SBCL -- Steel Bank Common Lisp

SBCL is a free Common Lisp programming environment. It is derived from the free CMU CL programming environment. (The name is intended to acknowledge the connection: steel and banking are the industries where Carnegie and Mellon made the big bucks.)

It is free software, mostly in the public domain, but with some subsystems under BSD-style licenses which allow modification and reuse as long as credit is given. It is provided "as is", with no warranty of any kind.

For more information about license issues, see the COPYING file in the distribution. For more information about history, see the CREDITS file in the distribution.

To run SBCL, type "sbcl" at the command line with no arguments. (SBCL understands command line arguments, but you probably won't need to use them unless you're a fairly advanced user. If you are, you should read the COMMAND LINE SYNTAX section, below.) You should see some startup messages, then a prompt ("CR*"). Type a Lisp expression at the prompt, and SBCL will read it, execute it, print any values returned, give you another prompt, and wait for your next input. For example, CR * (+ 1 2 3)

6 * (funcall (lambda (x y) (list x y y)) :toy :choo)

(:TOY :CHOO :CHOO) * "Hello World"

"Hello World" *

Many people like to run SBCL, like other Lisp systems, as a subprocess under Emacs. The Emacs "Slime" and "ilisp" modes provide many convenient features, like command line editing, tab completion, and various kinds of coupling between Common Lisp source files and the interactive SBCL subprocess, but they can be somewhat fragile wrt. packages and readtables, in which case SBCL in the Emacs "shell" mode can a useful substitute.

SBCL compiles Common Lisp to native code. (Even today, some 30 years after the MacLisp compiler, people will tell you that Lisp is an interpreted language. Ignore them.)

SBCL aims for but has not completely achieved compliance with the ANSI standard for Common Lisp. More information about this is available in the BUGS section below.

SBCL also includes various non-ANSI extensions, described more fully in the User Manual. Some of these are in the base system and others are "contrib" modules loaded on request using CRREQUIRE. For example, to load the CRSB-BSD-SOCKETS module that providces TCP/IP connectivity, CR * (require asdf) * (require sb-bsd-sockets)

Many Lispy extensions have been retained from CMU CL:

--

CMU-CL-style safe implementation of type declarations: "Declarations are assertions."

--

the source level debugger (very similar to CMU CL's)

--

the profiler (now somewhat different from CMU CL's)

--

saving the state of the running SBCL process, producing a "core" file which can be restarted later

--

Gray streams (a de-facto standard system of overloadable CLOS classes whose instances can be used wherever ordinary ANSI streams can be used)

--

weak pointers and finalization (which have unfortunately suffered from at least some code rot, so that e.g. weak hash tables don't work)

Fundamental system interface extensions are also provided:

--

calling out to C code (a.k.a. FFI, foreign function interface, with very nearly the same interface as CMU CL)

--

some simple support for operations with a "scripting language" flavor, e.g. reading POSIX CRargc and CRargv, or executing a subprogram

SBCL can be built from scratch using a plain vanilla ANSI Common Lisp system and a C compiler, and all of its properties are specified by the version of the source code that it was created from. This clean bootstrappability was the immediate motivation for forking off of the CMU CL development tree. A variety of implementation differences are motivated by this design goal.

Maintenance work in SBCL since the fork has diverged somewhat from the maintenance work in CMU CL. Many but not all bug fixes and improvements have been shared between the two projects, and sometimes the two projects disagree about what would be an improvement.

Most extensions supported by CMU CL have been unbundled from SBCL, including Motif support, the Hemlock editor, search paths, the low-level Unix interface, the WIRE protocol, various user-level macros and functions (e.g. CRLETF, CRITERATE, CRMEMQ, CRREQUIRED-ARGUMENT), and many others.

SBCL inplements multithreading, but in a completely different fashion from CMU CL: see the User Manual for details. As of 0.8.5 this is considered beta-quality and must be explicitly enabled at build time.

SBCL has retained some extensions from its parent CMU CL. Many of the retained extensions are in these categories:

--

things which might be in the new ANSI spec, e.g. safe type declarations, weak pointers, finalization, foreign function interface to C, and Gray streams;

--

things which are universally available in Unix scripting languages, e.g. CRRUN-PROGRAM and POSIX CRargv and CRgetenv;

--

hooks into the low level workings of the system which can be useful for debugging, e.g. requesting that a particular function be executed whenever GC occurs, or tuning compiler diagnostic output;

--

unportable performance hacks, e.g. CRFREEZE-TYPE and CRPURIFY. For more information about these, look at the online documentation for symbols in the CRSB-EXT package, and look at the user manual.

There are also a few retained extensions which don't fall into any particular category, e.g. the ability to save running Lisp images as executable files.

Some of the retained extensions have new names and/or different options than their CMU CL counterparts. For example, the SBCL function which saves a Lisp image to disk and kills the running process is called CRSAVE-LISP-AND-DIE instead of CRSAVE-LISP, and SBCL's CRSAVE-LISP-AND-DIE supports fewer keyword options than CMU CL's CRSAVE-LISP does.

(Why doesn't SBCL support more extensions natively? Why drop all those nice extensions from CMU CL when the code already exists? This is a frequently asked question on the mailing list. There are two principal reasons. First, it's a design philosophy issue: arguably SBCL has done its job by supplying a stable FFI, and the right design decision is to move functionality derived from that, like socket support, into separate libraries. Some of these are distributed with SBCL as "contrib" modules, others are distributed as separate software packages by separate maintainers. Second, it's a practical decision - focusing on a smaller number of things will, we hope, let us do a better job on them.)

SBCL inherits from CMU CL the "Python" native code compiler. (Though we often avoid that name in order to avoid confusion with the scripting language also called Python.) This compiler is very clever about understanding the type system of Common Lisp and using it to optimize code, and about producing notes to let the user know when the compiler doesn't have enough type information to produce efficient code. It also tries (almost always successfully) to follow the unusual but very useful principle that "declarations are assertions", i.e. type declarations should be checked at runtime unless the user explicitly tells the system that speed is more important than safety.

The compiler reportedly produces pretty good code for modern CPU architectures which have lots of registers, but its code for the X86 is marred by many extra loads and stores to stack-based temporary variables. Because of this, and because of the extra levels of indirection in Common Lisp relative to C, the performance of SBCL isn't going to impress people who are impressed by small constant factors. However, even on the X86 it tends to be faster than byte interpreted languages (and can be a lot faster).

The compiled code uses garbage collection to automatically manage memory. The garbage collector implementation varies considerably from CPU to CPU. In particular, on some CPUs the GC is nearly exact, while on others it's more conservative, and on some CPUs the GC is generational, while on others simpler stop and copy strategies are used.

For more information about the compiler, see the user manual.

Command line syntax can be considered an advanced topic; for ordinary interactive use, no command line arguments should be necessary.

In order to understand the command line argument syntax for SBCL, it is helpful to understand that the SBCL system is implemented as two components, a low-level runtime environment written in C and a higher-level system written in Common Lisp itself. Some command line arguments are processed during the initialization of the low-level runtime environment, some command line arguments are processed during the initialization of the Common Lisp system, and any remaining command line arguments are passed on to user code.

The full, unambiguous syntax for invoking SBCL at the command line is

sbcl [runtime options] --end-runtime-options [toplevel options] --end-toplevel-options [user options]

For convenience, the --end-runtime-options and --end-toplevel-options elements can be omitted. Omitting these elements can be convenient when you are running the program interactively, and you can see that no ambiguities are possible with the option values you are using. Omitting these elements is probably a bad idea for any batch file where any of the options are under user control, since it makes it impossible for SBCL to detect erroneous command line input, so that erroneous command line arguments will be passed on to the user program even if they was intended for the runtime system or the Lisp system.

Supported runtime options are

--core <corefilename>

Run the specified Lisp core file instead of the default. (See the FILES section for the standard core, or the system documentation for CRSB-INT:SAVE-LISP-AND-DIE for information about how to create a custom core.) Note that if the Lisp core file is a user-created core file, it may run a nonstandard toplevel which does not recognize the standard toplevel options.

