Jump to content

  • Log In with Google      Sign In   
  • Create Account

C# - Need help with choosing a way to solve a ridiculously easy 2D graphics task


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
10 replies to this topic

#1 Istrebitel   Members   -  Reputation: 105

Like
0Likes
Like

Posted 15 November 2013 - 04:55 AM

Greetings.

 

I have a pretty simple task before me Having coded some games in the past (using DelphiX back then, nowadays I'm C# .Net programmer) I thought it was going to be easy. I'm utterly stunned by how impossible it seems to me right now, because nothing works... at all! I will break it down to an absolute basic level (the app itself is more complex, but this is the key part I cannot achieve):

 

I need to write a full screen application (C# .Net) that pereodically shows a string of text on screen. This string of text must be shown for exactly 1 frame (computers this would be used on will all have 60Hz LCD monitors, but we can also assume 2 frames for 120Hz monitors). No more, no less. One frame, or 16,(6) ms. So, at t=0, screen is white, at t=16,6ms, "TEXT" appears, at t=33,3ms, screen is white again.

 

This program must run on Windows XP, Windows Vista and Windows 7.

 

However, I've spent about week studying different technologies and it seems that this task is nigh unsolvable! That's why I'm asking for your help.

 

Here's what I tried:

 

1) Windows Forms.

 

I tried just making a full screen borderless form, and showing/hiding a label. On a 60Hz monitor (refresh each 16,(6)ms), even when showing a label for 20ms, I had times when it wouldn't show at all. I used precise timer (Stopwatch class), I'd show the label, then run a timer for 20ms, and hide it when it ticks. Label wouldn't even appear on the screen. Obviously, other times, it would show for 2 frames total, because it can appear just before first refresh, say, 1ms before it, then ~16ms passes (total 17ms still less than 20ms) and second refresh still displays it, and then it gets hidden at 20ms. So it was no good.

 

I assumed this is because I have no direct control when stuff is getting painted - in terms of getting my label to appear on the screen, there was just too much links in the chain - first I tell it to .Show(), then OS will draw it onto form, then DWI will draw form onto screen, and then monitor will get new frame from the screen.

 

This is where I thought that the way to solve this is to use double buffering + vsync, because that would guarantee exposure, because buffers will only be swapped after a frame went to the monitor. I decided to look at available options among direct graphics output.

 

2) SDL

 

I tried SDL.Net, and I had no success with it. From what I gathered, you cannot enforce Vsync, but if it is not disabled in your graphics card drivers, it should be automatically used if you create a hardware double-buffered surface with SDL. Turns out, NOPE. As described here http://gamedev.stackexchange.com/questions/65841/c-sdl-net-how-to-enable-vsync I made a double-buffered hardware surface and am still getting 90 fps, meaning vsync isn't working, even though it's enabled - I checked in my drivers (using official nVidia ones, my GPU is GeForce GTX 460). Noone seems to be able to help me on that matter. Therefore, had to give up on it.

 

3) DirectX

 

I tried getting DirectX 9.0 SDK and it refused to install on my PC, saying I have to have an old OS (Windows 7 was not listed in the list of supported OS'es).

 

I installed SharpDX, I tried this tutorial http://katyscode.wordpress.com/2013/08/24/c-directx-api-face-off-slimdx-vs-sharpdx-which-should-you-choose/ but i crashes (I assume because it uses some 3D-2D operation only supported on Windows 8). I tried some other examples and they worked, but all of them are using DX11 or DX10 meaning my application won't run on XP! And I need that, because probably some of the users of my program will have computers that still have XP installed on them. Can't be sure, but for such a simple task, upgrade of a whole OS shouldn't be required - we're talking about outputting a string of text on the screen here! A simple string of simple text on a blank screen, dammit!

 

I tried to look around and everywhere I go examples are using DX11.

 

4) OpenGL

 

I tried to create OpenGL surface in SDL (supposedly you can set GL_swapcontrol and get vsync that way) but I just dont understand how do I output anything on that surface? There's no tutorials I found that explains it! I can swap buffers but that's all, I don't get how do I output any text there.

 

I tried SharpGL but there's not a single tutorial I found about a full screen application - it only uses windowed mode, and googling didn't provide anything useful. So once again, no idea what to do here.

 

5) XNA

 

I had past experience making small games in XNA so I know about the technology. I'd like to avoid XNA if possible because it requires installing a separate package on the user machine, and I'd like to avoid that if possible (I mean, for such a simple program, I'd like to keep it as simple as copying over an .exe and some .dll's). But I guess, if all else fails, and XNA works, I might have to go with it.

 

------------------------------------------------------------------------------------

 

Well, this is it. I feel like a total moron, not being able to do the simpliest of tasks. I kept pesteting google, stack overflow and other sources for answers to no avail. I'd really appreciate if someone could just tell me how to do this very simple thing, or at least tell me where I am wrong and hint me how can I do it.

