The assembler reads in an ascii file written in a Java assembly language and automatically tracks code lengths and attribute counts. The assembler then generates a binary .class file that conforms to the Java Virtual Machine specification.
The Java assembly language provides no means of directly manipulating the constant pool. Rather, operands that take constant pool offsets as arguments instead take the contents of that offset as an argument.
For example, the new operand in the Java VM Spec takes an index into the constant pool. This index must point to a constant of type CONSTANT_Class, which in turn points to a CONSTANT_Utf8 entry in the constant pool representing the class's name. Using the assembly language, a programmer need not deal with the constant pool. The instruction might simply state new java.lang.Object.
Since the assembler automatically generates the constant pool, the programmer also doesn't need to wonder if he should use an operand like ldc_w instead of ldc. The assembler automatically outputs the proper command based on the size of the generated constant pool index.
The assembly language allows programmers to declare and use named local variables rather than simply slot numbers. Although programmers still have the freedom to use slot numbers (and may wish to, in order to decrease the number of needed slots), named local variables offer a number of benefits.
When a local variable is declared (using syntax like int MyInt in the code portion of a method), the assembler automatically allocates a slot number to that local variable. When that local variable is then used in an operation (like iload MyInt), the assembler automatically substitutes the proper slot number. The assembler also tracks the count of max_locals and enters that in the class file (unless the programmer provided a value for max_locals). Named local variables also make an assembly program much easier to read, understand, and debug.
The programmer can also use the special operands of load and store. These commands are not in the Java VM Spec, but provide for easier programming. Following the above example, if the programmer types store MyInt the assembler knows that MyInt has a type of int and will automatically generate the istore command with the proper slot number. This might be particularly helpful in an assembly language class, since it reduces the number of instructions a student needs to learn.
Since manually calculating byte offsets is disliked by most programmers, the assembly language provides for use of labels to mark a position in a piece of code. (In fact, the assembly language does not allow the programmer to manually calculate offsets.)
A label declaration should be an identifier followed immediately be a colon (with no space between the identifier and colon). So one line of code might read MyLabel: istore MyInt. Later the programmer may include an operation such as goto MyLabel. The assembler will automatically calculate the offset to MyLabel and enter that into the code. It will also change the goto operand to goto_w if the offset is large enough. Thus the programmer need not worry about calculating the offset or changing the operation based on the size of the offset.
Labels are used anywhere a code offset is needed. This include commands such as tableswitch as well as the exceptions, line number, and local variable tables which come at the end of a method.