# Sieve31, my sieve of Eratosthenes returning IEnumerable<int>

This very fast, simple sieve quickly finds 31 bit primes. It uses a memory efficient BitArray for odd numbers.

How a 32 bit int is a 31 bit prime

Since a prime number must be > 1, the sign bit has no bearing, leaving 31 bits to store the positive numbers. This differs from uint which truly would be 32 bit primes.

See this Microsoft link on Magic Numbers.

If you find the value 2147483647 on that page you will see this text:

The maximum signed 32 bit value. Largest 31 bit prime.

On an unrelated note, the Magic Numbers link has some pretty interesting things in it.

public static class Sieve31
{
public static IEnumerable<int> Primes(int upperLimit)
{
if (upperLimit < 2)
{
throw new ArgumentException("Upper Limit be must greater than or equal to 2.");
}
yield return 2;
if (upperLimit == 2)
{
yield break;
}
// Check odd numbers for primality
const int offset = 3;
Func<int, int> ToNumber = delegate(int index)  { return (2 * index) + offset; };
Func<int, int> ToIndex  = delegate(int number) { return (number - offset) / 2; };
var bits = new BitArray(ToIndex(upperLimit) + 1, defaultValue: true);
var upperSqrtIndex = ToIndex((int)Math.Sqrt(upperLimit));
for (var i = 0; i <= upperSqrtIndex; i++)
{
// If this bit has already been turned off, then its associated number is composite.
if (!bits[i]) continue;
var number = ToNumber(i);
// The instant we have a known prime, immediately yield its value.
yield return number;
// Any multiples of number are composite and their respective bits should be turned off.
for (var j = ToIndex(number * number); (j > i) && (j < bits.Length); j += number)
{
bits[j] = false;
}
}
// Output remaining primes once bit array is fully resolved:
for (var i = upperSqrtIndex + 1; i < bits.Length; i++)
{
if (bits[i])
{
}
}
}
}


Example Usage

Here’s a simple example that counts the number of primes found and finds the largest one.

var count = 0;
var largest = 0;
var primes = Sieve31.Primes(int.MaxValue);
foreach (var prime in primes)
{
count++;
largest = prime;
}


Worst Case Scenario: int.MaxValue

The BitArray will have a Length of 1,073,741,823 bits, which is 128 megabytes. This will yield 105,097,565 primes.

If you want to store the primes to a List<int>, the code is quite easy:

var primeList = Sieve31.Primes(int.MaxValue).ToList();


The resulting list would require 409 megabytes, in addition to the 128 for the BitArray. On my laptop this takes about 40 seconds. Since BitArray is not thread safe, this is close to the best I can hope for on a single thread.

An indexed list is strongly recommended

When processing primes it is often, but not always, preferred to use an indexed list. Assuming, of course, that you have sufficient memory to do so. Worst case you need 530 megabytes to generate the list, and once it’s done, you get back 128 MB but now would have a fast Count and a fast index at your fingertips.

If you don’t use a list, and need to loop over the primes more than once, the BitArray is completely regenerated every time, which could degrade performance.

A Challenge Problem

Given an extremely large int upper limit, perhaps anything over 1 billion, build a random list of 100 primes. You cannot hardcode any known prime counts.

Short, Easy Solution

The easiest solution is to use the recommended indexed list, if memory allows.

private IList<int> GetRandom100Easy(int upperLimit)
{
var primeList = Sieve31.Primes(upperLimit).ToList();
var random = new Random();
for (var i = 0; i < 100; i++)
{
var index = random.Next(primeList.Count);
}
}


Longer, Low Memory Solution

If you don’t have sufficient memory to produce the largest possible list of primes, the solution is a lot longer and slower, as it requires two-passes over the enumerable collection. For the 2nd pass, you can exit early once the full answer is known.

