Why do these two identical codes, one in C++ and the other in Java, give different results?

Question

I’m trying to solve an issue in the URI Online Judge in Java but I’m afraid of 5% error, and what I’ve studied is being caused by the fact that the accuracy of variables in C++ and Java is different. I’ve tried everything to try to get Accept in Java but I couldn’t... I "converted" the Java code to C++ and it worked perfectly.

I’ve tried using the strictfp, I’ve tried using BigDecimal, I’ve tried to change the accuracy of BigDecimal, tried to change the variables, tried to change the methods of print, I tried to use other modifiers but I couldn’t get anything, so everything leads me to believe that the fault lies with the precision of the floating point values.

Java Code:

Scanner leitor = new Scanner(System.in);
float a = leitor.nextFloat();
float b = leitor.nextFloat();
double c = leitor.nextDouble();
double d = leitor.nextDouble(); 
System.out.printf("A = %f, B = %f\n",a,b);
System.out.printf("C = %f, D = %f\n",c,d);
System.out.printf("A = %.1f, B = %.1f\n",a,b);
System.out.printf("C = %.1f, D = %.1f\n",c,d);
System.out.printf("A = %.2f, B = %.2f\n",a,b);
System.out.printf("C = %.2f, D = %.2f\n",c,d);
System.out.printf("A = %.3f, B = %.3f\n",a,b);
System.out.printf("C = %.3f, D = %.3f\n",c,d);
System.out.printf("A = %.3E, B = %.3E\n",a,b);
System.out.printf("C = %.3E, D = %.3E\n",c,d);
System.out.printf("A = %.0f, B = %.0f\n",a,b);
System.out.printf("C = %.0f, D = %.0f\n",c,d);

C++ code:

float  a, b;
double c, d;
cin >> a;
cin >> b;
cin >> c;
cin >> d;
printf("A = %f, B = %f\n",a,b);
printf("C = %f, D = %f\n",c,d);
printf("A = %.1f, B = %.1f\n",a,b);
printf("C = %.1f, D = %.1f\n",c,d);
printf("A = %.2f, B = %.2f\n",a,b);
printf("C = %.2f, D = %.2f\n",c,d);
printf("A = %.3f, B = %.3f\n",a,b);
printf("C = %.3f, D = %.3f\n",c,d);
printf("A = %.3E, B = %.3E\n",a,b);
printf("C = %.3E, D = %.3E\n",c,d);
printf("A = %.0f, B = %.0f\n",a,b);
printf("C = %.0f, D = %.0f\n",c,d);

C++ results are 5% different from Java results. Unfortunately I do not have access to the results only have access to the website response.

Answer 1

It is not so unusual for languages to have floating point implementations differently from each other. Of course, they are not faithfully following the IEEE specification

A possible solution is to use in Java strictfp (linking to the compilation too) and may get the same result.

But anyway it doesn’t make much difference because binary floating point numbers were created to be fast and not to be exact. If you need accuracy you’re using the wrong type, you probably want to use the BigDecimal. See more in What is the correct way to use the float, double and decimal types?.

With the issue is said to have already made the suggested, but we have not seen what. It is possible that the exercise has only been tested with C++ and cannot be done in another language. We have no way of knowing.