 

Thanks in advance!


Edited by Istrebitel, 15 November 2013 - 04:55 AM.


Sponsor:

#2 Nik02   Crossbones+   -  Reputation: 2929

Like
0Likes
Like

Posted 15 November 2013 - 01:03 PM

OpenGL or Direct3D (with C or C++) are your best bets.

 

DirectX SDK June 2010 (the last separate SDK before its merge with the platform SDK) can definitely be installed on Windows 7, although the debug runtime has some troubles and you can't use PIX if you have up-to-date machine. The said SDK has headers and libs for D3D9, D3D10 and D3D11.

 

Raw OpenGL (without SDL) is very simple to set up, there are a lot of basic tutorials.

 

...

 

That said, modern desktop operating systems cannot guarantee that any operation finishes in a fixed time. Memory and processing resources are virtualized as far as the user mode (in which you run ordinary applications) can see, and other processes, including those running with the OS itself, can introduce random delays in your application.

 

This means that if it is absolutely critical that the graphic only shows for the duration of one physical frame (1/60 seconds), you are considering the wrong computing platform.

 

Can you tell us the reasoning behind that requirement?


Edited by Nik02, 15 November 2013 - 01:09 PM.

Niko Suni


#3 Stragen   Members   -  Reputation: 302

Like
0Likes
Like

Posted 17 November 2013 - 01:52 AM

For OGL i know the NEHE tutorials cover the display of a piece of text, though these are more C++ related... but the translation shouldn't be too difficult.

 

http://nehe.gamedev.net/tutorial/outline_fonts/15004/

 

I agree with Nik that you cant guarantee the delivery every 1/60th of a second, but i think you could get pretty darn close.



#4 Istrebitel   Members   -  Reputation: 105

Like
0Likes
Like

Posted 18 November 2013 - 03:44 AM

OpenGL or Direct3D (with C or C++) are your best bets.

 

DirectX SDK June 2010 (the last separate SDK before its merge with the platform SDK) can definitely be installed on Windows 7, although the debug runtime has some troubles and you can't use PIX if you have up-to-date machine. The said SDK has headers and libs for D3D9, D3D10 and D3D11.

 

Raw OpenGL (without SDL) is very simple to set up, there are a lot of basic tutorials.

 

...

 

That said, modern desktop operating systems cannot guarantee that any operation finishes in a fixed time. Memory and processing resources are virtualized as far as the user mode (in which you run ordinary applications) can see, and other processes, including those running with the OS itself, can introduce random delays in your application.

 

This means that if it is absolutely critical that the graphic only shows for the duration of one physical frame (1/60 seconds), you are considering the wrong computing platform.

 

Can you tell us the reasoning behind that requirement?

It is an application that has to do with testing humans for reactivity intervals to different stimuli. I can't go into much detail, but the reasoning is that in order for the results to be consistent, stimuli are to be presented for the exact time interval. That's not my decision or conclusion, I am just a programmer tasked with making it happen, and based on my experience and knowledge, I assumed that it is possible to make it happen - I mean, since we can output modern games like Counter-Strike GO for hours uninterrupted and without FPS drops, I thought that organising a simple text string would be possilble...

 

Are you asbolutely sure it's not possible to guarantee that?



#5 meeshoo   Members   -  Reputation: 508

Like
1Likes
Like

Posted 18 November 2013 - 06:55 AM

There is also WPF (Windows Presentation Foundation) which is rendered through hardware acceleration and you can even feed shaders to controls and such, measure FPS and so on.

 

But whatever you choose, i don't understand why don't you just start a time counter, display text, measure 1/60 seconds and hide text?


Edited by meeshoo, 18 November 2013 - 06:56 AM.


#6 Istrebitel   Members   -  Reputation: 105

Like
0Likes
Like

Posted 18 November 2013 - 07:38 AM

Because no text is shown. This is the biggest problem - sometimes, no text gets shown ever. In Windows Forms (fullscreen form), I was showing text for as long as 20 ms (that's 1/50 second) and it would still not appear on the screen.



#7 Nik02   Crossbones+   -  Reputation: 2929

Like
0Likes
Like

Posted 18 November 2013 - 01:55 PM

 

 Counter-Strike GO for hours uninterrupted and without FPS drops

 

That is an illusion. Most of the time, the framerate can be consistent, but any process on the system can create hiccups in which case you potentially miss some frames here and there.

 

Medical-type testing (and other medical applications) often requires dedicated hardware that can guarantee specific behavior without interruption. If you absolutely need to guarantee that a: the frame actually shows up and b: that it only stays visible for 1/60th of a second, you need such dedicated hardware.

 

Most monitors cannot guarantee every frame delivery either, though in practice they do usually succeed in drawing each frame.

 

...

