1 # BOLT - a post-link optimizer developed to speed up large applications
5 `llvm-bolt <executable> [-o outputfile] <executable>.bolt [-data=perf.fdata] [options]`
17 Display available options (--help-hidden for more)
21 Display all available options
25 Display list of available options (--help-list-hidden for more)
27 - `--help-list-hidden`
29 Display list of all available options
33 Display the version of this program
39 Write bolt info section in the output binary
47 Save recorded profile to a file
49 ### BOLT generic options:
51 - `--align-text=<uint>`
53 Alignment of .text section
57 Allow processing of stripped binaries
59 - `--alt-inst-feature-size=<uint>`
61 Size of feature field in .altinstructions
63 - `--alt-inst-has-padlen`
65 Specify that .altinstructions has padlen field
67 - `--asm-dump[=<dump folder>]`
69 Dump function into assembly
75 - `--bolt-id=<string>`
77 Add any string to tag this execution in the output binary via bolt info section
79 - `--break-funcs=<func1,func2,func3,...>`
81 List of functions to core dump on (debugging)
85 Perform verification of LLVM instruction encoding/decoding. Every instruction
86 in the input is decoded and re-encoded. If the resulting bytes do not match
87 the input, a warning message is printed.
89 - `--comp-dir-override=<string>`
91 Overrides DW_AT_comp_dir, and provides an alternative base location, which is
92 used with DW_AT_dwo_name to construct a path to *.dwo files.
94 - `--create-debug-names-section`
96 Creates .debug_names section, if the input binary doesn't have it already, for
99 - `--cu-processing-batch-size=<uint>`
101 Specifies the size of batches for processing CUs. Higher number has better
102 performance, but more memory usage. Default value is 1.
112 - `--debug-skeleton-cu`
114 Prints out offsets for abbrev and debug_info of Skeleton CUs that get patched.
116 - `--dot-tooltip-code`
118 Add basic block instructions as tool tips on nodes
120 - `--dump-alt-instructions`
122 Dump Linux alternative instructions info
124 - `--dump-cg=<string>`
126 Dump callgraph to the given file
130 Dump parsed bolt data for debugging
134 Dump function CFGs to graphviz format after each stage;enable '-print-loops'
135 for color-coded blocks
137 - `--dump-linux-exceptions`
139 Dump Linux kernel exception table
143 Dump raw ORC unwind information (sorted)
145 - `--dump-para-sites`
147 Dump Linux kernel paravitual patch sites
149 - `--dump-pci-fixups`
151 Dump Linux kernel PCI fixup table
155 Dump Linux kernel SMP locks
157 - `--dump-static-calls`
159 Dump Linux kernel static calls
161 - `--dump-static-keys`
163 Dump Linux kernel static keys jump table
165 - `--dwarf-output-path=<string>`
167 Path to where .dwo files or dwp file will be written out to.
171 Path and name to DWP file.
175 Print execution info based on profile
179 Print dyno stats after each stage
181 - `--dyno-stats-scale=<uint>`
183 Scale to be applied while reporting dyno stats
187 Write BOLT Address Translation tables
189 - `--force-data-relocations`
191 Force relocations to data sections to always be processed
195 Force patching of original entry points
197 - `--funcs=<func1,func2,func3,...>`
199 Limit optimizations to functions from the list
201 - `--funcs-file=<string>`
203 File with list of functions to optimize
205 - `--funcs-file-no-regex=<string>`
207 File with list of functions to optimize (non-regex)
209 - `--funcs-no-regex=<func1,func2,func3,...>`
211 Limit optimizations to functions from the list (non-regex)
215 Hot data symbols support (relocation mode)
217 - `--hot-functions-at-end`
219 If reorder-functions is used, order functions putting hottest last
223 Generate hot text symbols. Apply this option to a precompiled binary that
224 manually calls into hugify, such that at runtime hugify call will put hot code
225 into 2M pages. This requires relocation.
227 - `--hot-text-move-sections=<sec1,sec2,sec3,...>`
229 List of sections containing functions used for hugifying hot text. BOLT makes
230 sure these functions are not placed on the same page as the hot text.
231 (default='.stub,.mover').
