Compiler cannot remove unneccessary loop in case of unwrap_unchecked and Vec collect() optimization

[AngelicosPhosphoros]

I tried this code:

pub fn unwrap_all(vv: Vec<Option<Box<i32>>>)->Vec<Box<i32>>{
    vv
        .into_iter()
        .map(|x| unsafe{
            x.unwrap_unchecked()
        })
        .collect()
}

I expected to see this happen: code compiles into noop essentially, like transmute.

Expected code

example::unwrap_all_transmute:
        mov     rax, rdi
        mov     rcx, qword ptr [rsi + 16]
        mov     qword ptr [rdi + 16], rcx
        movups  xmm0, xmmword ptr [rsi]
        movups  xmmword ptr [rdi], xmm0
        ret

Instead, this happened: code iterates over the vec without toching real backing buffer. This is a loop without side effects so it should be removed by LLVM.

Generated code

example::unwrap_all:
        mov     rax, rdi
        mov     rcx, qword ptr [rsi]
        mov     r8, qword ptr [rsi + 8]
        mov     rdi, qword ptr [rsi + 16]
        mov     rsi, rcx
        test    rdi, rdi
        je      .LBB0_8
        lea     r9, [8*rdi - 8]
        mov     edx, r9d
        shr     edx, 3
        add     edx, 1
        and     rdx, 7
        je      .LBB0_2
        neg     rdx
        mov     rsi, rcx
.LBB0_4:
        add     rsi, 8
        inc     rdx
        jne     .LBB0_4
        cmp     r9, 56
        jae     .LBB0_6
        jmp     .LBB0_8
.LBB0_2:
        mov     rsi, rcx
        cmp     r9, 56
        jb      .LBB0_8
.LBB0_6:
        lea     rdx, [rcx + 8*rdi]
.LBB0_7:
        add     rsi, 64
        cmp     rsi, rdx
        jne     .LBB0_7
.LBB0_8:
        sub     rsi, rcx
        sar     rsi, 3
        mov     qword ptr [rax], rcx
        mov     qword ptr [rax + 8], r8
        mov     qword ptr [rax + 16], rsi
        ret

Meta

rustc --version --verbose:

rustc 1.58.0-nightly (936f2600b 2021-11-22)

Godbolt link with code

[the8472]

dup of #79308

[AngelicosPhosphoros]

Yeah, forgot about those.