Guide to Ghostscript Source Code

Conceptual overview

The Ghostscript source code is divided conceptually as follows:

PostScript interpreter

PostScript operators

z*.h and z*.c

Other interpreter code

i*.h and i*.c

PostScript code


PDF interpreter

PostScript code


Graphics library

Main library code

g*.h and g*.c


s*.h and s*.c

Device drivers

gdev*.h and gdev*.c

Platform-specific code

gp*.h and gp*.c

PostScript interpreter

gs.c is the main program for the interactive language interpreter; gserver.c is an alternative main program that is a rudimentary server. If you configure Ghostscript as a server rather than an interactive program, you will use gserver.c instead of gs.c.

Files named z*.c are Ghostscript operator files. The names of the files generally follow the section headings of the operator summary in section 6.2 (Second Edition) or 8.2 (Third Edition) of the PostScript Language Reference Manual. Each operator XXX is implemented by a procedure named zXXX, for example, zfill and zarray.

Files named i*.c, and *.h other than g*.h, are the rest of the interpreter. See the makefile for a little more information on how the files are divided functionally.

The main loop of the PostScript interpreter is the interp procedure in interp.c. When the interpreter is reading from an input file, it calls the token scanner in iscan*.c.

idebug.c contains a lot of debugger-callable routines useful for printing PostScript objects when debugging.

PDF interpreter

The PDF interpreter is written entirely in C meaning it’s faster than the old interpreter, uses less memory, is more robust and more secure because it can’t execute PostScript. Furthermore it is easier to maintain than previous versions.

Graphics library

Files beginning with gs, gx, or gz (both .c and .h), other than gs.c and gserver.c, are the Ghostscript library. Files beginning with gdev are device drivers or related code, also part of the library. Other files beginning with g are library files that don’t fall neatly into either the kernel or the driver category.

Files named s*.c and s*.h are a flexible stream package, including the Level 2 PostScript “filters” supported by Ghostscript. See stream.h, scommon.h, and strimpl.h for all the details.

Device drivers

The interface between the graphics library and device drivers is the only really well documented one in all of Ghostscript: see the documentation on drivers.

In addition to many real device and file format drivers listed in devs.mak and contrib.mak, a number of drivers are used for internal purposes. You can search lib.mak for files named gdev*.c to find almost all of them.

Drivers are divided into “printer” drivers, which support banding, and non-printer drivers, which don’t. The decision whether banding is required is made (by default on the basis of how much memory is available) in the procedure gdev_prn_alloc in gdevprn.c: it implements this decision by filling the virtual procedure table for the printer device in one of two different ways.

A good simple “printer” (bandable) driver to read is gdevmiff.c: it’s less than 100 lines, of which much is boilerplate. There are no simple non-printer drivers that actually drive devices: probably the simplest non-printer driver for reading is gdevm8.c, which implements 8-bit-deep devices that only store the bits in memory.

Platform-specific code

There are very few platform dependencies in Ghostscript. Ghostscript deals with them in three ways:

Files named *_.h substitute for the corresponding <*.h> file by adding conditionals that provide a uniform set of system interfaces on all platforms.

The file arch.h contains a set of mechanically-discovered platform properties like byte order, size of int, etc. These properties, not the names of specific platforms, are used to select between different algorithms or parameters at compile time.

Files named gp*.h define interfaces that are intended to be implemented differently on each platform, but whose specification is common to all platforms.

The platform-specific implementations of the gp*.h interfaces have names of the form gp_{platform}.c, specifically (this list may be out of date):

Platform-specific interfaces






DOS and MS Windows


DOS, Borland compilers


DOS, Watcom or Microsoft compiler


DOS and MS Windows


MS Windows NT






Unix, OS-9, and QNX


Unix and QNX




MS Windows NT

If you are going to extend Ghostscript to new machines or operating systems, check the *_.h files for ifdef on things other than DEBUG. You should probably plan to make a new makefile and a new gp_XXX.c file.


This section is only for advanced developers who need to integrate Ghostscript into a larger program at build time.



The Ghostscript makefiles are meant to be organized according to the following two principles:

  1. All the parameters that vary from platform to platform appear in the top-level makefile for a given platform. (“Platform” = OS + compiler + choice of interpreter vs. library).

  2. All the rules and definitions that can meaningfully be shared among more than 1 platform appear in a makefile that is “included” by a makefile (normally the top-level makefile) for those platforms.

Thus, for example:

  • Rules and definitions shared by all builds are in gs.mak.

  • Rules and definitions specific to the library (on all platforms) are in lib.mak. In principle this could be merged with gs.mak, but we wanted to leave open the possibility that gs.mak might be useful with hypothetical interpreter-only products.

  • Stuff specific to interpreters (on all platforms) is in int.mak.

  • Stuff specific to all Unix platforms should be in a single unix.mak file, but because make sometimes cares about the order of definitions, and because some of it is shared with DV/X, it got split between unix-aux.mak, unix-end.mak, unixhead.mak, unixinst.mak, and unixlink.mak.

For MS-DOS and MS Windows builds, there should be:

  • A makefile for all MS OS (DOS or Windows) builds, for all compilers and products.

    Perhaps a makefile for all MS-DOS builds, for all compilers and products, although since Watcom is the only such compiler we’re likely to support this may be overkill.

  • A makefile for all MS Windows builds, for all compilers and products.

  • A makefile for all Watcom builds (DOS or Windows), for all products.

  • A top-level makefile for the Watcom DOS interpreter product.

  • A top-level makefile for the Watcom Windows interpreter product.

  • A top-level makefile for the Watcom DOS library “product”.

  • A top-level makefile for the Watcom Windows library “product”.

  • A makefile for all Borland builds (DOS or Windows), for all products.

and so on.

This software is provided AS-IS with no warranty, either express or implied. This software is distributed under license and may not be copied, modified or distributed except as expressly authorized under the terms of that license. Refer to licensing information at or contact Artifex Software, Inc., 39 Mesa Street, Suite 108A, San Francisco, CA 94129, USA, for further information.Discord logo MS Word logo MS Word logo MS Word logo MS Word logo MS Word logo