1OCAMLOPT(1) General Commands Manual OCAMLOPT(1)
2
6 ocamlopt - The OCaml native-code compiler
7
10 ocamlopt [ options ] filename ...
11 ocamlopt.opt (same options)
13
16 The OCaml high-performance native-code compiler ocamlopt(1) compiles
17 OCaml source files to native code object files and link these object
18 files to produce standalone executables.
19 The ocamlopt(1) command has a command-line interface very close to that
21 of ocamlc(1). It accepts the same types of arguments and processes
22 them sequentially, after all options have been processed:
23 Arguments ending in .mli are taken to be source files for compilation
25 unit interfaces. Interfaces specify the names exported by compilation
26 units: they declare value names with their types, define public data
27 types, declare abstract data types, and so on. From the file x.mli, the
28 ocamlopt(1) compiler produces a compiled interface in the file x.cmi.
29 The interface produced is identical to that produced by the bytecode
30 compiler ocamlc(1).
31 Arguments ending in .ml are taken to be source files for compilation
33 unit implementations. Implementations provide definitions for the names
34 exported by the unit, and also contain expressions to be evaluated for
35 their side-effects. From the file x.ml, the ocamlopt(1) compiler pro‐
36 duces two files: x.o, containing native object code, and x.cmx, con‐
37 taining extra information for linking and optimization of the clients
38 of the unit. The compiled implementation should always be referred to
39 under the name x.cmx (when given a .o file, ocamlopt(1) assumes that it
40 contains code compiled from C, not from OCaml).
41 The implementation is checked against the interface file x.mli (if it
43 exists) as described in the manual for ocamlc(1).
44 Arguments ending in .cmx are taken to be compiled object code. These
46 files are linked together, along with the object files obtained by com‐
47 piling .ml arguments (if any), and the OCaml standard library, to pro‐
48 duce a native-code executable program. The order in which .cmx and .ml
49 arguments are presented on the command line is relevant: compilation
50 units are initialized in that order at run-time, and it is a link-time
51 error to use a component of a unit before having initialized it. Hence,
52 a given x.cmx file must come before all .cmx files that refer to the
53 unit x.
54 Arguments ending in .cmxa are taken to be libraries of object code.
56 Such a library packs in two files lib.cmxa and lib.a a set of object
57 files (.cmx/.o files). Libraries are build with ocamlopt -a (see the
58 description of the -a option below). The object files contained in the
59 library are linked as regular .cmx files (see above), in the order
60 specified when the library was built. The only difference is that if an
61 object file contained in a library is not referenced anywhere in the
62 program, then it is not linked in.
63 Arguments ending in .c are passed to the C compiler, which generates a
65 .o object file. This object file is linked with the program.
66 Arguments ending in .o or .a are assumed to be C object files and li‐
68 braries. They are linked with the program.
69 The output of the linking phase is a regular Unix executable file. It
71 does not need ocamlrun(1) to run.
72 ocamlopt.opt is the same compiler as ocamlopt, but compiled with itself
74 instead of with the bytecode compiler ocamlc(1). Thus, it behaves ex‐
75 actly like ocamlopt, but compiles faster. ocamlopt.opt is not avail‐
76 able in all installations of OCaml.
77
80 The following command-line options are recognized by ocamlopt(1).
81 -a Build a library (.cmxa/.a file) with the object files (.cmx/.o
83 files) given on the command line, instead of linking them into
84 an executable file. The name of the library must be set with the
85 -o option.
86 If -cclib or -ccopt options are passed on the command line,
88 these options are stored in the resulting .cmxa library. Then,
89 linking with this library automatically adds back the
90 -cclib and -ccopt options as if they had been provided on the
91 command line, unless the -noautolink option is given. Addition‐
92 ally, a substring $CAMLORIGIN inside a -ccopt options will be
93 replaced by the full path to the .cma library, excluding the
94 filename.
95 -absname
97 Show absolute filenames in error messages.
98 -annot Deprecated since OCaml 4.11. Please use -bin-annot instead.
100 -bin-annot
102 Dump detailed information about the compilation (types, bind‐
103 ings, tail-calls, etc) in binary format. The information for
104 file src.ml is put into file src.cmt. In case of a type error,
105 dump all the information inferred by the type-checker before the
106 error. The annotation files produced by -bin-annot contain more
107 information and are much more compact than the files produced by
108 -annot.
109 -c Compile only. Suppress the linking phase of the compilation.
111 Source code files are turned into compiled files, but no exe‐
112 cutable file is produced. This option is useful to compile mod‐
113 ules separately.