private IList<int> GetRandom100LowMemory(int upperLimit)
{
// To produce this answer without a list of primes requires two-passes.
// The 2nd pass can exit early.
var primes = Sieve31.Primes(upperLimit);
// Fully loop over to get count
var primeCount = 0;
foreach (var prime in primes)
{
primeCount++;
}
// Initialize dictionary of 100 random sequences.
IDictionary<int, int> dict = new Dictionary<int, int>();
var random = new Random();
for (var i = 0; i < 100; i++)
{
}
// For early loop termination, find max sequence.
var maxSequence = dict.Keys.Max();
// Loop again to assign primes to dict.
var sequence = 0;
foreach (var prime in primes)
{
sequence++;
if (dict.ContainsKey(sequence))
{
dict[sequence] = prime;
if (sequence == maxSequence)
{
break;
}
}
}
return dict.Values.ToList();
}


As you review both solutions, I’m sure you understand why working with an indexed list is preferred.

Questions

Being this is CR, there is always an implied question of “Do you have any constructive comments?”

Are there better ways to address the challenge problem?

• If you're worried about a List taking up space, you can write a Primes method that, instead of returning the prime numbers, just returns the BitArray and the count. Then, you can select 100 primes from the BitArray. No need to store all the prime numbers, just the 100 – Dennis_E Jun 1 '15 at 14:35
• @Dennis_E Not really worried about if for 31 bit primes. If I follow you suggestion, the sieve no longer becomes simple. Counting adds 8 seconds. And someone could modify the bit flags of the returned array.Not that these issue can't be overcome, except for the keep it simple part. – Rick Davin Jun 1 '15 at 15:12
• It was just an idea... The BitArray doesn't have to be public. You just have to split up the implementation. One (private) method generates the BitArray, and GetRandom100() uses it to generate 100 primes. But I don't see how counting could add 8 whole seconds. Instead of yield return number, you have count++ – Dennis_E Jun 1 '15 at 16:07

About the only huge, performance-impacting change I can see that would be necessary is replacing the BitArray for a literal bool array.

This actually has a huge speed impact.

So, making that change, and inverting the related conditions causes us to have the following method:

public static IEnumerable<int> Primes(int upperLimit)
{
if (upperLimit < 2)
{
throw new ArgumentException("Upper Limit be must greater than or equal to 2.");
}
yield return 2;
if (upperLimit == 2)
{
yield break;
}
// Check odd numbers for primality
const int offset = 3;
Func<int, int> ToNumber = delegate (int index) { return (2 * index) + offset; };
Func<int, int> ToIndex = delegate (int number) { return (number - offset) / 2; };
var bits = new bool[ToIndex(upperLimit) + 1];
var upperSqrtIndex = ToIndex((int)Math.Sqrt(upperLimit));
for (var i = 0; i <= upperSqrtIndex; i++)
{
// If this bit has already been turned off, then its associated number is composite.
if (bits[i]) continue;
var number = ToNumber(i);
// The instant we have a known prime, immediately yield its value.
yield return number;
// Any multiples of number are composite and their respective bits should be turned off.
for (var j = ToIndex(number * number); (j > i) && (j < bits.Length); j += number)
{
bits[j] = true;
}
}
// Output remaining primes once bit array is fully resolved:
for (var i = upperSqrtIndex + 1; i < bits.Length; i++)
{
if (!bits[i])
{
}
}
}


You'll notice that I inverted all the booleans on the bits value, to make sure that requiring a default of false instead of true now would not create issues.

As far as speed goes, the difference seems to be performing in 63.3% of the time of your original method. (For a limit of 100,000,000 the modified version I have here is about 2486ms, yours is about 3917ms.)

Other than that, there is really not much you can do about performance overall.

As far as readability, I would never one-line an if statement without braces.

if (bits[i]) continue;


Break that down into two-lines, it won't hurt anything. ;)

if (bits[i])
continue;


You could also do for a little whitespace to break logical portions of your algorithm out.

 public static IEnumerable<int> Primes(int upperLimit)
{
if (upperLimit < 2)
{
throw new ArgumentException("Upper Limit be must greater than or equal to 2.");
}

yield return 2;

if (upperLimit == 2)
{
yield break;
}

// Check odd numbers for primality
const int offset = 3;
Func<int, int> ToNumber = delegate (int index) { return (2 * index) + offset; };
Func<int, int> ToIndex = delegate (int number) { return (number - offset) / 2; };

var bits = new bool[ToIndex(upperLimit) + 1];
var upperSqrtIndex = ToIndex((int)Math.Sqrt(upperLimit));

for (var i = 0; i <= upperSqrtIndex; i++)
{
// If this bit has already been turned off, then its associated number is composite.
if (bits[i])
continue;

