Thread limitation before kernel invocation


Suppose we have a kernel invocation like this: mykernel<<<10, 128>>>() and we don’t want to interfere with the invocation, I mean we don’t want to change that line in the code (the invocation line), but we want to prevent the kernel invocation from having 128 blocks. Is there any way, like a cuda function, which we can use before the invocation to change the number of threads? For example :

mykernel<<<10, 128>>>();

Thank you in advance.