https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96252
--- Comment #1 from Will Wray <wjwray at gmail dot com> ---
Here's the code, compiler invocation and codegen output.
The longer codegen expands memcpy to copy the std::array by-value arguments.
-fno-inline shows the compiler call the first function from the second, then,
when it does so, it has to copy the arguments as both functions pass by value.
/** compare_differing_codegen.cpp ******************/
#include <algorithm>
#include <array>
using cmp = std::array<int,64>;
bool cmp_x(cmp l, cmp r) noexcept {
return std::lexicographical_compare(begin(l),end(l)
,begin(r),end(r));
}
bool cmp_y(cmp l, cmp r) noexcept {
return std::lexicographical_compare(begin(l),end(l)
,begin(r),end(r));
}
/** compiler invocation **************************/
> g++ --version
g++ (GCC) 10.1.1 20200507 (Red Hat 10.1.1-1)
> g++ -std=c++11 -O2 compare_differing_codegen.cpp -S
> cat compare_differing_codegen.s
.file "compare_differing_codegen.cpp"
.text
.p2align 4
.globl _Z5cmp_xSt5arrayIiLm64EES0_
.type _Z5cmp_xSt5arrayIiLm64EES0_, @function
_Z5cmp_xSt5arrayIiLm64EES0_:
.LFB890:
.cfi_startproc
leaq 264(%rsp), %rcx
leaq 8(%rsp), %rax
movq %rcx, %rdx
.p2align 4,,10
.p2align 3
.L4:
movl (%rdx), %esi
cmpl %esi, (%rax)
jl .L12
jg .L7
addq $4, %rax
addq $4, %rdx
cmpq %rcx, %rax
jne .L4
leaq 520(%rsp), %rax
cmpq %rax, %rdx
setne %al
ret
.p2align 4,,10
.p2align 3
.L12:
movl $1, %eax
ret
.p2align 4,,10
.p2align 3
.L7:
xorl %eax, %eax
ret
.cfi_endproc
.LFE890:
.size _Z5cmp_xSt5arrayIiLm64EES0_, .-_Z5cmp_xSt5arrayIiLm64EES0_
.p2align 4
.globl _Z5cmp_ySt5arrayIiLm64EES0_
.type _Z5cmp_ySt5arrayIiLm64EES0_, @function
_Z5cmp_ySt5arrayIiLm64EES0_:
.LFB909:
.cfi_startproc
subq $400, %rsp
.cfi_def_cfa_offset 408
movdqu 408(%rsp), %xmm0
leaq -120(%rsp), %rdx
movdqu 424(%rsp), %xmm1
leaq 136(%rsp), %rax
movdqu 440(%rsp), %xmm2
movdqu 456(%rsp), %xmm3
movdqu 472(%rsp), %xmm4
movups %xmm0, -120(%rsp)
movdqu 488(%rsp), %xmm5
movdqu 504(%rsp), %xmm6
movups %xmm1, -104(%rsp)
movdqu 520(%rsp), %xmm7
movdqu 536(%rsp), %xmm0
movups %xmm2, -88(%rsp)
movdqu 552(%rsp), %xmm1
movdqu 568(%rsp), %xmm2
movups %xmm3, -72(%rsp)
movdqu 584(%rsp), %xmm3
movups %xmm4, -56(%rsp)
movdqu 600(%rsp), %xmm4
movups %xmm5, -40(%rsp)
movdqu 616(%rsp), %xmm5
movups %xmm6, -24(%rsp)
movdqu 632(%rsp), %xmm6
movups %xmm7, -8(%rsp)
movdqu 648(%rsp), %xmm7
movups %xmm0, 8(%rsp)
movups %xmm1, 24(%rsp)
movups %xmm2, 40(%rsp)
movups %xmm3, 56(%rsp)
movups %xmm4, 72(%rsp)
movups %xmm5, 88(%rsp)
movups %xmm6, 104(%rsp)
movups %xmm7, 120(%rsp)
movdqu 664(%rsp), %xmm0
movdqu 680(%rsp), %xmm1
movdqu 696(%rsp), %xmm2
movdqu 712(%rsp), %xmm3
movdqu 728(%rsp), %xmm4
movdqu 744(%rsp), %xmm5
movups %xmm0, 136(%rsp)
movdqu 760(%rsp), %xmm6
movups %xmm1, 152(%rsp)
movdqu 776(%rsp), %xmm7
movdqu 792(%rsp), %xmm0
movups %xmm2, 168(%rsp)
movdqu 808(%rsp), %xmm1
movdqu 824(%rsp), %xmm2
movups %xmm3, 184(%rsp)
movdqu 840(%rsp), %xmm3
movups %xmm4, 200(%rsp)
movdqu 856(%rsp), %xmm4
movups %xmm5, 216(%rsp)
movdqu 872(%rsp), %xmm5
movups %xmm6, 232(%rsp)
movdqu 888(%rsp), %xmm6
movups %xmm7, 248(%rsp)
movdqu 904(%rsp), %xmm7
movups %xmm0, 264(%rsp)
movups %xmm1, 280(%rsp)
movups %xmm2, 296(%rsp)
movups %xmm3, 312(%rsp)
movups %xmm4, 328(%rsp)
movups %xmm5, 344(%rsp)
movups %xmm6, 360(%rsp)
movups %xmm7, 376(%rsp)
.p2align 4,,10
.p2align 3
.L15:
movl (%rax), %ecx
cmpl %ecx, (%rdx)
jl .L16
jg .L17
addq $4, %rax
leaq 392(%rsp), %rsi
addq $4, %rdx
cmpq %rsi, %rax
jne .L15
.L17:
xorl %eax, %eax
addq $400, %rsp
.cfi_remember_state
.cfi_def_cfa_offset 8
ret
.p2align 4,,10
.p2align 3
.L16:
.cfi_restore_state
movl $1, %eax
addq $400, %rsp
.cfi_def_cfa_offset 8
ret
.cfi_endproc
.LFE909:
.size _Z5cmp_ySt5arrayIiLm64EES0_, .-_Z5cmp_ySt5arrayIiLm64EES0_
.ident "GCC: (GNU) 10.1.1 20200507 (Red Hat 10.1.1-1)"
.section .note.GNU-stack,"",@progbits
