This isn't a terribly precise answer to your question, but an aha! moment, when learning about compilers, is learning that gcc (the GNU Compiler Collection), like llvm, is an example of a three-stage compiler.
- The ‘front end’ parses the syntax of whichever language you're interested, and spits out an Abstract Syntax Tree (AST), which represents your program in a language-independent way.
- Then the ‘middle end’ (terrible name, but ‘the clever bit’) reorganises that AST into another AST which is semantically equivalent but easier to turn into machine code.
- Then the ‘back end’ turns that reorganised AST into assembler for one-or-other processor, possibly doing platform-specific micro-optimisations along the way.
That's why the (huge number of) gcc/llvm options are unexpectedly common to (apparently wildly) different languages. A few of the options are specific to C, or Fortran, or Objective-C, or whatever, but the majority of them (probably) are concerned with the middle and last bits, and so are common to all of the languages that gcc/llvm supports.
Thus the various options are specific to stage 1, 2 or 3, but may not be conveniently labelled as such; with this in mind, however, you might reasonably intuit what is and isn't relevant to the particular language you're interested in.
(It's for this sort of reason that I will dogmatically claim that CC++FortranJavaPerlPython is essentially a single language, with only trivial syntactical and library minutiae to distinguish between dialects).