[Bug middle-end/108506] bit_cast from 32-byte vector generates worse code than memcpy

[m.cencora at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108506

--- Comment #5 from m.cencora at gmail dot com ---
Almost, there is still some unnecessary stack spill with bit_cast case:

"deserialize(unsigned char const*, Foo*)":
        push    rbp
        xor     eax, eax
        mov     rbp, rsp
        and     rsp, -32
        vmovdqa ymm1, YMMWORD PTR .LC0[rip]
.L2:
        vmovdqu ymm0, YMMWORD PTR [rdi+rax]
        vpshufb ymm0, ymm0, ymm1
        vmovdqa YMMWORD PTR [rsp-32], ymm0 // unnecessary
        vmovdqu YMMWORD PTR [rsi+rax], ymm0
        add     rax, 32
        cmp     rax, 32768
        jne     .L2
        vzeroupper
        leave
        ret
"deserialize2(unsigned char const*, Foo*)":
        vmovdqa ymm1, YMMWORD PTR .LC0[rip]
        xor     eax, eax
.L7:
        vmovdqu ymm0, YMMWORD PTR [rdi+rax]
        vpshufb ymm0, ymm0, ymm1
        vmovdqu YMMWORD PTR [rsi+rax], ymm0
        add     rax, 32
        cmp     rax, 32768
        jne     .L7
        vzeroupper
        ret