1 llc - LLVM static compiler
2 ==========================
9 :program:`llc` [*options*] [*filename*]
14 The :program:`llc` command compiles LLVM source inputs into assembly language
15 for a specified architecture. The assembly language output can then be passed
16 through a native assembler and linker to generate a native executable.
18 The choice of architecture for the output assembly code is automatically
19 determined from the input file, unless the :option:`-march` option is used to
25 If ``filename`` is "``-``" or omitted, :program:`llc` reads from standard input.
26 Otherwise, it will read from ``filename``. Inputs can be in either the LLVM
27 assembly language format (``.ll``) or the LLVM bitcode format (``.bc``).
29 If the :option:`-o` option is omitted, then :program:`llc` will send its output
30 to standard output if the input is from standard input. If the :option:`-o`
31 option specifies "``-``", then the output will also be sent to standard output.
33 If no :option:`-o` option is specified and an input file other than "``-``" is
34 specified, then :program:`llc` creates the output filename by taking the input
35 filename, removing any existing ``.bc`` extension, and adding a ``.s`` suffix.
37 Other :program:`llc` options are described below.
44 Print a summary of command line options.
46 .. option:: -o <filename>
48 Use ``<filename>`` as the output filename. See the summary above for more
53 Generate code at different optimization levels. These correspond to the
54 ``-O0``, ``-O1``, ``-O2``, and ``-O3`` optimization levels used by
57 .. option:: -mtriple=<target triple>
59 Override the target triple specified in the input file with the specified
62 .. option:: -march=<arch>
64 Specify the architecture for which to generate assembly, overriding the target
65 encoded in the input file. See the output of ``llc -help`` for a list of
66 valid architectures. By default this is inferred from the target triple or
67 autodetected to the current architecture.
69 .. option:: -mcpu=<cpuname>
71 Specify a specific chip in the current architecture to generate code for.
72 By default this is inferred from the target triple and autodetected to
73 the current architecture. For a list of available CPUs, use:
77 llvm-as < /dev/null | llc -march=xyz -mcpu=help
79 .. option:: -filetype=<output file type>
81 Specify what kind of output ``llc`` should generated. Options are: ``asm``
82 for textual assembly ( ``'.s'``), ``obj`` for native object files (``'.o'``)
83 and ``null`` for not emitting anything (for performance testing).
85 Note that not all targets support all options.
87 .. option:: -mattr=a1,+a2,-a3,...
89 Override or control specific attributes of the target, such as whether SIMD
90 operations are enabled or not. The default set of attributes is set by the
91 current CPU. For a list of available attributes, use:
95 llvm-as < /dev/null | llc -march=xyz -mattr=help
97 .. option:: --frame-pointer
99 Specify effect of frame pointer elimination optimization (all,non-leaf,none).
101 .. option:: --disable-excess-fp-precision
103 Disable optimizations that may produce excess precision for floating point.
104 Note that this option can dramatically slow down code on some systems
107 .. option:: --enable-no-infs-fp-math
109 Enable optimizations that assume no Inf values.
111 .. option:: --enable-no-nans-fp-math
113 Enable optimizations that assume no NAN values.
115 .. option:: --enable-no-signed-zeros-fp-math
117 Enable FP math optimizations that assume the sign of 0 is insignificant.
119 .. option:: --enable-no-trapping-fp-math
121 Enable setting the FP exceptions build attribute not to use exceptions.
123 .. option:: --enable-unsafe-fp-math
125 Enable optimizations that make unsafe assumptions about IEEE math (e.g. that
126 addition is associative) or may not work for all input ranges. These
127 optimizations allow the code generator to make use of some instructions which
128 would otherwise not be usable (such as ``fsin`` on X86).
132 Print statistics recorded by code-generation passes.
134 .. option:: --time-passes
136 Record the amount of time needed for each pass and print a report to standard
139 .. option:: --load=<dso_path>
141 Dynamically load ``dso_path`` (a path to a dynamically shared object) that
142 implements an LLVM target. This will permit the target name to be used with
143 the :option:`-march` option so that code can be generated for that target.
145 .. option:: -meabi=[default|gnu|4|5]
147 Specify which EABI version should conform to. Valid EABI versions are *gnu*,
148 *4* and *5*. Default value (*default*) depends on the triple.
150 .. option:: -stack-size-section
152 Emit the .stack_sizes section which contains stack size metadata. The section
153 contains an array of pairs of function symbol values (pointer size) and stack
154 sizes (unsigned LEB128). The stack size values only include the space allocated
155 in the function prologue. Functions with dynamic stack allocations are not
158 .. option:: -remarks-section
160 Emit the __remarks (MachO) section which contains metadata about remark
163 Tuning/Configuration Options
164 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
166 .. option:: --print-after-isel
168 Print generated machine code after instruction selection (useful for debugging).
170 .. option:: --regalloc=<allocator>
172 Specify the register allocator to use.
173 Valid register allocators are:
177 Basic register allocator.
181 Fast register allocator. It is the default for unoptimized code.
185 Greedy register allocator. It is the default for optimized code.
189 Register allocator based on 'Partitioned Boolean Quadratic Programming'.
191 .. option:: --spiller=<spiller>
193 Specify the spiller to use for register allocators that support it. Currently
194 this option is used only by the linear scan register allocator. The default
195 ``spiller`` is *local*. Valid spillers are:
205 Intel IA-32-specific Options
206 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
208 .. option:: --x86-asm-syntax=[att|intel]
210 Specify whether to emit assembly code in AT&T syntax (the default) or Intel
216 If :program:`llc` succeeds, it will exit with 0. Otherwise, if an error
217 occurs, it will exit with a non-zero value.