114 -cc ccomp
116 Use ccomp as the C linker called to build the final executable
117 and as the C compiler for compiling .c source files.
118 -cclib -llibname
120 Pass the -llibname option to the linker. This causes the given C
121 library to be linked with the program.
122 -ccopt option
124 Pass the given option to the C compiler and linker. For in‐
125 stance, -ccopt -Ldir causes the C linker to search for C li‐
126 braries in directory dir.
127 -color mode
129 Enable or disable colors in compiler messages (especially warn‐
130 ings and errors). The following modes are supported:
131 auto use heuristics to enable colors only if the output supports
133 them (an ANSI-compatible tty terminal);
134 always enable colors unconditionally;
136 never disable color output.
138 The environment variable "OCAML_COLOR" is considered if -color
140 is not provided. Its values are auto/always/never as above.
141 If -color is not provided, "OCAML_COLOR" is not set and the en‐
143 vironment variable "NO_COLOR" is set, then color output is dis‐
144 abled. Otherwise, the default setting is auto, and the current
145 heuristic checks that the "TERM" environment variable exists and
146 is not empty or "dumb", and that isatty(stderr) holds.
147 -error-style mode
150 Control the way error messages and warnings are printed. The
151 following modes are supported:
152 short only print the error and its location;
154 contextual like "short", but also display the source code snip‐
156 pet corresponding to the location of the error.
157 The default setting is contextual.
159 The environment variable "OCAML_ERROR_STYLE" is considered if
161 -error-style is not provided. Its values are short/contextual as
162 above.
163 -compact
166 Optimize the produced code for space rather than for time. This
167 results in smaller but slightly slower programs. The default is
168 to optimize for speed.
169 -config
171 Print the version number of ocamlopt(1) and a detailed summary
172 of its configuration, then exit.
173 -config-var
175 Print the value of a specific configuration variable from the
176 -config output, then exit. If the variable does not exist, the
177 exit code is non-zero.
178 -depend ocamldep-args
180 Compute dependencies, as ocamldep would do.
181 -for-pack module-path
183 Generate an object file (.cmx and .o files) that can later be
184 included as a sub-module (with the given access path) of a com‐
185 pilation unit constructed with -pack. For instance, ocam‐
186 lopt -for-pack P -c A.ml will generate a.cmx and a.o files that
187 can later be used with ocamlopt -pack -o P.cmx a.cmx.
188 -g Add debugging information while compiling and linking. This op‐
190 tion is required in order to produce stack backtraces when the
191 program terminates on an uncaught exception (see ocamlrun(1)).
192 -i Cause the compiler to print all defined names (with their in‐
194 ferred types or their definitions) when compiling an implementa‐
195 tion (.ml file). No compiled files (.cmo and .cmi files) are
196 produced. This can be useful to check the types inferred by the
197 compiler. Also, since the output follows the syntax of inter‐
198 faces, it can help in writing an explicit interface (.mli file)
199 for a file: just redirect the standard output of the compiler to
200 a .mli file, and edit that file to remove all declarations of
201 unexported names.
202 -cmi-file filename
204 Type-check the source implementation to be compiled against the
205 specified interface file (by-passes the normal lookup for .mli
206 and .cmi files).
207 -I directory
209 Add the given directory to the list of directories searched for
210 compiled interface files (.cmi), compiled object code files
211 (.cmx), and libraries (.cmxa). By default, the current directory
212 is searched first, then the standard library directory. Directo‐
213 ries added with -I are searched after the current directory, in
214 the order in which they were given on the command line, but be‐
215 fore the standard library directory. See also option -nostdlib.
216 If the given directory starts with +, it is taken relative to
218 the standard library directory. For instance, -I +compiler-libs
219 adds the subdirectory compiler-libs of the standard library to
220 the search path.
221 -impl filename
223 Compile the file filename as an implementation file, even if its
224 extension is not .ml.
225 -inline n
227 Set aggressiveness of inlining to n, where n is a positive inte‐
228 ger. Specifying -inline 0 prevents all functions from being in‐
229 lined, except those whose body is smaller than the call site.
230 Thus, inlining causes no expansion in code size. The default ag‐
231 gressiveness, -inline 1, allows slightly larger functions to be
232 inlined, resulting in a slight expansion in code size. Higher
233 values for the -inline option cause larger and larger functions
234 to become candidate for inlining, but can result in a serious
235 increase in code size.
