https://computers-are-fast.github.io/ cache example w/ asm

assembly.diff

1c1
< 	.file	"c1.c"
---
> 	.file	"c2.c"
16d15
< 	pushl	%edi
28a28
> 	xorl	%ecx, %ecx
31,32d30
< 	movl	%eax, %ecx
< 	leal	(%eax,%ebx), %edi
34c32
< 	jle	L5
---
> 	jle	L4
43,44c41,42
< 	movb	%dl, -1(%ecx)
< 	cmpl	%edi, %ecx
---
> 	cmpl	%ebx, %ecx
> 	movb	%dl, (%esi,%edx)
46c44
< L5:
---
> L4:
58c56
< 	leal	-12(%ebp), %esp
---
> 	leal	-8(%ebp), %esp
62d59
< 	popl	%edi

c1.c

#include <stdlib.h>
#include <stdio.h>

// Number to guess: How big of an array (in bytes)
// can we allocate and fill in a second?

// this is intentionally more complicated than it needs to be
// so that it matches the out-of-order version :)

int main(int argc, char **argv) {
    int NUMBER, i;
    NUMBER = atoi(argv[1]);

    char* array = malloc(NUMBER);
    int j = 1;
    for (i = 0; i < NUMBER; ++i) {
        j = j * 2;
        if (j > NUMBER) {
            j = j - NUMBER;
        }
        array[i] = j;
    }

    printf("%d", array[NUMBER / 7]);
    // so that -O2 doesn't optimize out the loop

    return 0;
}

c1.s

	.file	"c1.c"
	.def	___main;	.scl	2;	.type	32;	.endef
	.section .rdata,"dr"
LC0:
	.ascii "%d\0"
	.section	.text.unlikely,"x"
LCOLDB1:
	.section	.text.startup,"x"
LHOTB1:
	.p2align 4,,15
	.globl	_main
	.def	_main;	.scl	2;	.type	32;	.endef
_main:
	pushl	%ebp
	movl	%esp, %ebp
	pushl	%edi
	pushl	%esi
	pushl	%ebx
	andl	$-16, %esp
	subl	$16, %esp
	call	___main
	movl	12(%ebp), %eax
	movl	4(%eax), %eax
	movl	%eax, (%esp)
	call	_atoi
	movl	%eax, %ebx
	movl	%eax, (%esp)
	call	_malloc
	testl	%ebx, %ebx
	movl	%eax, %esi
	movl	%eax, %ecx
	leal	(%eax,%ebx), %edi
	movl	$1, %edx
	jle	L5
	.p2align 4,,10
L6:
	addl	%edx, %edx
	movl	%edx, %eax
	subl	%ebx, %eax
	cmpl	%edx, %ebx
	cmovl	%eax, %edx
	addl	$1, %ecx
	movb	%dl, -1(%ecx)
	cmpl	%edi, %ecx
	jne	L6
L5:
	movl	%ebx, %eax
	movl	$-1840700269, %edx
	movl	$LC0, (%esp)
	imull	%edx
	addl	%ebx, %edx
	sarl	$31, %ebx
	sarl	$2, %edx
	subl	%ebx, %edx
	movsbl	(%esi,%edx), %eax
	movl	%eax, 4(%esp)
	call	_printf
	leal	-12(%ebp), %esp
	xorl	%eax, %eax
	popl	%ebx
	popl	%esi
	popl	%edi
	popl	%ebp
	ret
	.section	.text.unlikely,"x"
LCOLDE1:
	.section	.text.startup,"x"
LHOTE1:
	.ident	"GCC: (i686-posix-sjlj-rev2, Built by MinGW-W64 project) 4.9.2"
	.def	_atoi;	.scl	2;	.type	32;	.endef
	.def	_malloc;	.scl	2;	.type	32;	.endef
	.def	_printf;	.scl	2;	.type	32;	.endef

c2.c

#include <stdlib.h>
#include <stdio.h>

// Number to guess: How big of an array (in bytes)
// can we allocate and fill with 5s in a second?
// The catch: We do it out of order instead of in order.
int main(int argc, char **argv) {
    int NUMBER, i;
    NUMBER = atoi(argv[1]);

    char* array = malloc(NUMBER);
    int j = 1;
    for (i = 0; i < NUMBER; ++i) {
        j = j * 2;
        if (j > NUMBER) {
            j = j - NUMBER;
        }
        array[j] = j;
    }

    printf("%d", array[NUMBER / 7]);
    // so that -O2 doesn't optimize out the loop

    return 0;
}

c2.s

	.file	"c2.c"
	.def	___main;	.scl	2;	.type	32;	.endef
	.section .rdata,"dr"
LC0:
	.ascii "%d\0"
	.section	.text.unlikely,"x"
LCOLDB1:
	.section	.text.startup,"x"
LHOTB1:
	.p2align 4,,15
	.globl	_main
	.def	_main;	.scl	2;	.type	32;	.endef
_main:
	pushl	%ebp
	movl	%esp, %ebp
	pushl	%esi
	pushl	%ebx
	andl	$-16, %esp
	subl	$16, %esp
	call	___main
	movl	12(%ebp), %eax
	movl	4(%eax), %eax
	movl	%eax, (%esp)
	call	_atoi
	movl	%eax, %ebx
	movl	%eax, (%esp)
	call	_malloc
	xorl	%ecx, %ecx
	testl	%ebx, %ebx
	movl	%eax, %esi
	movl	$1, %edx
	jle	L4
	.p2align 4,,10
L6:
	addl	%edx, %edx
	movl	%edx, %eax
	subl	%ebx, %eax
	cmpl	%edx, %ebx
	cmovl	%eax, %edx
	addl	$1, %ecx
	cmpl	%ebx, %ecx
	movb	%dl, (%esi,%edx)
	jne	L6
L4:
	movl	%ebx, %eax
	movl	$-1840700269, %edx
	movl	$LC0, (%esp)
	imull	%edx
	addl	%ebx, %edx
	sarl	$31, %ebx
	sarl	$2, %edx
	subl	%ebx, %edx
	movsbl	(%esi,%edx), %eax
	movl	%eax, 4(%esp)
	call	_printf
	leal	-8(%ebp), %esp
	xorl	%eax, %eax
	popl	%ebx
	popl	%esi
	popl	%ebp
	ret
	.section	.text.unlikely,"x"
LCOLDE1:
	.section	.text.startup,"x"
LHOTE1:
	.ident	"GCC: (i686-posix-sjlj-rev2, Built by MinGW-W64 project) 4.9.2"
	.def	_atoi;	.scl	2;	.type	32;	.endef
	.def	_malloc;	.scl	2;	.type	32;	.endef
	.def	_printf;	.scl	2;	.type	32;	.endef