If you're not required to produce the actual primes but only test if there exists a pair of primes p and q such that p+q == N, you could make this very simple based on the Goldbach conjecture. All even numbers can be expressed as the sum of two primes. So return True if the number is even and check if N-2 is prime for odd numbers (because 2 is the only even prime and that's the only prime that will produce another odd number when starting from an odd number). This will boil down to a single prime test of N-2 only for odd numbers.
def primePart(N):
return N%2==0 or all((N-2)%p for p in range(3,int(N**0.5)+1,2))
primePart(3432) # True
primePart(37+2) # True
primePart(13+41) # True
primePart(123) # False
If you want to actually find a pair of primes that add up to N, you can generate primes up to N and return the first prime >= N/2 where N - prime is one of the primes already found:
def findPQ(N):
if not primePart(N): return
if N%2: return 2,N-2
isPrime = [0]+[1]*N
for p in range(3,N,2):
if not isPrime[p]: continue
if 2*p>=N and isPrime[N-p]: return p,N-p
isPrime[p*p::p] = [0]*len(isPrime[p*p::p])
output:
findPQ(3432) # (1723, 1709)
findPQ(12345678) # (6172879, 6172799)
To go beyond 10^9 you will need a more memory efficient algorithm than the sieve of Eratosthenes that is just as fast. This can be achieved with a dictionary of multiples of primes to skip:
def findPQ(N):
if not primePart(N): return
if N%2: return 2,N-2
skip,primes = {},{2}
for p in range(3,N,2):
if p in skip:
prime = skip.pop(p)
mult = p + 2*prime
while mult in skip: mult += 2*prime
if mult <= N: skip[mult] = prime
else:
if 2*p>=N and N-p in primes: return p,N-p
if p*p<=N: skip[p*p]=p
if 2*p<=N: primes.add(p)
output (takes a while but doesn't bust memory space):
findPQ(1234567890) # (617283983, 617283907)