I have a long list of device pointers in the kernel argument list, sth. like:
global void kernel(float *a, float b, floatc, float *d…).
They really are the same value for all threads and they won’t be changed in the kernel.
But it seems to me they each are assigned a register in every thread, which causes a lot of register pressure.
It there a way to share them among all threads?
Thanks!