cuda header files .cuh private host functions - cuda

In cuda we can create header files with .cuh extension and we can call the functions from anywhere like,
__device__ void doSomething()
{
....................
}
void doSomthingOnHost()
{
....................
}
these two functions are public. How can i make the host function to private?

I find that what works best for me is to
Make .CU files with my CUDA kernels, their public C/C++ wrappers and any private/encapsulated C/C++ functions I need to make the device code work.
Make .H files which provide access to the C/C++ wrappers inside my .CU files, #including them in the .CU files and any .C/.CPP files I need to call the device code from
Make .C/.CPP files which handle the high-level application logic and which invoke device code through interfaces supplied through the header files described in step 2.
To make host functions private in this scheme, just don't put prototypes for them in the header... a pretty neat scheme if you ask me.

Strictly speaking, there's no way to make a free function private in C++ -- any client which can see a function's signature can call it.
Instead, you could make doSomethingOnHost a private, static member function of some class:
class my_class
{
private:
static void doSomethingOnHost(); // only my_class or friends of my_class may use this function
};

Related

How to pass configuration parameters to SimpleInjector packaging?

So SimpleInjector now has a packaging nuget that you can use to isolate different aspects of root composition.
Say I have a configurable composition root in a library that is reused by multiple projects in an application. For example, in an Azure solution I might have a Web role and a Worker role which share a large set of the same dependencies for the most part, but with slightly different configuration options depending on the consumer. When I compose the root, I can pass in a plain old RootCompositionSettings object with properties that tell SimpleInjector how to register dependencies.
However, I am not sure how (or if) I can pass these settings to an IPackage instance. Is it possible to pass custom settings to a SimpleInjector package, and if so, how?
I see that the standard practices for registering packages is to invoke either
container.RegisterPackages(); // scans all loaded assemblies for IPackage
// or
container.RegisterPackages(IEnumerable<Assembly>) // specific assemblies only
...so how can we pass parameters into the packaging instance(s)? Is there some way to do it via the container?
The trick here is to pass the information on with the container to the package. You can do this by using the container's Items dictionary, that is much like ASP.NET's HttpContext.Items collection. This can be done as follows:
using SimpleInjector.Advanced;
container.SetItem(typeof(RootCompositionSettings), settings);
container.RegisterPackages();
Now inside your packages, you can do the following:
var settings =
(RootCompositionSettings)container.GetItem(typeof(RootCompositionSettings));
Please note that:
SetItem and GetItem are extension methods that are located in the SimpleInjector.Advanced namespace. Those methods allow you to access the (internal) Items dictionary.
You can pass in any key you like. Passing in typeof(RootCompositionSettings) is just convenient in this case, but not required.
If you need to call the settings in more places, it might be useful to create a more specific extension method that allows you to access the setting instance, but that's up to you.
Another option is to not use the IPackage interface and the SimpleInjector.Packaging library at all. In most cases it doesn't really add anything and you could simply define a public static method in the assembly that does the same as a package does. For instance:
public static class BusinessLayerBootstrapper
{
public static void Bootstrap(Container container, ScopedLifestyle scopedLifestyle,
RootCompositionSettings settings)
{
// Here the same logic as what you would write in your package.
}
}
Most applications are not that dynamic that you need to load assemblies dynamically and the startup project usually has a hard reference to all the other assemblies. In that case it is perfectly sane to simply call a static method.
And even if you have the requirement of dynamically loading assemblies and allowing them to register their stuff in the container, it's quite trivial to build your own IPackage abstraction instead:\
// Your own IPackage interface
public interface IPackage
{
void RegisterServices(Container container, RootCompositionSettings settings);
}
// Your own extension method
public static void RegisterPackages(this Container container,
RootCompositionSettings settings)
{
var packages =
from assembly in AppDomain.CurrentDomain.GetAssemblies()
from type in assembly.GetExportedTypes()
where typeof(IPackage).IsAssignableFrom(type)
where !type.IsAbstract
select (IPackage)Activator.CreateInstance(type);
packages.ToList().ForEach(p => p.RegisterServices(container, settings));
}
In fact, except for some extra validations and filtering out dynamic assemblies, the SimpleInjector.Packaging project is not much more than this.

How do I define 'out' parameters a in C++CX Windows Runtime Component?

By the looks of Google it seems like this might not be possible, but:
How do I define an 'out' parameter in a C++/CX 'ref class'?
If your answer is that this isn't possible, please provide a reference.
Any parameter which is of type T* (where T is a ABI-legal type) will be treated by the compiler as an out parameter, and decorated in metadata as such. The following code:
namespace TestMakePublic {
public ref class Class1 sealed
{
public:
void foo(int* out1, Object^* out2){}
};
}
Produces a function in metadata which looks like this (ildasm output):
.method public hidebysig newslot virtual final
instance void foo([out] int32& out1,
[out] object& out2) runtime managed
{
.override TestMakePublic.__IClass1PublicNonVirtuals::foo
} // end of method Class1::foo
Note that WinRT does not support "in/out" parameters, so the value of out1 and out2 are only valid for returning from the function, and cannot be trusted as inputs to foo.
It is a C# specific keyword, COM has it too in the IDL syntax. The equivalent in MSVC++ is the [out] attribute.
But no, the compiler is going to reject that with C3115 if you try to use it. Keep in mind that you use the C++/CX language extension to write code that's used by other languages. Which in general support to notion of [out] very poorly. Neither C++, Javascript or .NET languages like vb.net support it. You can see this as well in the .h files in C:\Program Files (x86)\Windows Kits\8.0\Include\WinRT, generated from the .idl files in that same directory that does have the [out] attribute. It was stripped in the .h file by midl.
It doesn't matter anyway since your code will be used in-process so there's no benefit at all from [out] being able to optimize the marshaling of the argument value. Just a plain pointer gets the job done. Having to initialize the argument value in C# code is however inevitable lossage.
You can use:
_Out_opt_
_Out_
But these are available only for private, internal, and protected members AFAIK.

byte array parameter in visual c++

i am new to visual c++.. I had a method in .h file something like this:
public:
void DoSomething();
Here i need to pass byte array as parameter and i need to implement it in .cpp file.. I am working on windows phone 8 for this i need to include visual c++ project of windows phone run time component. I need to use this method in c# class and pass the byte array from there. But i dont know how to declare a byte array method in c++. can any one please help me to find the solution.
In C++/CX, what you're using for a Runtime Component, the signature would look like this (assuming you have a ref class):
void DoSomething(const Platform::Array<uint8>^ something);
This could be called from C# directly by passing in a byte[].
public:
void DoSomething(Byte *);
Or
public:
void DoSomething(unsigned char *);
for exemple.

Calling C# method from C++ code in WP8

I'm interested in calling a C# method from C++ code in Windows Phone 8. I have already learned how to pass a callback function to C++ code from C# via delegate declarations in my C++ code, but I am looking to see if I can do any of the following:
Call certain methods directly from the C++ code. This would involve somehow inspecting the C# object makeup from C++, and seems unlikely to me, but I thought I'd ask you all anyway
Trigger events in the C# code, which can then be handled by C# methods
Use a dispatcher to call C# callbacks in the Main UI thread so that the callbacks can modify UI elements
Use a dispatcher to trigger events in the C# code, (Essentially a merging of the above two points)
In short, I am looking for as many C++ -->C# communication tips as you guys can throw me, I want to learn it all. :)
By getting an object in C# code to implement a Windows RT interface, and passing down a reference to this object, it is possible to do all of the above with a bit of set-up (if I understand correctly - not sure about exactly what you want to do with your Dispatcher examples - you might want to wrap the Dispatcher on the C# side).
Create a Windows Runtime component library.
Define a public interface class in a C++/CX header for the C# to implement (C++ to call) (e.g. ICallback).
Define a public ref class in a C++/CX header for the C++ to implement (C# to call) (e.g. CppCxClass).
Add a method in CppCxClass that passes and stores an ICallback. (A C++ global variable is shown for consiseness, I recommend you review this to see if you can find a better place to store this in your code-base).
ICallback^ globalCallback;
...
void CppCxClass::SetCallback(ICallback ^callback)
{
globalCallback = callback;
}
Reference the WinRT library in your C# code.
C# code: create an instance of CppCxClass using var cppObject = new CppCxClass().
C# code: create a class which implements ICallback (e.g. CSharpCallbackObject).
C# code: pass an instance of CSharpCallbackObject down to C++. E.g. cppObject.SetCallback(new CSharpCallbackObject()).
You can now call C# with globalCallback->CallCsharp(L"Hello C#");. You should be able to extend either ICallback and/or CppCxObject to do the rest of your tasks.
After a lot of headaches trying to figure out the required code, I think it's worth posting the final version here
C++/CX
//.h
[Windows::Foundation::Metadata::WebHostHidden]
public interface class ICallback
{
public:
virtual void Exec( Platform::String ^Command, Platform::String ^Param);
};
//.cpp
ICallback ^CSCallback = nullptr;
void Direct3DInterop::SetCallback( ICallback ^Callback)
{
CSCallback = Callback;
}
//...
if (CSCallback != nullptr)
CSCallback->Exec( "Command", "Param" );
C#
public class CallbackImpl : ICallback
{
public void Exec(String Command, String Param)
{
//Execute some C# code, if you call UI stuff you will need to call this too
//Deployment.Current.Dispatcher.BeginInvoke(() => {
// //Lambda code
//}
}
}
//...
CallbackImpl CI = new CallbackImpl();
D3DComponent.SetCallback( CI);

hooking via hotpatching... non exported class method in dll

I have been studying this method of API hooking using the mechanisms for hotpatching in windows dlls.
http://www.codeproject.com/KB/winsdk/0xF9EB_Hooking.aspx
I was wondering if anyone would know of a way to extend that to hooking non exported functions such as a C++ constructor for an internal class inside of a DLL. I have already know the address via dis-assembly... the problem I am having is how to set up the right calling conventions so that I can call the original function inside of my hook function.
I'm already to the point to where my hook function gets called... the program crashes because I can't return the results of calling the original function.
Lets assume we are talking about hooking an internal class constructor with a prototype something like this:
public __thiscall <class_name>::<class_name>(<single pointer arg to another object>)
depending on how your module is loaded, you can generally just overwrite the relative or absolute addresses at their respective call sites, else you need to make a trampolining function, for which its easier to use something like MS Detours.
In terms of the correct prototype for __thiscall based class member functions, you need some trickery, as you can't generally use __thiscall outside classes. The fastest and easiest way is to use __fastcall and ignore the second parameter. So your first definition becomes void __fastcall myctor(myobj* pObj).
Define it as a typical __stdcall function except that you'll have this pointer in ecx register. If you need this pointer, then use the __asm keyword to get the value:
void __stdcall HookedConstructor( SomeObject *pObject){
HookedClass *pClass;
__asm mov pClass, ecx;
...
}
Note that you'll have to do this at the beginning of the call. Otherwise, the value of ecx register may be overwritten.