Gpu.Default.Launch(() =>
{
// Note you are calling a kernel inside a kernel.
var lp = new LaunchParam(1, 1024);
DeviceRuntime.Launch(YourKernel, lp)
}, new LaunchParam(1, 1));
What you cannot do, (unfortunately), is call Cublas inside a kernel.
For Cublas I urge you to use the latest version which in now a separate nuget package: https://www.nuget.org/packages/Alea.CudaToolkit/
AleaGpu supports dynamic parallelism.
You can do it this way:
Use the latest AleaGpu release: https://www.nuget.org/packages/Alea/3.0.4-beta3
What you cannot do, (unfortunately), is call Cublas inside a kernel. For Cublas I urge you to use the latest version which in now a separate nuget package: https://www.nuget.org/packages/Alea.CudaToolkit/