236 -insn-sched
238 Enables the instruction scheduling pass in the compiler backend.
239 -intf filename
241 Compile the file filename as an interface file, even if its ex‐
242 tension is not .mli.
243 -intf-suffix string
245 Recognize file names ending with string as interface files (in‐
246 stead of the default .mli).
247 -keep-docs
249 Keep documentation strings in generated .cmi files.
250 -keep-locs
252 Keep locations in generated .cmi files.
253 -labels
255 Labels are not ignored in types, labels may be used in applica‐
256 tions, and labelled parameters can be given in any order. This
257 is the default.
258 -linkall
260 Force all modules contained in libraries to be linked in. If
261 this flag is not given, unreferenced modules are not linked in.
262 When building a library (-a flag), setting the -linkall flag
263 forces all subsequent links of programs involving that library
264 to link all the modules contained in the library. When compil‐
265 ing a module (option -c), setting the -linkall option ensures
266 that this module will always be linked if it is put in a library
267 and this library is linked.
268 -linscan
270 Use linear scan register allocation. Compiling with this allo‐
271 cator is faster than with the usual graph coloring allocator,
272 sometimes quite drastically so for long functions and modules.
273 On the other hand, the generated code can be a bit slower.
274 -match-context-rows
276 Set number of rows of context used during pattern matching com‐
277 pilation. Lower values cause faster compilation, but less opti‐
278 mized code. The default value is 32.
279 -no-alias-deps
281 Do not record dependencies for module aliases.
282 -no-app-funct
284 Deactivates the applicative behaviour of functors. With this op‐
285 tion, each functor application generates new types in its result
286 and applying the same functor twice to the same argument yields
287 two incompatible structures.
288 -noassert
290 Do not compile assertion checks. Note that the special form as‐
291 sert false is always compiled because it is typed specially.
292 This flag has no effect when linking already-compiled files.
293 -noautolink
295 When linking .cmxa libraries, ignore -cclib and -ccopt options
296 potentially contained in the libraries (if these options were
297 given when building the libraries). This can be useful if a li‐
298 brary contains incorrect specifications of C libraries or C op‐
299 tions; in this case, during linking, set -noautolink and pass
300 the correct C libraries and options on the command line.
301 -nodynlink
303 Allow the compiler to use some optimizations that are valid only
304 for code that is never dynlinked.
305 -no-insn-sched
307 Disables the instruction scheduling pass in the compiler back‐
308 end.
309 -nostdlib
311 Do not automatically add the standard library directory to the
312 list of directories searched for compiled interface files
313 (.cmi), compiled object code files (.cmx), and libraries
314 (.cmxa). See also option -I.
315 -nolabels
317 Ignore non-optional labels in types. Labels cannot be used in
318 applications, and parameter order becomes strict.
319 -o exec-file
321 Specify the name of the output file produced by the linker. The
322 default output name is a.out, in keeping with the Unix tradi‐
323 tion. If the -a option is given, specify the name of the library
324 produced. If the -pack option is given, specify the name of the
325 packed object file produced. If the -output-obj option is
326 given, specify the name of the output file produced. If the
327 -shared option is given, specify the name of plugin file pro‐
328 duced. This can also be used when compiling an interface or im‐
329 plementation file, without linking, in which case it sets the
330 name of the cmi or cmo file, and also sets the module name to
331 the file name up to the first dot.
332 -opaque
334 When compiling a .mli interface file, this has the same effect
335 as the -opaque option of the bytecode compiler. When compiling a
336 .ml implementation file, this produces a .cmx file without
337 cross-module optimization information, which reduces recompila‐
338 tion on module change.
339 -open module
341 Opens the given module before processing the interface or imple‐
342 mentation files. If several -open options are given, they are
343 processed in order, just as if the statements open! module1;;
344 ... open! moduleN;; were added at the top of each file.
345 -output-obj
347 Cause the linker to produce a C object file instead of an exe‐
348 cutable file. This is useful to wrap OCaml code as a C library,
349 callable from any C program. The name of the output object file
350 must be set with the -o option. This option can also be used to
351 produce a compiled shared/dynamic library (.so extension).
352 -output-complete-obj Same as -output-obj except the object file
353 produced includes the runtime and autolink libraries.