 

Modern desktop operating systems accommodate user-mode software by implementing a message queue (window messages or equivalent) that separates the OS core processes from the other processes that use the services of the OS (loose coupling). This way, the kernel doesn't have to call the user-mode processes directly, and therefore a program hang or crash does not bring down the kernel. A message queue makes it possible to make cross-process and cross-thread messages without having to synchronize the participating threads (which is a cornerstone of any modern OS).

 

Processes compete with each other over the "attention" of the OS, and WPF even has its own user-mode event queue system (the Dispatcher) on the top of the OS scheduler.

 

Because the queue is observed by the processes when the processor gives them the time to do that, the timing of user-mode programs is effectively non-deterministic. You also need to remember that the other processes share all the other hardware of your system as well, including cache, memory, buses, mass memory and the video card, and they can interrupt your frames at random intervals and for random amounts of time (as far as the user process knows).

 

Dedicated hardware would avoid this situation by only running your drawing code at fixed time intervals, and doing nothing else. You'd also need to obtain a monitor that is guaranteed to display all frames as they come in, without latency or inter-frame leak.

 

Your description of the problem hints that the system needs to have medical precision.


Edited by Nik02, 18 November 2013 - 02:18 PM.

Niko Suni


#8 Istrebitel   Members   -  Reputation: 105

Like
0Likes
Like

Posted 19 November 2013 - 03:52 AM

Thanks for a detailed explanation. Yes I see the problem now.

 

However, I understand that I cannot guarantee 100% reliability - and I don't really need to (like, if monitor will skip a frame, like, once in a million, its fine - the person undergoing the test can also choose to blink at the exact moment the image pops on the screen for the same result). Thing is - it seems to be possible to achieve way more than I currently can.

 

I found that OpenTK seems to be easy to use to do what I need, and my application actually worked without flaws as far as I could test it (like, ~60 times in a row without a single hiccup, I know 60 is not a big number, but it's hard to force yourself to sit and stare at a screen blinking every second for over a minute...). However, that was on the old version of the lib (1.0). That version does not support re-opening window (so once you close the OpenGL window, you can no longer use OpenGL until you restart the program). Since it's an old one. But then... newer version (1.1) introduces hiccups! For no apparent reason. Here's where I discuss it with the developer:

http://www.opentk.com/node/3436#comment-14685

It seems he also has no idea why it would be the case (I hope he finds out, eventually).

 

I understand that I'm kinda trying to do with the PC something that it's not designed for. And that means, I should not expect for it to work out correctly. But then again, that's how it usually happens with technology, isn't it? If you would ask people about playing games or taking photos using mobile phones in 1990's, they'd tell you "phones aren't designed for that, buy a gameboy or a camera", but then, someone decided to try it out, and even though it was clearly not designed for playing games or taking photos, mobile phones started to be used that way, and eventually, we have a huge mobile games market and most of the casual photos nowadays taken by your phone camera.

 

Also, if I can achieve what I need (well, I cannot scientifically confirm it because I haven't tested it on large enough number of expositions, but still) with an old version of the OpenTK library, but not with modern version, there must be some reason for it, and that reason most likely is NOT the message queue / monitor eating frames / anything else.

 

So yeah, I'm not giving up on this problem, because I feel I can achieve a reasonable compromise of a solution.



#9 Nik02   Crossbones+   -  Reputation: 2929

Like
0Likes
Like

Posted 19 November 2013 - 09:55 AM

If you can afford compromises, I agree with you. It's often the case with medical systems, though, that compromises are not allowed.

Niko Suni


#10 Istrebitel   Members   -  Reputation: 105

Like
0Likes
Like

Posted 20 November 2013 - 12:59 AM

Well, this time it's a prototype of sorts, and not a critical system (that has human lives depending on it), so I guess it's okay.

 

Anyways, on topic, I think the problem is solved - in that I found OpenGL is the simpliest way to do it. Right now, it remains to solve the issues with OpenTK (the library for C# I chose to leverage OpenGL).



#11 Nik02   Crossbones+   -  Reputation: 2929

Like
0Likes
Like

Posted 20 November 2013 - 02:12 PM

I am inclined to think that testing human subjects' response timing to stimuli is a type of health assessment, and where I have worked, this type of software is also considered very critical - even though lives aren't directly at risk.

 

False or biased readings from medical equipment, no matter how unlikely they are, can affect the subject's life in a dramatic way. Even though the results may seem insignificant to someone, somebody else might consider them very significant to the point of regarding them as authority (even if they were actually faulty). Consider if employment or insurability of an individual is determined by the test, and he/she fails because the equipment didn't work...

 

The above points may seem far-fetched but they are very real concerns in serious medical sw development.

 

Anyway, native OpenGL or Direct3D are the best technologies for the stuff you're doing, if the PC platform itself is sufficient.


Niko Suni





Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS