Internals of GCC

Basically, this podcast talks about the internal implementation of GCC as a compiler that is able to compile from languages such as java, c, c++ and how it does it. Morgan Deters, who is a compiler programmer with a PHD says that it is flexible because of its modularity. The front end, middle and back end are modules that work together but are architecture and language independent, so it can translate from one language to another (assembler) using the steps of compilers design that we’ve seen in classes without being specific about the language.

  Morgan says that the compiler can crash in runtime if we do not make the programs properly because even though it has lexical and syntax analysis to prevent that, it’s impossible to do it perfectly.

  He says that the importance of compilers programming is about performance because there are many ways to translate from one language to another but the way we do it defines if its performance improves or goes down. He talked so much about trees that define the functions, operations, variables and other elements of the program that let the compiler make decisions to optimize the processes. It all depends on the levels of representations that the language has.

  I didn’t understand at all when Morgan talked about optimization using registers, he says that the compilers assumes that he has many of them and uses them as much as he can to optimize processes. But I believe that there should be some kind of control to use those registers and memory.

  He also talked about something called the RTL that has a way to translate some instructions directly to assembly. It matches the instructions in a language such as a sum with the defined instructions in machine code so the gcc is able to translate it easily, more optimized.