How can a cuda kernel send informations to others cuda kernels ?

Hi , i want to know how can cuda kernels communicate with each other , and if a cuda kernel can send information to another kernel ( which memory is made for this type of communication ? ) .
thank u for ur help.

Hi, I think it’s a wrong place for this question!
First of all what do you mean by kernel ?
If you mean separate instantiations of the same kernel in form of blocks of threads:
you need to use global memory with atomic functions to assure consistence.
If you mean entirely different kernels (functions) you need to use global memory as a storage and pass a pointer to it to both kernels. Of course, since there’s a chance your kernels(functions) can be executed concurrently you should use the same stream or explicitly synchronize their execution.

Hope it helps,

thanks a lot Glupol :)