--noinform

Suppress the printing of any banner or other informational message at startup. (This makes it easier to write Lisp programs which work cleanly in Unix pipelines. See also the "--noprint" and "--disable-debugger" options.)

--help

Print some basic information about SBCL, then exit.

--version

Print SBCL's version information, then exit.

In the future, runtime options may be added to control behavior such as lazy allocation of memory.

Runtime options, including any --end-runtime-options option, are stripped out of the command line before the Lisp toplevel logic gets a chance to see it.

The toplevel options supported by the standard SBCL core are

--sysinit <filename>

Load filename instead of the default system-wide initialization file. (See the FILES section.) There is no special option to cause no system-wide initialization file to be read, but on a Unix system "--sysinit /dev/null" can be used to achieve the same effect.

--userinit <filename>

Load filename instead of the default user initialization file. (See the FILES section.) There is no special option to cause no user initialization file to be read, but on a Unix system "--userinit /dev/null" can be used to achieve the same effect.

--eval <command>

After executing any initialization file, but before starting the read-eval-print loop on standard input, read and evaluate the command given. More than one --eval option can be used, and all will be read and executed, in the order they appear on the command line.

--load <filename>

This is equivalent to --eval (load "<filename>"). The special syntax is intended to reduce quoting headaches when invoking SBCL from shell scripts.

--noprint

When ordinarily the toplevel "read-eval-print loop" would be executed, execute a "read-eval loop" instead, i.e. don't print a prompt and don't echo results. Combined with the --noinform runtime option, this makes it easier to write Lisp "scripts" which work cleanly in Unix pipelines.

--disable-debugger

This is equivalent to --eval (sb-ext:disable-debugger). By default, a Common Lisp system tries to ask the programmer for help when it gets in trouble (by printing a debug prompt, then listening, on CR*DEBUG-IO*). However, this is not useful behavior for a system running with no programmer available, and this option tries to set up more appropriate behavior for that situation. This is implemented by redefining CRINVOKE-DEBUGGER so that any call exits the process with a failure code after printing a backtrace. (Note that because it is implemented by modifying special variables and CRFDEFINITIONs, its effects persist in .core files created by CRSB-EXT:SAVE-LISP-AND-DIE. If you want to undo its effects, e.g. if you build a system unattended and then want to operate a derived system interactively, see the CRSB-EXT:ENABLE-DEBUGGER command.)

Regardless of the order in which --sysinit, --userinit, and --eval options appear on the command line, the sysinit file, if it exists, is loaded first; then the userinit file, if it exists, is loaded; then any --eval commands are read and executed in sequence; then the read-eval-print loop is started on standard input. At any step, error conditions or commands such as CRSB-EXT:QUIT can cause execution to be terminated before proceeding to subsequent steps.

Note that when running SBCL with the --core option, using a core file created by a user call to the CRSB-EXT:SAVE-LISP-AND-DIE, the toplevel options may be under the control of user code passed as arguments to CRSB-EXT:SAVE-LISP-AND-DIE. For this purpose, the --end-toplevel-options option itself can be considered a toplevel option, i.e. the user core, at its option, may not support it.

In the standard SBCL startup sequence (i.e. with no user core involved) toplevel options and any --end-toplevel-options option are stripped out of the command line argument list before user code gets a chance to see it.

SBCL currently runs on X86 (Linux, FreeBSD, OpenBSD, and NetBSD), X86-64 (Linux), Alpha (Linux, Tru64), PPC (Linux, Darwin/MacOS X), SPARC (Linux and Solaris 2.x), and MIPS (Linux). For information on other ongoing and possible ports, see the sbcl-devel mailing list, and/or the web site.

SBCL requires on the order of 16Mb RAM to run on X86 systems, though all but the smallest programs would be happier with 32Mb or more.

This section attempts to list the most serious and long-standing bugs. For more detailed and current information on bugs, see the BUGS file in the distribution.

