[Bug middle-end/85721] bad codegen for looped copy of primitives at -O2 and -O3 (differently bad)

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85721

--- Comment #2 from Jonathan Wakely <redi at gcc dot gnu.org> ---
#include <algorithm>
void copy2(char* out, const char* in, SIZE_T n) {
    std::copy(in, in+n, out);
}

At -O3:

copy2(char*, char const*, unsigned long):
  test rdx, rdx
  jne .L21
  ret
.L21:
  jmp memmove


#include <memory>
void copy3(char* out, const char* in, SIZE_T n) {
    std::uninitialized_copy(in, in+n, out);
}

At -O3:

copy3(char*, char const*, unsigned long):
  test rdx, rdx
  jne .L21
  ret
.L21:
  jmp memmove


Even with -O1 these beat your loop hands down:

copy2(char*, char const*, unsigned long):
  test rdx, rdx
  jne .L11
  ret
.L11:
  sub rsp, 8
  call memmove
  add rsp, 8
  ret
copy3(char*, char const*, unsigned long):
  test rdx, rdx
  jne .L18
  ret
.L18:
  sub rsp, 8
  call memmove
  add rsp, 8
  ret