he executable takes two command line arguments: N and blockSize, which means we like to

find all prime numbers in the range of $3$ to N inclusively. We ignore the edge cases of checking

number of $0,1$ and $2$ to simplify your program. Also, we like to launch a grid of threads on GPU

that consists of ceil$(($N $+1)/2\mathrm{.}0/$ blockSize$)$ blocks. You have to explore and use the long

long int type in CUDA.