Used to check ASM generation, filtered via Compiler Explorer. The only major differences are:
eq_ignore_case
This is insignificant, because it is always called with literals for u, and this produces the same ASM in that case, and otherwise the initial implementation is logically unsound.
2c2,5
< test rdx, rdx
---
> mov r9, qword ptr [rdi + 8]
> cmp r9, rdx
> cmova r9, rdx
> test r9, r9When we use it with a literal u, we get the same ASM.
step_by
This can't be trivially fixed, since the compiler can't decipher the bounds checks are redundant.
0a1,17
> example::AsciiStr::step_by:
> push rax
> mov rax, qword ptr [rdi + 8]
> sub rax, rsi
> jb .LBB0_1
> add qword ptr [rdi], rsi
> mov qword ptr [rdi + 8], rax
> mov rax, rdi
> pop rcx
> ret
> .LBB0_1:
> lea rdi, [rip + .L__unnamed_1]
> lea rdx, [rip + .L__unnamed_2]
> mov esi, 43
> call qword ptr [rip + core::panicking::panic@GOTPCREL]
> ud2
>
1a19
> push rax
3c21
< je .LBB0_4
---
> je .LBB1_5
5c23
< .LBB0_2:
---
> .LBB1_2:
9c27
< ja .LBB0_4
---
> ja .LBB1_5
14a33,35
> mov rcx, qword ptr [rdi + 8]
> test rcx, rcx
> je .LBB1_6
18,20c39,43
< add qword ptr [rdi + 8], -1
< jne .LBB0_2
< .LBB0_4:
---
> add rcx, -1
> mov qword ptr [rdi + 8], rcx
> jne .LBB1_2
> .LBB1_5:
> pop rax
21a45,50
> .LBB1_6:
> lea rdi, [rip + .L__unnamed_1]
> lea rdx, [rip + .L__unnamed_2]
> mov esi, 43
> call qword ptr [rip + core::panicking::panic@GOTPCREL]
> ud2
24a54,63
>
> .L__unnamed_1:
> .ascii "called `Option::unwrap()` on a `None` value"
>
> .L__unnamed_3:
> .ascii "/app/example.rs"
>
> .L__unnamed_2:
> .quad .L__unnamed_3
> .asciz "\017\000\000\000\000\000\000\000u\000\000\000&\000\000"Safety go brrr.