// The instant we have a known prime, immediately yield its value.
var number = ToNumber(i);
yield return number;

// Any multiples of number are composite and their respective bits should be turned off.
for (var j = ToIndex(number * number); (j > i) && (j < bits.Length); j += number)
{
bits[j] = true;
}
}

// Output remaining primes once bit array is fully resolved:
for (var i = upperSqrtIndex + 1; i < bits.Length; i++)
{
if (!bits[i])
{
}
}
}


I know, old question, but I felt it deserved a little attention.

• It can be faster. See my answer to your answer :-) – t3chb0t Aug 27 '15 at 19:57
• Sorry for late response. Been out of town. Very nice answer. For past 18 months of reading CR I have believed that BitArray is the "best" route. It is still the smallest memory route but certainly not the fastest. You are right about speed boost, but it does come at price of using 6x memory. I discovered that internally a BitArray actually uses a int[] as its backing field so that means a lot of internal bit twiddling is done to make it appears as a stream of bits. – Rick Davin Aug 29 '15 at 22:47
• This answer works fine for Sieve31 but would produce an out of memory exception for Sieve32, where one must still use BitArray. But since this question is about Sieve31, I have marked your answer as the accepted answer. – Rick Davin Aug 29 '15 at 23:02
• @RickDavin That's very curious - you could potentially make an int[] of int[(ToIndex(UpperLimit) + 1) / 8] and use some "cute" math to get the values/etc you need. I might try an experimentation involving just that when I am back home - I'm on vacation right now so it won't happen until at least Monday, if you are interested in the result. :) – Der Kommissar Aug 29 '15 at 23:10
• No need. What you have just described is what BitArray already does. See BitArray Reference Source at referencesource.microsoft.com/#mscorlib/system/collections/… – Rick Davin Aug 29 '15 at 23:15

You can still make this function faster and gain another second if you don't use IEnumerable and an enumerator but for example a List<int>.

Like this (based on @EBrown's code):

public static List<int> Primes(int upperLimit)
{
var result = new List<int>();

if (upperLimit < 2)
{
throw new ArgumentException("Upper Limit be must greater than or equal to 2.");
}

if (upperLimit == 2)
{
return result;
}

// Check odd numbers for primality
const int offset = 3;

Func<int, int> toNumber = index => (2 * index) + offset;
Func<int, int> toIndex = number => (int)((number - offset) * 0.5);

var bits = new bool[toIndex(upperLimit) + 1];

var upperSqrtIndex = toIndex((int)Math.Sqrt(upperLimit));

for (var i = 0; i <= upperSqrtIndex; i++)
{
// If this bit has already been turned off, then its associated number is composite.
if (bits[i]) { continue;}

var number = toNumber(i);

// The instant we have a known prime, immediately yield its value.

// Any multiples of number are composite and their respective bits should be turned off.
for (var j = toIndex(number * number); (j > i) && (j < bits.Length); j += number)
{
bits[j] = true;
}
}
// Output remaining primes once bit array is fully resolved:
for (var i = upperSqrtIndex + 1; i < bits.Length; i++)
{
if (!bits[i])
{
}
}

return result;
}


Here are two profiler screenshots:

@EBrown's optimized method:

Without enumerator:

Original method:

Tested code:

for (int i = 0; i < 3; i++)
{
//var primes1 = Primes(100000000);
var primes1 = Primes1(100000000).ToList();
//var primes1 = PrimesOriginal(100000000).ToList();
}

