I worked out an implementation of the Sieve of Atkin for use in solving Project Euler problems. It’s probably not as efficient as it could be, but it does the trick for me.

set nocount on
drop table #atkin
create table #atkin
(
num float
, isPrime bit
)
create clustered index primenums
on #atkin(num)
declare @limit as bigint
declare @sqrt as float
declare @x as int
declare @y as int
declare @n as int
set @limit = 1000000
set @sqrt = sqrt(@limit)
set @x = 1
print 'begin outer loop'
/*
put in candidate primes:
integers which have an odd number of representations by certain quadratic forms
*/
while (@x <= @sqrt)
begin
set @y = 1
while (@y <=@sqrt)
begin
set @n = (4 * (@x * @x)) + (@y * @y)
if ((select num from #atkin where num =@n) is null)
insert into #atkin (num, isPrime) values (@n,0)
if (@n <= @limit AND (@n % 12 = 1 or @n % 12 =5))
update #atkin set isPrime = isPrime ^ 1 where num = @n
set @n = (3 * (@x * @x)) + (@y * @y)
if ((select num from #atkin where num =@n) is null)
insert into #atkin (num, isPrime) values (@n,0)
if (@n <= @limit AND @n % 12 = 7)
update #atkin set isPrime = isPrime ^ 1 where num = @n
set @n = (3 * (@x * @x)) - (@y * @y)
if ((select num from #atkin where num =@n) is null)
insert into #atkin (num, isPrime) values (@n,0)
if (@x >= @y AND @n <= @limit AND @n % 12 = 11)
update #atkin set isPrime = isPrime ^ 1 where num = @n
set @y = @y + 1
end
set @x = @x + 1
end
declare @nsqr as bigint
declare @k as bigint
/*
eliminate composites by sieving
*/
set @n = 5
while ( @n <= @sqrt)
begin
print @n
if ((select isprime from #atkin where num = @n)=1)
begin
set @nsqr = @n * @n
set @k = @nsqr
while (@k <= @limit)
begin
update #atkin set isPrime = 0 where num = @k
set @k = @k + @nsqr
print '@k increment ' + str(@k)
end
end
set @n = @n + 1
print '@n increment ' + str(@n)
end
insert into #atkin (num, isPrime) values (2,1)
insert into #atkin (num, isPrime) values (3,1)
delete from #atkin where isprime <> 1

This is a solution for Project Euler Problem 37. It makes uses of a temporary table called #atkin which is populated with prime numbers using the TSQL routine that uses the Sieve of Atkin, which is also located on this site.

Easy! There are, however, a couple of deliberate errors, that can be easily located with some patience. I donâ€™t want to make things to easy for you little scamps, do I?

/*
process the prime number list
*/
/*
the number we are examining can't be single digits or contain zeros
*/
drop table #candidates
select num
into #candidates
from #atkin
where charindex('0', cast(num as varchar)) = 0
declare @ctr as integer
declare @del as integer
set @ctr = 1
set @del = 0
/*
delete from the candidate list any number that doesn't have a
shorter counterpart in the list of prime numbers, increasing
the string size by 1 with every loop, and bailing out of when
we stop deleting things
*/
while (@del > 0)
begin
delete from #candidates
from #candidates c1
left outer join #atkin a1
on left(c1.num,@ctr) = a1.num
left outer join #atkin a2
on right(c1.num,@ctr) = a2.num
where a1.num is null
or a2.num is not null
set @del = @@rowCount
set @ctr = @ctr + 1
end
/*
output the answer
*/
select sum(num)
from #candidates

/*
generate the number range
*/
create table #polygonalnumbers
(
ctr int
, val bigint
, poltype varchar(3)
)
declare @ctr as integer
set @ctr = 1
while (@ctr < 1000)
begin
insert into #polygonalnumbers
select @ctr,(@ctr*(@ctr+1))/2,'tri'
insert into #polygonalnumbers
select @ctr,@ctr * @ctr *@ctr,'sqr'
insert into #polygonalnumbers
select @ctr,(@ctr*((3*@ctr)-1))/2 ,'pnt'
insert into #polygonalnumbers
select @ctr,(@ctr*((2*@ctr)-1)) ,'hex'
insert into #polygonalnumbers
select @ctr,(@ctr*((5*@ctr)-3))/2 ,'hep'
insert into #polygonalnumbers
select @ctr,(@ctr*((3*@ctr)-2)) ,'oct'
set @ctr = @ctr + 1
end
/*
disregard
*/
delete from #polygonalnumbers
where LEN(val) <> 4
/*
output the answer
*/
select distinct num1.val + num2.val + num3.val + num4.val + num5.val + num6.val
from #polygonalnumbers num1
inner join #polygonalnumbers num2
on RIGHT (num1.val,2) = LEFT(num2.val,1)
and num2.poltype <> num1.poltype
inner join #polygonalnumbers num3
on RIGHT (num2.val,2) = LEFT(num3.val,2)
and num3.poltype <> num2.poltype
and num3.poltype <> num1.poltype
inner join #polygonalnumbers num4
on RIGHT (num3.val,2) = LEFT(num4.val,2)
and num4.poltype <> num3.poltype
and num4.poltype <> num2.poltype
and num4.poltype <> num1.poltype
inner join #polygonalnumbers num5
on RIGHT (num4.val,2) = LEFT(num5.val,2)
and num5.poltype <> num4.poltype
and num5.poltype <> num3.poltype
and num5.poltype <> num2.poltype
and num5.poltype <> num1.poltype
inner join #polygonalnumbers num6
on RIGHT (num5.val,2) = LEFT(num6.val,2)
and num6.poltype <> num5.poltype
and num6.poltype <> num4.poltype
and num6.poltype <> num3.poltype
and num6.poltype <> num2.poltype
and num6.poltype <> num1.poltype
and RIGHT (num6.val,2) = LEFT(num1.val,2)

Easy! There are, however, a couple of deliberate errors, that can be easily located with some patience. I don’t want to make things to easy for you little scamps, do I?