5 lli - directly execute programs from LLVM bitcode
9 B<lli> [I<options>] [I<filename>] [I<program args>]
13 B<lli> directly executes programs in LLVM bitcode format. It takes a program
14 in LLVM bitcode format and executes it using a just-in-time compiler, if one is
15 available for the current architecture, or an interpreter. B<lli> takes all of
16 the same code generator options as L<llc|llc>, but they are only effective when
17 B<lli> is using the just-in-time compiler.
19 If I<filename> is not specified, then B<lli> reads the LLVM bitcode for the
20 program from standard input.
22 The optional I<args> specified on the command line are passed to the program as
25 =head1 GENERAL OPTIONS
29 =item B<-fake-argv0>=I<executable>
31 Override the C<argv[0]> value passed into the executing program.
33 =item B<-force-interpreter>=I<{false,true}>
35 If set to true, use the interpreter even if a just-in-time compiler is available
36 for this architecture. Defaults to false.
40 Print a summary of command line options.
42 =item B<-load>=I<puginfilename>
44 Causes B<lli> to load the plugin (shared object) named I<pluginfilename> and use
49 Print statistics from the code-generation passes. This is only meaningful for
50 the just-in-time compiler, at present.
54 Record the amount of time needed for each code-generation pass and print it to
59 Print out the version of B<lli> and exit without doing anything else.
67 =item B<-mtriple>=I<target triple>
69 Override the target triple specified in the input bitcode file with the
70 specified string. This may result in a crash if you pick an
71 architecture which is not compatible with the current system.
73 =item B<-march>=I<arch>
75 Specify the architecture for which to generate assembly, overriding the target
76 encoded in the bitcode file. See the output of B<llc -help> for a list of
77 valid architectures. By default this is inferred from the target triple or
78 autodetected to the current architecture.
80 =item B<-mcpu>=I<cpuname>
82 Specify a specific chip in the current architecture to generate code for.
83 By default this is inferred from the target triple and autodetected to
84 the current architecture. For a list of available CPUs, use:
85 B<llvm-as E<lt> /dev/null | llc -march=xyz -mcpu=help>
87 =item B<-mattr>=I<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:
92 B<llvm-as E<lt> /dev/null | llc -march=xyz -mattr=help>
97 =head1 FLOATING POINT OPTIONS
101 =item B<-disable-excess-fp-precision>
103 Disable optimizations that may increase floating point precision.
105 =item B<-enable-no-infs-fp-math>
107 Enable optimizations that assume no Inf values.
109 =item B<-enable-no-nans-fp-math>
111 Enable optimizations that assume no NAN values.
113 =item B<-enable-unsafe-fp-math>
115 Causes B<lli> to enable optimizations that may decrease floating point
120 Causes B<lli> to generate software floating point library calls instead of
121 equivalent hardware instructions.
125 =head1 CODE GENERATION OPTIONS
129 =item B<-code-model>=I<model>
131 Choose the code model from:
133 default: Target default code model
134 small: Small code model
135 kernel: Kernel code model
136 medium: Medium code model
137 large: Large code model
139 =item B<-disable-post-RA-scheduler>
141 Disable scheduling after register allocation.
143 =item B<-disable-spill-fusing>
145 Disable fusing of spill code into instructions.
147 =item B<-enable-correct-eh-support>
149 Make the -lowerinvoke pass insert expensive, but correct, EH code.
151 =item B<-jit-enable-eh>
153 Exception handling should be enabled in the just-in-time compiler.
155 =item B<-join-liveintervals>
157 Coalesce copies (default=true).
159 =item B<-nozero-initialized-in-bss>
160 Don't place zero-initialized symbols into the BSS section.
162 =item B<-pre-RA-sched>=I<scheduler>
164 Instruction schedulers available (before register allocation):
166 =default: Best scheduler for the target
167 =none: No scheduling: breadth first sequencing
168 =simple: Simple two pass scheduling: minimize critical path and maximize processor utilization
169 =simple-noitin: Simple two pass scheduling: Same as simple except using generic latency
170 =list-burr: Bottom-up register reduction list scheduling
171 =list-tdrr: Top-down register reduction list scheduling
172 =list-td: Top-down list scheduler -print-machineinstrs - Print generated machine code
174 =item B<-regalloc>=I<allocator>
176 Register allocator to use (default=linearscan)
178 =bigblock: Big-block register allocator
179 =linearscan: linear scan register allocator =local - local register allocator
180 =simple: simple register allocator
182 =item B<-relocation-model>=I<model>
184 Choose relocation model from:
186 =default: Target default relocation model
187 =static: Non-relocatable code =pic - Fully relocatable, position independent code
188 =dynamic-no-pic: Relocatable external references, non-relocatable code
192 Spiller to use (default=local)
194 =simple: simple spiller
195 =local: local spiller
197 =item B<-x86-asm-syntax>=I<syntax>
199 Choose style of code to emit from X86 backend:
201 =att: Emit AT&T-style assembly
202 =intel: Emit Intel-style assembly
208 If B<lli> fails to load the program, it will exit with an exit code of 1.
209 Otherwise, it will return the exit code of the program it executes.
217 Maintained by the LLVM Team (L<http://llvm.org/>).