concurrency - Simultaneous launch of Multiple Kernels using CUDA for a GPU -

Is it possible to launch two colonels, which work independently, simultaneously for example if I have this quad code

  // Host and device initialization ....... ....... // Launch kernel 1 myMethod1 & lt; & Lt; & Lt;. ... & gt; & Gt; & Gt; (Parameter); // Launch Kernel 2 myMethod2 & lt; & Lt; & Lt; ..... & gt; & Gt; & Gt; (Parameter);  

Assuming that these kernels are free, they have the facility to launch them at the same time allocated for each grid / block. CUDA / OpenCL has this provision. Device with CUDA compute capability 2.0 and better (i.e. Fermi)

Some compute capabilities can perform several kernels simultaneously at 2.0 simultaneously. You can query this capability by calling the application cudaGetDeviceProperties () and check the concurrent kernel property.

Launch the maximum number of kernels that can execute a device simultaneously four.

A CUDA reference can not execute simultaneously with a kernel from another CUDA reference with the kernel.

Using a large amount of texture or local memory is unlikely to be executed with the kernel, along with the other kernel.