simple loop program.

loop.c

#include <stdio.h>

int main(void)
{
  int total = 0;
  volatile int i;
  for( i = 0 ; i < (1000000 * 1000) ; i++ ) {
    total += i+1;
  }
  printf( "total = %d\n", total );
  return 0;
}

loop.s

; ModuleID = 'loop.c'
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-pc-linux-gnu"

@.str = private unnamed_addr constant [12 x i8] c"total = %d\0A\00", align 1

define i32 @main() nounwind uwtable {
  %1 = alloca i32, align 4
  %total = alloca i32, align 4
  %i = alloca i32, align 4
  store i32 0, i32* %1
  store i32 0, i32* %total, align 4
  store volatile i32 0, i32* %i, align 4
  br label %2

; <label>:2                                       ; preds = %10, %0
  %3 = load volatile i32* %i, align 4
  %4 = icmp slt i32 %3, 1000000000
  br i1 %4, label %5, label %13

; <label>:5                                       ; preds = %2
  %6 = load volatile i32* %i, align 4
  %7 = add nsw i32 %6, 1
  %8 = load i32* %total, align 4
  %9 = add nsw i32 %8, %7
  store i32 %9, i32* %total, align 4
  br label %10

; <label>:10                                      ; preds = %5
  %11 = load volatile i32* %i, align 4
  %12 = add nsw i32 %11, 1
  store volatile i32 %12, i32* %i, align 4
  br label %2

; <label>:13                                      ; preds = %2
  %14 = load i32* %total, align 4
  %15 = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([12 x i8]* @.str, i32 0, i32 0), i32 %14)
  ret i32 0
}

declare i32 @printf(i8*, ...)