354 -pack Build an object file (.cmx and .o files) and its associated com‐
357 piled interface (.cmi) that combines the .cmx object files given
358 on the command line, making them appear as sub-modules of the
359 output .cmx file. The name of the output .cmx file must be
360 given with the -o option. For instance, ocam‐
361 lopt -pack -o P.cmx A.cmx B.cmx C.cmx generates compiled files
362 P.cmx, P.o and P.cmi describing a compilation unit having three
363 sub-modules A, B and C, corresponding to the contents of the ob‐
364 ject files A.cmx, B.cmx and C.cmx. These contents can be refer‐
365 enced as P.A, P.B and P.C in the remainder of the program.
366 The .cmx object files being combined must have been compiled
368 with the appropriate -for-pack option. In the example above,
369 A.cmx, B.cmx and C.cmx must have been compiled with ocam‐
370 lopt -for-pack P.
371 Multiple levels of packing can be achieved by combining -pack
373 with -for-pack. See The OCaml user's manual, chapter "Native-
374 code compilation" for more details.
375 -pp command
377 Cause the compiler to call the given command as a preprocessor
378 for each source file. The output of command is redirected to an
379 intermediate file, which is compiled. If there are no compila‐
380 tion errors, the intermediate file is deleted afterwards.
381 -ppx command
383 After parsing, pipe the abstract syntax tree through the pre‐
384 processor command. The module Ast_mapper(3) implements the ex‐
385 ternal interface of a preprocessor.
386 -principal
388 Check information path during type-checking, to make sure that
389 all types are derived in a principal way. All programs accepted
390 in -principal mode are also accepted in default mode with equiv‐
391 alent types, but different binary signatures.
392 -rectypes
394 Allow arbitrary recursive types during type-checking. By de‐
395 fault, only recursive types where the recursion goes through an
396 object type are supported. Note that once you have created an
397 interface using this flag, you must use it again for all depen‐
398 dencies.
399 -runtime-variant suffix
401 Add suffix to the name of the runtime library that will be used
402 by the program. If OCaml was configured with option -with-de‐
403 bug-runtime, then the d suffix is supported and gives a debug
404 version of the runtime.
405 -S Keep the assembly code produced during the compilation. The as‐
407 sembly code for the source file x.ml is saved in the file x.s.
408 -stop-after pass
410 Stop compilation after the given compilation pass. The currently
411 supported passes are: parsing, typing, scheduling, emit.
412 -save-ir-after pass
414 Save intermediate representation after the given compilation
415 pass. The currently supported passes are: scheduling.
416 -safe-string
418 Enforce the separation between types string and bytes, thereby
419 making strings read-only. This is the default.
420 -shared
422 Build a plugin (usually .cmxs) that can be dynamically loaded
423 with the Dynlink module. The name of the plugin must be set with
424 the -o option. A plugin can include a number of OCaml modules
425 and libraries, and extra native objects (.o, .a files). Build‐
426 ing native plugins is only supported for some operating system.
427 Under some systems (currently, only Linux AMD 64), all the OCaml
428 code linked in a plugin must have been compiled without the -no‐
429 dynlink flag. Some constraints might also apply to the way the
430 extra native objects have been compiled (under Linux AMD 64,
431 they must contain only position-independent code).
432 -short-paths
434 When a type is visible under several module-paths, use the
435 shortest one when printing the type's name in inferred inter‐
436 faces and error and warning messages.
437 -strict-sequence
439 The left-hand part of a sequence must have type unit.
440 -unboxed-types
442 When a type is unboxable (i.e. a record with a single argument
443 or a concrete datatype with a single constructor of one argu‐
444 ment) it will be unboxed unless annotated with [@@ocaml.boxed].
445 -no-unboxed-types
447 When a type is unboxable it will be boxed unless annotated with
448 [@@ocaml.unboxed]. This is the default.
449 -unsafe
451 Turn bound checking off for array and string accesses (the
452 v.(i)ands.[i] constructs). Programs compiled with -unsafe are
453 therefore faster, but unsafe: anything can happen if the program
454 accesses an array or string outside of its bounds. Additionally,
455 turn off the check for zero divisor in integer division and mod‐
456 ulus operations. With -unsafe, an integer division (or modulus)
457 by zero can halt the program or continue with an unspecified re‐
458 sult instead of raising a Division_by_zero exception.
459 -unsafe-string
461 Identify the types string and bytes, thereby making strings
462 writable. This is intended for compatibility with old source
463 code and should not be used with new software.
464 -v Print the version number of the compiler and the location of the
466 standard library directory, then exit.
467 -verbose
469 Print all external commands before they are executed, in partic‐
470 ular invocations of the assembler, C compiler, and linker.
471 -version or -vnum
473 Print the version number of the compiler in short form (e.g.
474 "3.11.0"), then exit.
