In the following CUDA kernel code, number the operations from fastest $(1)$ to slowest $(3).($hint: identify global, shared and local variables and then compare access costs$).$

$\_\_$global$\_\_$ void foo $($float x$,$ float y$,$ float z$)$

$\{$

$\_\_$shared$\_\_$ float a$,$b$,$c;

$$ float s$,$t$,$u;

$$

$$

$$ s$=$x;

$$ a$=$b;

$$

$$ y$=$z;$$

$\}$