In other processes it doesn't make sense to do that so I don't.[/QUOTE]
What is your definition of "make sense"? Does saving a bit of time and effort now in exchange for expending more time and effort in the future "make sense" to you?
[QUOTE=Cold;49586411]GPGPU is your best bet, beyond that point its HPC. Parallela is interesting because it doesn't have some of the restrictions a GPU has and has a decent performance/wat, but its also much slower then a modern GPU.
You could find a way to cluster your workload and run it on EC2/Compute Engine, but generally that's hard with fluid equations, especially if you don't want to cut corners and lose out on accuracy.[/QUOTE]
[QUOTE=Fourier;49586562]You can do octree stuff do speed up greatly fluid simulations.[/QUOTE]
Right o, I'll give gpu platforms + octree optimization more of a look. Currently GYRO is the most common code base for testing reactors, and has shown accuracy to within a few % (!). It actually helped find one of the more common instabilities. The code seems fairly open too, so I'm hoping my academic reasoning is enough to justify getting a tarball of the source.
I will say that as per the usual high end EMF stuff is fucking sorcery. The term "Six dimensional Phase space" shows up multiple times. The function used a ton is apparently recursive and expands to n dimensions unless artificially constrained (introducing inaccuracy). Also, I can't figure out how my Uni models a spheromak.
Insane Clown Posse asked the eternal question- magnets, how the fuck do they work?
i have decided
i will give birth to the shittiest action rpg ever, with unity
no i wont use unity primitives this time
its time
time for
[t]http://i.imgur.com/IPIQ2FP.png[/t]
mspaint.
After 3 days after of after-work mucking about I finally got my DirectX 12 engine to work (read: display the clear colour) in C#! I've never written a C++/CLI wrapper for a native C++ DLL before.. Christ that was quite an experience.
In theory I just need to "port" over the renderable classes and I could make a little level editor, assuming I also figure out how to squeeze the renderer into a sub-area of the main window.
[QUOTE=adnzzzzZ;49588120]Caring about all those things costs more than just doing things the easy way first. Things don't go wrong as often as I'm having to add new things to my code base, so it makes mental economic sense to do the easy thing as often as possible. Now, it's important that "doing the easy thing" has some structure to it. This doesn't mean I'm going to use globals all the time or abandon commenting code forever or whatever. It just means that you have to look at situations on a case by case basis and not be afraid to do the easy and "wrong" thing if it turns out that that thing is the one that makes most sense. For instance, I'm only copy pasting things in a few processes (like adding a new attack) where it makes sense to do that. In other processes it doesn't make sense to do that so I don't.[/QUOTE]
Where do you want to go with your programming? Are you studying/looking to be employed etc?
Sorry, you need to Log In to post a reply to this thread.
