[Bug tree-optimization/115040] New: Missed optimization opportunity in std::find

[Bug tree-optimization/115040] New: Missed optimization opportunity in std::find

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115040

            Bug ID: 115040
           Summary: Missed optimization opportunity in std::find
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hiraditya at msn dot com
  Target Milestone: ---

https://gcc.godbolt.org/z/s3hv15935


```
#include <algorithm>
#include <vector>
#include <cstdint>

bool find_epi8(const std::vector<int8_t>& v) {
    return std::find(v.begin(), v.end(), 42) != v.end();
}

bool find_epi32(const std::vector<int32_t>& v) {
    return std::find(v.begin(), v.end(), 42) != v.end();
}
```

$ gcc -O3 -ftree-vectorize -march=pantherlake

```
find_epi8(std::vector<signed char, std::allocator<signed char> > const&):
        mov     rcx, QWORD PTR [rdi+8]
        mov     rdx, QWORD PTR [rdi]
        mov     rsi, rcx
        sub     rsi, rdx
        mov     rax, rsi
        sar     rax, 2
        test    rax, rax
        jle     .L2
        lea     rax, [rdx+rax*4]
        jmp     .L8
.L3:
        cmp     BYTE PTR [rdx+1], 42
        je      .L23
        cmp     BYTE PTR [rdx+2], 42
        je      .L24
        cmp     BYTE PTR [rdx+3], 42
        je      .L25
        add     rdx, 4
        cmp     rdx, rax
        je      .L26
.L8:
        cmp     BYTE PTR [rdx], 42
        jne     .L3
.L21:
        cmp     rcx, rdx
        setne   al
        ret
.L26:
        mov     rsi, rcx
        sub     rsi, rdx
.L2:
        cmp     rsi, 2
        je      .L9
        cmp     rsi, 3
        je      .L10
        cmp     rsi, 1
        je      .L11
        xor     eax, eax
        ret
.L10:
        cmp     BYTE PTR [rdx], 42
        je      .L21
        add     rdx, 1
.L9:
        cmp     BYTE PTR [rdx], 42
        je      .L21
        add     rdx, 1
.L11:
        cmp     BYTE PTR [rdx], 42
        sete    al
        cmp     rcx, rdx
        setne   dl
        and     eax, edx
        ret
.L23:
        add     rdx, 1
        cmp     rcx, rdx
        setne   al
        ret
.L24:
        add     rdx, 2
        cmp     rcx, rdx
        setne   al
        ret
.L25:
        add     rdx, 3
        cmp     rcx, rdx
        setne   al
        ret
find_epi32(std::vector<int, std::allocator<int> > const&):
        mov     rcx, QWORD PTR [rdi+8]
        mov     rdx, QWORD PTR [rdi]
        mov     rax, rcx
        sub     rax, rdx
        mov     rsi, rax
        sar     rax, 4
        sar     rsi, 2
        test    rax, rax
        jle     .L28
        sal     rax, 4
        add     rax, rdx
        jmp     .L34
.L29:
        cmp     DWORD PTR [rdx+4], 42
        je      .L48
        cmp     DWORD PTR [rdx+8], 42
        je      .L49
        cmp     DWORD PTR [rdx+12], 42
        je      .L50
        add     rdx, 16
        cmp     rdx, rax
        je      .L51
.L34:
        cmp     DWORD PTR [rdx], 42
        jne     .L29
.L47:
        cmp     rcx, rdx
        setne   al
        ret
.L51:
        mov     rsi, rcx
        sub     rsi, rdx
        sar     rsi, 2
.L28:
        cmp     rsi, 2
        je      .L35
        cmp     rsi, 3
        je      .L36
        cmp     rsi, 1
        je      .L37
        xor     eax, eax
        ret
.L36:
        cmp     DWORD PTR [rdx], 42
        je      .L47
        add     rdx, 4
.L35:
        cmp     DWORD PTR [rdx], 42
        je      .L47
        add     rdx, 4
.L37:
        cmp     DWORD PTR [rdx], 42
        sete    al
        cmp     rcx, rdx
        setne   dl
        and     eax, edx
        ret
.L48:
        add     rdx, 4
        cmp     rcx, rdx
        setne   al
        ret
.L49:
        add     rdx, 8
        cmp     rcx, rdx
        setne   al
        ret
.L50:
        add     rdx, 12
        cmp     rcx, rdx
        setne   al
        ret
```


clang lowers both the calls to (w)memchr

--- Comment #1 from Sam James <sjames at gcc dot gnu.org> ---
*** Bug 115041 has been marked as a duplicate of this bug. ***

[Bug tree-optimization/115040] Missed optimization opportunity in std::find of std::vector elements

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115040

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement

[Bug tree-optimization/115040] Missed optimization opportunity in std::find of std::vector elements

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115040

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
> clang lowers both the calls to (w)memchr

Is that with libc++ or libstdc++?

[Bug tree-optimization/115040] Missed optimization opportunity in std::find of std::vector elements

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115040

--- Comment #3 from Sam James <sjames at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> > clang lowers both the calls to (w)memchr
> 
> Is that with libc++ or libstdc++?

It's libc++. Checked locally too..

[Bug libstdc++/115040] Missed optimization opportunity in std::find of std::vector elements

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115040

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Sam James from comment #3)
> (In reply to Andrew Pinski from comment #2)
> > > clang lowers both the calls to (w)memchr
> > 
> > Is that with libc++ or libstdc++?
> 
> It's libc++. Checked locally too..

Then this is partly a dup of bug 88545

[Bug libstdc++/115040] Missed optimization opportunity in std::find of std::vector elements

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115040

--- Comment #5 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #4)
> Then this is partly a dup of bug 88545

Yes, that would manually transform the find_epi8 case into a memchr call, but
Clang doesn't need a manual transform in the library, it does it in the
compiler.

Tamar's plan to remove the manual loop unrolling in our std::find_if might help
here, but I'm not sure.