• I am the OP. Glad to see that you properly credited @EBrown. Despite that, I consider you answer to be poor and weak since it throws an out of memory exception for an upper limit of int.MaxValue. – Rick Davin Aug 30 '15 at 12:52
• @RickDavin consider that people give their reviews for free, you'd do well to be slightly less rude. This implementation appears to be O(n) which would mean that accessing int.MaxValue would require 4 billion bytes of memory, or 4GB of RAM - plus the extra array that OP has, which brings it to 8GB. Perhaps your PC just isn't up to scratch? :-) Even the wikipedia article acknowledges memory limitations as the prime issue with SoE. – Dan Pantry Nov 11 '15 at 15:19
• I'm also informed by @EBrown that C# will implicitly throw OOM errors when you create objects of > 2GB in size unless you enable a flag, gcAllowVeryLargeObjects. – Dan Pantry Nov 11 '15 at 15:21

A Challenge Problem

Given an extremely large int upper limit, perhaps anything over 1 billion, build a random list of 100 primes. You cannot hardcode any known prime counts.

As always with challenges of this type, the trick lies in doing only the work necessary instead of lots more. That's why the solutions proposed here need a thousand times as much CPU and memory as necessary, despite the brilliant micro-optimisations - a fact that calls to mind The Sieve of Smith (I hope that little gem stays around for a while, as a warning about what happens when you start micro-optimising before you're done with algorithmical and architectural optimisation).

The key here is this: the range from which the random primes are to be drawn is huge but the number of random primes to be drawn is minuscule.

Sieving all numbers up to 2^32 requires on the order of several CPU seconds even when done in C/C++ (2 seconds on a single core for an optimised implementation), and the particulars of CLR and C# compiler slow things down further by about one order of magnitude. However, the challenge doesn't ask for the primality classification of four billion numbers, and it doesn't ask for 203,280,221 primes either. It asks only for a hundred random primes.

Hence a quick, easy and brutally simple solution is to implement a segmented (or rather windowed) sieve. The primes up to 64K - the square root of the upper limit - need to be sieved out in the classic manner and kept around, preferably as an array of integers. But that range is tiny and the list needs to hold only a few thousand primes; generating it should take only a few milliseconds even with the most simple and straightforward (unoptimised) code.

To draw a random prime, draw a random number k between 0 and 2^32-1 and sieve the window [k-335, k]. Then scan the window backwards from k and return the first prime you encounter. The window size is based on the fact that the gaps between primes up to 2^32 are no wider than 336; the value needs to be adjusted for processing ranges beyond 4,302,407,359. See The Gaps Between Primes and First Occurrence Prime Gaps, for example.

In this way, drawing a hundred primes requires only the sieving of 2^16 + 100 * 336 = 99136 numbers instead of 2^32, or about 1/43323th (four orders of magnitude less). The windows are so small compared to the total range that you could even scan them using trial division and still be lots faster than with a full-range sieve.

Here's a Delphi implementation (not my weapon of choice but I gotta learn it somehow) that draws 100 random primes up to 1,000,000,000 in less than a millisecond:

function HighestPrimeNotAfter (n: Cardinal): Cardinal;
const
WINDOW_SIZE = 336;
var
is_composite: array of Boolean;
max_factor, window_start: Cardinal;
i, prime, start, stride, j, rest: Cardinal;
begin
SetLength(is_composite, WINDOW_SIZE);
max_factor := Trunc(sqrt(n));

if n > WINDOW_SIZE then
window_start := n - WINDOW_SIZE + 1
else
window_start := 2;

for i := Low(primes_up_to_64K) to High(primes_up_to_64K) do begin
prime := primes_up_to_64K[i];
if prime > max_factor then
BREAK;
start := prime * prime;
stride := prime shl (prime and 1);  // no double-stepping for p == 2, because of the squaring
if start >= window_start then
j := start - window_start
else begin
rest := (window_start - start) mod stride;
if rest = 0 then j := 0 else j := stride - rest;
end;
while j < WINDOW_SIZE do begin
is_composite[j] := true;
Inc(j, stride);
end;
end;

for j := High(is_composite) downto Low(is_composite) do
if not is_composite[j] then begin
result := window_start + j;
EXIT;
end;

assert(false);
result := 0;  // suppress compiler warning
end;


The timing includes 0.2 ms for sieving the primes up to 64K into an array of Cardinal, the code for which is trivial and not shown here.

• Just brilliant :) – pkuderov Mar 14 '16 at 13:37