Modern C++ Lamentations

[Steven Schveighoffer]

On 12/31/18 1:43 PM, Walter Bright wrote:
> On 12/31/2018 8:05 AM, Steven Schveighoffer wrote:
>  > (see my reply to that post).
> 
> Your reply is:
> 
> "Ooh, that's interesting. Same issue with the D version.
> I had to work a bit on it, but this does work and is 172ms vs the ~1000ms:
> 
>    return
>      recurrence!"a[n-1]+1"(1)
>      .then!((z) {
>         auto ztotal = z * z;
>         return iota(1, z + 1).then!((x) {
>             auto xtotal = x * x;
>             return iota(x, z + 1)
>                .filter!(y => y * y + xtotal == total)
>                .map!(y => tuple(x,y,z));
>                });
>         });
> "
> 
> Atila, can you please update the blog post?

This isn't a fair comparison though -- I'm doing work here that in the 
other languages the compiler is doing (hoisting the multiplications 
outside the inner loops). It's not a straight port.

I just wanted to point out that this accounts for how the rust and C++ 
versions are faster than the D range versions. It would be good to look 
into why the D compilers are not seeing that optimization possibility.

-Steve