It is possible to get in deep trouble by exhausting heap memory. The SBCL system overcommits memory at startup, so, on typical Unix-alikes like Linux and FreeBSD, this means that if the SBCL system turns out to use more virtual memory than the system has available for it, other processes tend to be killed randomly (!).

The compiler's handling of function return values unnecessarily violates the "declarations are assertions" principle that it otherwise adheres to. Using CRPROCLAIM or CRDECLAIM to specify the return type of a function causes the compiler to believe you without checking. Thus compiling a file containing CR (DECLAIM (FTYPE (FUNCTION (T) NULL) SOMETIMES)) (DEFUN SOMETIMES (X) (ODDP X)) (DEFUN FOO (X) (IF (SOMETIMES X) THIS-TIME NOT-THIS-TIME))

then running CR(FOO 1) gives CRNOT-THIS-TIME, because the compiler relied on the truth of the CRDECLAIM without checking it.

Some things are implemented very inefficiently.

--

Multidimensional arrays are inefficient, especially multidimensional arrays of floating point numbers.

--

CLOS isn't particularly efficient. (In part, CLOS is so dynamic that it's slow for fundamental reasons, but beyond that, the SBCL implementation of CLOS doesn't do some important known optimizations.)

--

SBCL, like most (maybe all?) implementations of Common Lisp on stock hardware, has trouble passing floating point numbers around efficiently, because a floating point number, plus a few extra bits to identify its type, is larger than a machine word. (Thus, they get "boxed" in heap-allocated storage, causing GC overhead.) Within a single compilation unit, or when doing built-in operations like CRSQRT and CRAREF, or some special operations like structure slot accesses, this is avoidable: see the user manual for some efficiency hints. But for general function calls across the boundaries of compilation units, passing the result of a floating point calculation as a function argument (or returning a floating point result as a function value) is a fundamentally slow operation.

To report a bug, please send mail to the mailing lists sbcl-help or sbcl-devel. You can find the complete mailing list addresses on the web pages at <CRhttp://sbcl.sourceforge.net/>. (You may also find fancy SourceForge bug-tracking machinery there, but don't be fooled. As of 2002-07-25 anyway, we don't actively monitor that machinery, and it exists only because we haven't been able to figure out how to turn it off.)

As with any software bug report, it's most helpful if you can provide enough information to reproduce the symptoms reliably, and if you say clearly what the symptoms are. For example, "There seems to be something wrong with TAN of very small negative arguments. When I execute CR(TAN LEAST-NEGATIVE-SINGLE-FLOAT) interactively on sbcl-1.2.3 on my Linux 4.5 X86 box, I get an CRUNBOUND-VARIABLE error."

Various information about SBCL is available at <CRhttp://www.sbcl.org/>. The mailing lists there are the recommended place to look for support.

SBCL_HOME

This variable controls where files like "sbclrc", "sbcl.core", and the add-on "contrib" systems are searched for. If it is not set, then sbcl sets it from a compile-time default location which is usually /usr/local/lib/sbcl/ but may have been changed e.g. by a third-party packager.

sbcl

executable program containing some low-level runtime support and a loader, used to read sbcl.core

sbcl.core

dumped memory image containing most of SBCL, to be loaded by the `sbcl' executable. Looked for in $CRSBCL_HOME, unless overridden by the CR--core option.

sbclrc

optional system-wide startup script, looked for in $CRSBCL_HOME/sbclrc then /etc/sbclrc, unless overridden by the CR--sysinit command line option.

.sbclrc

optional per-user customizable startup script (in user's home directory, or as specified by CR--userinit)

Dozens of people have made substantial contributions to SBCL and its subsystems, and to the CMU CL system on which it was based, over the years. See the CREDITS file in the distribution for more information.

Full SBCL documentation is maintained as a Texinfo manual. If is has been installed, the command

info sbcl

should give you access to the complete manual. Depending on your installation it may also be available in HTML and PDF formats in eg.

/usr/local/share/doc/sbcl/

See the SBCL homepage

<CRhttp://www.sbcl.org/>

for more information, including directions on how to subscribe to the sbcl-devel and sbcl-help mailing-lists.