475 -w warning-list
477 Enable, disable, or mark as fatal the warnings specified by the
478 argument warning-list. See ocamlc(1) for the syntax of warning-
479 list.
480 -warn-error warning-list
482 Mark as fatal the warnings specified in the argument warn‐
483 ing-list. The compiler will stop with an error when one of
484 these warnings is emitted. The warning-list has the same mean‐
485 ing as for the -w option: a + sign (or an uppercase letter)
486 marks the corresponding warnings as fatal, a - sign (or a lower‐
487 case letter) turns them back into non-fatal warnings, and a @
488 sign both enables and marks as fatal the corresponding warnings.
489 Note: it is not recommended to use the -warn-error option in
491 production code, because it will almost certainly prevent com‐
492 piling your program with later versions of OCaml when they add
493 new warnings or modify existing warnings.
494 The default setting is -warn-error -a+31 (only warning 31 is fa‐
496 tal).
497 -warn-help
499 Show the description of all available warning numbers.
500 -where Print the location of the standard library, then exit.
502 -with-runtime
504 Include the runtime system in the generated program. This is the
505 default.
506 -without-runtime
508 The compiler does not include the runtime system (nor a refer‐
509 ence to it) in the generated program; it must be supplied sepa‐
510 rately.
511 - file Process file as a file name, even if it starts with a dash (-)
513 character.
514 -help or --help
516 Display a short usage summary and exit.
517
520 When the Flambda code generator has been enabled at configuration time,
521 its behavior may be tuned up with the following additional options:
522 -02 Perform more optimisation than usual. Compilation times may be
524 lengthened.
525 -03 Perform even more optimisation than usual, possibly including
527 unrolling of recursive functions. Compilation times may be sig‐
528 nificantly lengthened.
529 -Oclassic
531 Makes inlining decisions at the point of definition of a func‐
532 tion rather than at the call site(s). This mirrors the behaviour
533 of OCaml compilers not using Flambda. Compared to compilation
534 using the new Flambda inlining heuristics (for example at -O2)
535 it produces smaller .cmx files, shorter compilation times and
536 code that probably runs rather slower.
537 -inlining-report
539 Emit .inlining files (one per round of optimisation) showing all
540 of the inliner's decisions.
541
544 The IA32 code generator (Intel Pentium, AMD Athlon) supports the fol‐
545 lowing additional option:
546 -ffast-math
548 Use the IA32 instructions to compute trigonometric and exponen‐
549 tial functions, instead of calling the corresponding library
550 routines. The functions affected are: atan, atan2, cos, log,
551 log10, sin, sqrt and tan. The resulting code runs faster, but
552 the range of supported arguments and the precision of the result
553 can be reduced. In particular, trigonometric operations cos,
554 sin, tan have their range reduced to [-2^64, 2^64].
555
558 The AMD64 code generator (64-bit versions of Intel Pentium and AMD
559 Athlon) supports the following additional options:
560 -fPIC Generate position-independent machine code. This is the de‐
562 fault.
563 -fno-PIC
565 Generate position-dependent machine code.
566
569 The PowerPC code generator supports the following additional options:
570 -flarge-toc
572 Enables the PowerPC large model allowing the TOC (table of con‐
573 tents) to be arbitrarily large. This is the default since 4.11.
574 -fsmall-toc
576 Enables the PowerPC small model allowing the TOC to be up to 64
577 kbytes per compilation unit. Prior to 4.11 this was the default
578 behaviour. \nd{options}
579
582 The ARM code generator supports the following additional options:
583 -farch=armv4|armv5|armv5te|armv6|armv6t2|armv7
585 Select the ARM target architecture
586 -ffpu=soft|vfpv2|vfpv3-d16|vfpv3
588 Select the floating-point hardware
589 -fPIC Generate position-independent machine code.
591 -fno-PIC
593 Generate position-dependent machine code. This is the default.
594 -fthumb
596 Enable Thumb/Thumb-2 code generation
597 -fno-thumb
599 Disable Thumb/Thumb-2 code generation
600 The default values for target architecture, floating-point hardware and
602 thumb usage were selected at configure-time when building ocamlopt it‐
603 self. This configuration can be inspected using ocamlopt -config. Tar‐
604 get architecture depends on the "model" setting, while floating-point
605 hardware and thumb support are determined from the ABI setting in "sys‐
606 tem" ( linux_eabiorlinux_eabihf).
607
610 ocamlc(1).
611 The OCaml user's manual, chapter "Native-code compilation".
612 OCAMLOPT(1)