Timing too good to be true

[Paul Jurczak]

On Saturday, 21 September 2013 at 22:47:16 UTC, bearophile wrote:
> Paul Jurczak:
>
>> I'm running this test program with multiple compilers on 
>> Xubuntu and I'm getting impossibly short timings with LDC 
>> compared to DMD and GDC (timing details in comments below). I 
>> would appreciate any ideas about what is going on here?
>
> Use -output-s on ldc and show us the asm of just the two 
> functions.
>
> Bye,
> bearophile

Here it is:

	.globl	_D4main5e28_0FiZi
	.align	16, 0x90
	.type	_D4main5e28_0FiZi, at function
_D4main5e28_0FiZi:
	.cfi_startproc
	movl	$1, %eax
	cmpl	$3, %edi
	jl	.LBB0_3
	movl	$2, %r8d
	movl	$20, %edx
	movl	$8, %esi
	movl	%eax, %ecx
	.align	16, 0x90
.LBB0_2:
	addl	%edx, %eax
	leal	(%rax,%rcx,4), %eax
	addl	%esi, %ecx
	addl	$8, %esi
	addl	$20, %edx
	addl	$2, %r8d
	cmpl	%edi, %r8d
	jl	.LBB0_2
.LBB0_3:
	ret
.Ltmp0:
	.size	_D4main5e28_0FiZi, .Ltmp0-_D4main5e28_0FiZi
	.cfi_endproc

	.globl	_D4main5e28_1FiZi
	.align	16, 0x90
	.type	_D4main5e28_1FiZi, at function
_D4main5e28_1FiZi:
	.cfi_startproc
	movl	$1, %eax
	cmpl	$3, %edi
	jl	.LBB1_4
	leal	-3(%rdi), %ecx
	shrl	$31, %ecx
	leal	-3(%rdi,%rcx), %ecx
	andl	$-2, %ecx
	negl	%ecx
	leal	-3(%rdi,%rcx), %ecx
	leal	1(%rdi), %edx
	subl	%ecx, %edx
	cmpl	$2, %edx
	je	.LBB1_4
	decl	%edi
	subl	%ecx, %edi
	movl	$1, %ecx
	movl	$20, %edx
	movl	$8, %esi
	movl	$1, %eax
	.align	16, 0x90
.LBB1_3:
	addl	%edx, %eax
	leal	(%rax,%rcx,4), %eax
	addl	%esi, %ecx
	addl	$8, %esi
	addl	$20, %edx
	addl	$-2, %edi
	jne	.LBB1_3
.LBB1_4:
	ret
.Ltmp1:
	.size	_D4main5e28_1FiZi, .Ltmp1-_D4main5e28_1FiZi
	.cfi_endproc