Sunday, August 6, 2017

The structure of a GCC back end

This is part one of a series “Writing a GCC back end”.

The GCC back end is configured in gcc/ and the implementation is placed in directories machine under gcc/config and gcc/common/config where “machine” is the name of the back end (for example, i386 for the x86 architecture).

The back end places some functionality in libgcc. For example, architectures that do not have an instruction for integer division will instead generate a call to a function __divsi3 in libgcc. libgcc is configured in libgcc/ and target-specific files are located in a directory machine under libgcc/config.


config.gcc is a shell script that parses the target string (e.g. x86_64-linux-gnu) and sets variables pointing out where to find the rest of the back end and how to compile it. The variables that can be set are documented at the top of the config.gcc file.

The only variable that must be set is cpu_type that specifies machine. Most targets also set extra_objs that specifies extra object files that should be linked into the compiler, tmake_file that contains makefile fragments that compiles those extra objects (or sets makefile variables modifying the build), and tm_file that adds header files containing target-specific information.

A typical configuration for a simple target (such as ft32-unknown-elf) looks something like
tm_file="dbxelf.h elfos.h newlib-stdint.h ${tm_file}"


The main part of the back end is located in gcc/config/machine. It consists of eight different components, each implemented in a separate file:
  • machine.h is included all over the compiler and contains macros defining properties of the target, such as the size of integers and pointers, number of registers, data alignment rules, etc.
  • GCC implements a generic backend where machine.c can override most of the functionality. The backend is written in C,1 so the virtual functions are handled manually with function pointers in a structure, and machine.c overrides the defaults using code of the form
    #define TARGET_FRAME_POINTER_REQUIRED ft32_frame_pointer_required
    static bool
    ft32_frame_pointer_required (void)
      return cfun->calls_alloca;
  • machine-protos.h contains prototypes for the external functions defined in machine.c.
  • machine.opt adds target-specific command-line options to the compiler using a record format specifying the option name, properties, and a documentation string for the --help output. For example,
    Target Joined Separate UInteger Var(g_switch_value) Init(8)
    -msmall-data-limit=N    Put global and static data smaller than <number> bytes into a special section.
  • adds a command-line option -msmall-data-limit that has a default value 8, and is generated as an unsigned variable named g_switch_value.
  •,, and contain the machine description consisting of rules for instruction selection and register allocation, pipeline description, and peephole optimizations. These will be covered in detail in parts 3–7 of this series.
  • machine-modes.def defines extra machine modes for use in the low-level IR (a “machine mode” in the GCC terminology defines the size and representation of a data object. That is, it is a data type.). This is typically used for condition codes and vectors.
The GCC configuration is very flexible and everything can be overridden, so some back ends look slightly different as they, for example, add several .opt files by setting extra_options in config.gcc.


The gcc/common/config/machine directory contains a file machine-common.c that can add/remove optimization passes, change the defaults for --param values, etc.

Many back ends do not need to do anything here, and this file can be disabled by setting
in config.gcc.


The libgcc works in the same way as config.gcc, but with different variables.

The only variable that must be set is cpu_type that specifies machine. Most targets also set extra_parts that specifies extra object files to include in the library and tmake_file that contains makefile fragments that add extra functionality (such as soft-float support).

A typical configuration for a simple target looks something like
tmake_file="$tmake_file t-softfp"
extra_parts="$extra_parts crti.o crtn.o crtbegin.o crtend.o"


The libgcc/config/machine directory contains extra files that may be needed for the target architecture. Simple implementations typically only contain a crti.S and crtn.S (crtbegin/crtend and the makefile support for building all of these have default implementation) and a file sfp-machine.h containing defaults for the soft-float implementation.

1. GCC is written in C++03 these days, but the structure has not been changed since it was written in C.

No comments:

Post a Comment