Table of Contents
Yasm, according to the classifications given by [Saloman92], is a one-pass macro meta-assembler. In essence this jumble of words means that yasm reads the source file once, supports macros, and can target many different instruction sets. However, yasm also falls outside the set of definitions given by Saloman: while it is a one-pass assembler in that it only reads the source file once, it performs many in-memory passes over the source contents during assembly, and usually performs a number of out-of-order passes as well.
Nearly all large assemblers prior to yasm were forced to make multiple passes over the source code due to memory limitations. Yasm was designed with the modern system in mind, in which the amount of memory available in the system is vastly greater than the size of an entire executable, and certainly greater than the size of a single object file.
The design of yasm thus relies on the fundamental assumption that the parse of the source file is only performed once, and the entire source file contents are available in memory for use by later stages of assembly.
This book is about the internal structure of the yasm assembler: how the core libyasm library, modules, and frontend interoperate and the algorithms used in the process of turning a source file into an object file. While there will be some mention of yasm’s user interface, that is not the primary focus of this book.
As yasm started out as an x86 architecture, NASM syntax assembler, most examples in this book will use x86 instructions and the NASM assembler syntax. The concepts will generally apply to all architectures and syntaxes supported by yasm, but these provide a convenient point of reference.