1llc - LLVM static compiler
2==========================
3
4SYNOPSIS
5--------
6
7:program:`llc` [*options*] [*filename*]
8
9DESCRIPTION
10-----------
11
12The :program:`llc` command compiles LLVM source inputs into assembly language
13for a specified architecture.  The assembly language output can then be passed
14through a native assembler and linker to generate a native executable.
15
16The choice of architecture for the output assembly code is automatically
17determined from the input file, unless the :option:`-march` option is used to
18override the default.
19
20OPTIONS
21-------
22
23If ``filename`` is "``-``" or omitted, :program:`llc` reads from standard input.
24Otherwise, it will from ``filename``.  Inputs can be in either the LLVM assembly
25language format (``.ll``) or the LLVM bitcode format (``.bc``).
26
27If the :option:`-o` option is omitted, then :program:`llc` will send its output
28to standard output if the input is from standard input.  If the :option:`-o`
29option specifies "``-``", then the output will also be sent to standard output.
30
31If no :option:`-o` option is specified and an input file other than "``-``" is
32specified, then :program:`llc` creates the output filename by taking the input
33filename, removing any existing ``.bc`` extension, and adding a ``.s`` suffix.
34
35Other :program:`llc` options are described below.
36
37End-user Options
38~~~~~~~~~~~~~~~~
39
40.. option:: -help
41
42 Print a summary of command line options.
43
44.. option:: -O=uint
45
46 Generate code at different optimization levels.  These correspond to the
47 ``-O0``, ``-O1``, ``-O2``, and ``-O3`` optimization levels used by
48 :program:`clang`.
49
50.. option:: -mtriple=<target triple>
51
52 Override the target triple specified in the input file with the specified
53 string.
54
55.. option:: -march=<arch>
56
57 Specify the architecture for which to generate assembly, overriding the target
58 encoded in the input file.  See the output of ``llc -help`` for a list of
59 valid architectures.  By default this is inferred from the target triple or
60 autodetected to the current architecture.
61
62.. option:: -mcpu=<cpuname>
63
64 Specify a specific chip in the current architecture to generate code for.
65 By default this is inferred from the target triple and autodetected to
66 the current architecture.  For a list of available CPUs, use:
67
68 .. code-block:: none
69
70   llvm-as < /dev/null | llc -march=xyz -mcpu=help
71
72.. option:: -filetype=<output file type>
73
74 Specify what kind of output ``llc`` should generated.  Options are: ``asm``
75 for textual assembly ( ``'.s'``), ``obj`` for native object files (``'.o'``)
76 and ``null`` for not emitting anything (for performance testing).
77
78 Note that not all targets support all options.
79
80.. option:: -mattr=a1,+a2,-a3,...
81
82 Override or control specific attributes of the target, such as whether SIMD
83 operations are enabled or not.  The default set of attributes is set by the
84 current CPU.  For a list of available attributes, use:
85
86 .. code-block:: none
87
88   llvm-as < /dev/null | llc -march=xyz -mattr=help
89
90.. option:: --disable-fp-elim
91
92 Disable frame pointer elimination optimization.
93
94.. option:: --disable-excess-fp-precision
95
96 Disable optimizations that may produce excess precision for floating point.
97 Note that this option can dramatically slow down code on some systems
98 (e.g. X86).
99
100.. option:: --enable-no-infs-fp-math
101
102 Enable optimizations that assume no Inf values.
103
104.. option:: --enable-no-nans-fp-math
105
106 Enable optimizations that assume no NAN values.
107
108.. option:: --enable-unsafe-fp-math
109
110 Enable optimizations that make unsafe assumptions about IEEE math (e.g. that
111 addition is associative) or may not work for all input ranges.  These
112 optimizations allow the code generator to make use of some instructions which
113 would otherwise not be usable (such as ``fsin`` on X86).
114
115.. option:: --stats
116
117 Print statistics recorded by code-generation passes.
118
119.. option:: --time-passes
120
121 Record the amount of time needed for each pass and print a report to standard
122 error.
123
124.. option:: --load=<dso_path>
125
126 Dynamically load ``dso_path`` (a path to a dynamically shared object) that
127 implements an LLVM target.  This will permit the target name to be used with
128 the :option:`-march` option so that code can be generated for that target.
129
130.. option:: -meabi=[default|gnu|4|5]
131
132 Specify which EABI version should conform to.  Valid EABI versions are *gnu*,
133 *4* and *5*.  Default value (*default*) depends on the triple.
134
135
136Tuning/Configuration Options
137~~~~~~~~~~~~~~~~~~~~~~~~~~~~
138
139.. option:: --print-machineinstrs
140
141 Print generated machine code between compilation phases (useful for debugging).
142
143.. option:: --regalloc=<allocator>
144
145 Specify the register allocator to use.
146 Valid register allocators are:
147
148 *basic*
149
150  Basic register allocator.
151
152 *fast*
153
154  Fast register allocator. It is the default for unoptimized code.
155
156 *greedy*
157
158  Greedy register allocator. It is the default for optimized code.
159
160 *pbqp*
161
162  Register allocator based on 'Partitioned Boolean Quadratic Programming'.
163
164.. option:: --spiller=<spiller>
165
166 Specify the spiller to use for register allocators that support it.  Currently
167 this option is used only by the linear scan register allocator.  The default
168 ``spiller`` is *local*.  Valid spillers are:
169
170 *simple*
171
172  Simple spiller
173
174 *local*
175
176  Local spiller
177
178Intel IA-32-specific Options
179~~~~~~~~~~~~~~~~~~~~~~~~~~~~
180
181.. option:: --x86-asm-syntax=[att|intel]
182
183 Specify whether to emit assembly code in AT&T syntax (the default) or Intel
184 syntax.
185
186EXIT STATUS
187-----------
188
189If :program:`llc` succeeds, it will exit with 0.  Otherwise, if an error
190occurs, it will exit with a non-zero value.
191
192SEE ALSO
193--------
194
195lli
196
197