Can we restrict number of SM for a kernel in cuda

Hi ,
I am using an RTX5000 GPU for my Cuda application, and I have 48 SM on my GPU card.
I need to run 4 kernels, and I need to restrict 12SM for each kernel.
Is it possible to restrict the SM per kernel?
If possible, please share with me the procedure to do it?
If not, please give me the reason?

CUDA 12.4 introduced Green Contexts that support partitioning the GPU resources.

Prior to Green Contexts the only method would be to limit each kernel to 12 blocks and have each block do additional work.

Or, of course, instead of having each block do additional work, invoking the kernel several time for different data sets.