Cross-Compilation
To understand the concept of cross-compilation, let’s revisit the definition of a compiler. A compiler is a program that converts a program from one language to another. But the word ‘compiler’ is often used to refer to a program that translates a program to a machine language to create an executable that runs on a computing device. Commonly, a compiler is used to generate machine code for the same machine that the compiler itself is running. By the same machine, we mean the same architecture. For example, a compiler running on a linux-x64 machine compiling a C++ program and generating machine code for the same linux-x64 machine. This program can run on all linux-x64 machines, as long as a similar environment is provided.
However, there are situations where we want to generate binaries for a machine type other than the compiler it is running on. For example, if the target machine is not powerful enough. This is often the case while generating binaries for embedded devices, mobile apps, etc. A cross-compiler generates binaries that will run on a different machine (target machine) than the one where the compiler itself is running (the host machine). This is a slightly complicated process, as it requires all the dependencies of the target machine to be present on the host machine.
As an example, when compiling a simple hello-world program for a host machine, the stdio.h header file in place like /usr/include/stdio.h is used. For generating a cross-compiled hello-world program, the stdio.h will be found in a different sysroot. So, the compiler invocation may look like:
gcc --sysroot=/path/to/aarch64/sysroot -march=armv8-a hello.cAn even more convoluted setup is the , where there are two cross-compilers involved. In this setup, there are three machines A, B, and C. The cross-compiler in A (CA) will generate another cross-compiler (CB) that will run on B. CB will generate code for machine C.