How does a compiler convert high to low level language?

Answer 1

It depends on the compiler and operating system. A lot of compilers generate C(++) objects or code instead of assembly language. It seems to be easier for compiler developers to develop and support a C(++) platform instead of assembler.

Answer 2

First, it is important to note that a compiler might, but does not need to, generate output in a low level language. Many traditional compilers generate assembly language output and meet the question's assumption, but more modern designs often skip this intermediate step.

The general process of compiling a high level language into the desired output is a multi-step process. The essence of this process is outlined in the following.

In one step, the compiler reads and pre-conditions the source code. For example, extraneous spaces or source code comments might be removed at this stage.

In another step, the compiler scans the remaining source code and constructs tokens from groups of input characters, and categorizes each token. For example, the letters a, b, and c might combine to a token of value "abc" and type variable reference, and characters 1, 2 and 3 might combine to a token of value 123 and type integer constant.

In the next step, the steam of tokens is subject to syntactic analysis. For example, three tokens "abc," "=" and "123" can form an expression when used in this order, but placing the constant number on the left hand side does not result in a valid expression. Syntactic analysis is generally driven by rules defined in the grammar, and implemented with a parser. The parser generally results in a tree of semantic nodes. For example, the expression in the previous example might be folded into a semantic node of type assignment, consisting of a variable reference and an arithmetic expression.

In the next step, semantic analysis is performed. The purpose is to determine whether each construct makes sense in its context. For example, one language may not allow assigning a variable of type "string" to a variable of type "int."

Now follows a series of optimizing steps, aimed at reducing the runtime-complexity of the semantic tree.

Finally, the semantic tree is translated into the destination code. Code generation is typically implemented using multiple steps, and can be closely interwoven with optimization.