Print

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

[Mego]

Am I doing this right?

http://pastebin.com/87ZUe64p

I'm not sure why defined next to #if is blue, but defined next to #define isn't... @_@

1. Get rid of the #defines that set stuff to 0
2. Move the #defines and #undefs to the top, before you start including the headers.
3. The directives should look like this:

Code: [Select]

#define CL_VERSION_1_0                              1
#undef CL_VERSION_1_1
#undef CL_VERSION_1_2

defined() is a macro function that returns 1 if the macro is defined, and 0 if it's not. You can't assign values to it; instead, assign values directly to the macros. Since the C++ binding checks if the macro is defined, rather than its value, you have to undef the unwanted macros and never redefine them.

[Skyrunner]

Exe
misc dlls

Delete the exe inside the zip and use the deviceprofile5.exe. I don't think it's helping >_>

[MagmaMcFry]

So I was away over the weekend, and it looks like people missed me. Skyrunner, what exactly are the problems you currently have?

[olemars]

I don't think it's helping >_>

Certainly doesn't help that you're getting contradictory information   

2. Move the #defines and #undefs to the top, before you start including the headers.

Nooooo. When you're trying to redefine, or in this case undefine, a #define from an included file, the redefinition has to be after the #include or it won't have any effect.

3. The directives should look like this:

Code: [Select]

#define CL_VERSION_1_0                              1
#undef CL_VERSION_1_1
#undef CL_VERSION_1_2

defined() is a macro function that returns 1 if the macro is defined, and 0 if it's not. You can't assign values to it; instead, assign values directly to the macros. Since the C++ binding checks if the macro is defined, rather than its value, you have to undef the unwanted macros and never redefine them.

This is correct.

[Max White]

So I was away over the weekend, and it looks like people missed me. Skyrunner, what exactly are the problems you currently have?

Actually if you want to write up a tutorial on using OpenGL in lwjgl I'm sure I'm not the only one who will ever use it.
I like most of your class structure, by the way.

[MagmaMcFry]

Actually if you want to write up a tutorial on using OpenGL in lwjgl I'm sure I'm not the only one who will ever use it.

About half done, will finish tomorrow.

[Mego]

2. Move the #defines and #undefs to the top, before you start including the headers.

Nooooo. When you're trying to redefine, or in this case undefine, a #define from an included file, the redefinition has to be after the #include or it won't have any effect.
[/quote]

Isn't he trying to define the version so that, when he includes the header, the preprocessor will read it and generate the proper code? Or am I misunderstanding the problem?

Like so:

Code: (opencl.h) [Select]

#if defined(CL_VERSION_1_0)
// code he wants
#endif
#if defined(CL_VERSION_1_1)
// code he doesn't want
#endif
// and so on


[olemars]

No, the defines are already in the header and in turn generates the improper code in the c++ bindings, which are including that header. Thus the need to undef some of those defines to target the older version. Which is slightly hacky, but I can't see any better way.

The real problem is that the Khronos sample implementation, and the AMD SDK with it, doesn't have any good mechanism for targeting a specific API version. So if you have the newest SDK (which includes OpenCL 1.2) but want to target OpenCL 1.0 or 1.1 to support a wider array of hardware you have to do hacks like this.

[Nadaka]

No I remember why I abandoned my last couple coding projects outside of work... I was trying to do comprehensive unit tests.. and its boring as hell.

[MagmaMcFry]

OpenGL 1.1 with LWJGL! Go learn something.

EDIT: By the way, that's an Eclipse project too.

[forsaken1111]

Just posting to watch and say I've recently started learning Java myself. Doing well so far, I can make things add!

[Skyrunner]

OpenGL 1.1 with LWJGL! Go learn something.

EDIT: By the way, that's an Eclipse project too.

I was totally like "OpenCL 1.1 with LWJGL?! AWESOME " then I saw that it's Java then I saw that it's GL. :<



edit: Yessss, it works. I am now ready to proceed to create an unoptimized version of the simulation. I need to decide whether to keep the object oriented original or just unpack it all into arrays.

[alway]

Whelp you're right.


Edit: Screw me, of course you can't find the maximum of an array in log(n) time, since if no assumptions can be made about the array, every element needs to be checked.

Fun fact addendum to this particular discussion:
While you cannot find an element in an unsorted array in less than O(n) time with classical computing, you can find it in O(n^(1/2)) time using quantum computing. http://en.wikipedia.org/wiki/Grover's_algorithm

Though practical quantum computers are still likely a decade or two away. But still; it's something we will definitely need to learn how to use at some point in the not-distant future.

[Siquo]

Well, I doubt quantum computing will make the current computer-systems obsolete for a very long while, but a plug-in quantum card (just like you have a video card now) could become a thing, perhaps.

Are there any quantum-computer-emulators out there, yet? It'll be magnitudes slower than the actual thing, but you'd be able to test algorithms on small samples.
Fake edit: of course there are, google is my friend.

[Skyrunner]

What are the odds? I was thinking of quantum computing and that no matter what people may think, it's going to be as hard or even harder to use in virtual reality or gaming aspects as GPGPU is now Quantum computers could revolutionize the simulation industry or such, though, with speedup ...