Re: Toy computational "benchmark" in Ada (new blog post)

[David Trudgett <dktrudgett@gmail.com>]

Il giorno venerdì 7 giugno 2019 15:57:36 UTC+10, Paul Rubin ha scritto:
> 
> I think this is it below.  I don't understand why it's accumulating the
> values into a stack slot instead of a register, but I don't know the x86
> all that well.
> 
>     .L17:
>             addq    $1, %rax
>     .LEHB4:
>             movq    416(%r12), %rdx
>             movsd   -8(%rdx,%rax,8), %xmm0
>     .LEHE4:
>             mulsd   %xmm0, %xmm0
>             cmpq    %rax, %rbp
>             addsd   8(%rsp), %xmm0
>             movsd   %xmm0, 8(%rsp)
>             jne     .L17
>             jmp     .L19

Instead, you should have something like:

.L5:
    movsd (%rax), %xmm0
    addq $8, %rax
    cmpq %rbx, %rax
    mulsd %xmm0, %xmm0
    addsd %xmm0, %xmm1
    jne .L5

It would appear you did not turn on optimisations since, as you noticed, it is directly using the stack frame variable rather than keeping it in a register.

Cheers,
David