Assembler code:

Assembler code:

     movw    r0, #:lower16:y
     movw    r0, #:upper16:y
     ldr     r0, [r0]  ???

Is this translation to ARM assembler code always correct ???

No, if the data type of the variable y was byte, we must use:

     movw    r0, #:lower16:y
     movw    r0, #:upper16:y
     ldrsb   r0, [r0]  !!!

Otherwise, we will get a runtime error !!!

And if the data type of the variable y was short, we must use:

     movw    r0, #:lower16:y
     movw    r0, #:upper16:y
     ldrsh   r0, [r0]  !!!

Otherwise, we will get a runtime error !!!

Answer:

Two-pass compilers:

Evidence: the Java compiler can compile this code:

  public class demo
  {
     public static void main(String[] args)
     {
       System.out.println(x); // How can Java compiler     
                              // know the data type of x ? 
     }                // Answer: read input file twice !

     static int x = 1234;
  }
   

DEMO:   /home/cs255001/demo/tmp/demo.java

One-pass compilers:

Notable fact about 1 pass compilers:

Evidence: the C compiler can compile this code:

  #include <stdio.h>

  int x = 1234;

  int main( )
  {
     printf("%d\n", x);
  }
   

DEMO:   /home/cs255001/demo/tmp/demo.c

Evidence: the C compiler can not compile this code:

  #include <stdio.h>

  int main( )
  {
     printf("%d\n", x);  // <-- error !
  }

  int x = 1234; 

Reason: the data type of x is not (yet) known when the C compiler processes the highlight line !!