help with prime pairs code

[Jabari Zakiya]

On Monday, 3 February 2025 at 00:04:17 UTC, Sergey wrote:
> On Sunday, 2 February 2025 at 22:40:41 UTC, Jabari Zakiya wrote:
>> I am really impressed!
>> D is phenomenally  memory efficient with this code.
>> I just ran the D code for some really large n values.
>> On my older Lenovo Legion 5, using Linux (June 2024) w/16GB
>> I was able to go up to n = 1,000,000,0000 and it took up 
>> ~14.4GB max, out of ~14.9GB usable.
>> I got a new, fast TUXEDO Gen 3, w/64GB (Jan 2025), so I'm 
>> going to how high I can go on it.
>
> Nice. Maybe create a github repo with results from different 
> languages. It will be interesting to see the comparison.
>
> I haven't checked other parts of the code - maybe it is 
> possible to improve the performance a bit more (to reduce some 
> allocations, use `appender` instead of `~=` operator)
>
> But will see what other will produce first :)

Good idea. I'll see if I'm up to it.

Another code opt for you.

I translated this Ruby code:


       lhr_del = lhr_mults.map { |r_del| (r_del > ndiv2 ? n - 
r_del : r_del) }

       lhr_rmax = Integer.sqrt(n-2) + 2                 # the 
lhr_rmax high bound
       lhr_rmax_cnt = lhr.count { |r| r <= lhr_rmax }   # count of 
lhr <= lhr_rmax
       lhr -= lhr_del                                   # lhr pcp 
prime residues
       pcp_rmax = lhr.select { |r| r if r <= lhr_rmax } # lhr pcp 
prime residues <= lhr_rmax


To this D code. It works, seems fast, can it be done shorter, 
faster, more idiomatic?

       auto lhr_del = lhr_mults.map!((r_del) => r_del > ndiv2 ? n 
- r_del : r_del).array;
       lhr_del.sort!("a < b");

       auto lhr_rmax = 2 + cast(uint) sqrt(cast(double) n-2);
       auto lhr_rmax_cnt = 0;
       foreach(r; lhr) { if (r > lhr_rmax) break; lhr_rmax_cnt += 
1;}
       lhr = setDifference(lhr, lhr_del).array;
       uint[] pcp_rmax;
       foreach(pcp; lhr) { if (pcp > lhr_rmax) break; pcp_rmax ~= 
pcp; }