233 - `--insert-retpolines`
235 Run retpoline insertion pass
239 Keep or generate .debug_aranges section if .gdb_index is written
243 Preserve intermediate .o file
247 Skip processing of cold functions
249 - `--log-file=<string>`
251 Redirect journaling to a file instead of stdout/stderr
253 - `--long-jump-labels`
255 Always use long jumps/nops for Linux kernel static keys
257 - `--match-profile-with-function-hash`
259 Match profile with function hash
261 - `--max-data-relocations=<uint>`
263 Maximum number of data relocations to process
265 - `--max-funcs=<uint>`
267 Maximum number of functions to process
271 Use regular size pages for code alignment
275 Disable multithreading
277 - `--pad-funcs=<func1:pad1,func2:pad2,func3:pad3,...>`
279 List of functions to pad with amount of bytes
283 Print mappings in the legend, between characters/blocks and text sections
287 - `--profile-format=<value>`
289 Format to dump profile output in aggregation mode, default is fdata
290 - `fdata`: offset-based plaintext format
291 - `yaml`: dense YAML representation
293 - `--r11-availability=<value>`
295 Determine the availability of r11 before indirect branches
296 - `never`: r11 not available
297 - `always`: r11 available before calls and jumps
298 - `abi`: r11 available before calls but not before jumps
302 Use relocations in the binary (default=autodetect)
306 Remove .symtab section
308 - `--reorder-skip-symbols=<symbol1,symbol2,symbol3,...>`
310 List of symbol names that cannot be reordered
312 - `--reorder-symbols=<symbol1,symbol2,symbol3,...>`
314 List of symbol names that can be reordered
316 - `--retpoline-lfence`
318 Determine if lfence instruction should exist in the retpoline
320 - `--skip-funcs=<func1,func2,func3,...>`
322 List of functions to skip
324 - `--skip-funcs-file=<string>`
326 File with list of functions to skip
330 Trust the input to be from a well-formed source
332 - `--tasks-per-thread=<uint>`
334 Number of tasks to be created per thread
338 Assume that execution stops at trap instruction
340 - `--thread-count=<uint>`
344 - `--top-called-limit=<uint>`
346 Maximum number of functions to print in top called functions section
350 In relocation mode trap upon entry to any function that uses AVX-512
355 Insert traps in old function bodies (relocation mode)
357 - `--update-debug-sections`
359 Update DWARF debug sections of the executable
363 Use GNU_STACK program header for new segment (workaround for issues with
368 Re-use space in old .text if possible (relocation mode)
372 Set verbosity level for diagnostic output
376 Output a single dwarf package file (dwp) instead of multiple non-relocatable
377 dwarf object files (dwo).
379 ### BOLT optimization options:
385 - `--align-blocks-min-size=<uint>`
387 Minimal size of the basic block that should be aligned
389 - `--align-blocks-threshold=<uint>`
391 Align only blocks with frequency larger than containing function execution
392 frequency specified in percent. E.g. 1000 means aligning blocks that are 10
393 times more frequently executed than the containing function.
395 - `--align-functions=<uint>`
397 Align functions at a given value (relocation mode)
399 - `--align-functions-max-bytes=<uint>`
401 Maximum number of bytes to use to align functions
405 Assume the ABI is never violated
407 - `--block-alignment=<uint>`
409 Boundary to use for alignment of basic blocks
411 - `--bolt-seed=<uint>`
413 Seed for randomization
415 - `--cg-from-perf-data`
417 Use perf data directly when constructing the call graph for stale functions
419 - `--cg-ignore-recursive-calls`
421 Ignore recursive calls when constructing the call graph
423 - `--cg-use-split-hot-size`
425 Use hot/cold data on basic blocks to determine hot sizes for call graph
428 - `--cold-threshold=<uint>`
430 Tenths of percents of main entry frequency to use as a threshold when
431 evaluating whether a basic block is cold (0 means it is only considered cold
432 if the block has zero samples). Default: 0
434 - `--elim-link-veneers`
436 Run veneer elimination pass
438 - `--eliminate-unreachable`
440 Eliminate unreachable code
442 - `--equalize-bb-counts`
444 Use same count for BBs that should have equivalent count (used in non-LBR and
447 - `--execution-count-threshold=<uint>`
449 Perform profiling accuracy-sensitive optimizations only if function execution
450 count >= the threshold (default: 0)
452 - `--fix-block-counts`
454 Adjust block counts based on outgoing branch counts
456 - `--fix-func-counts`
458 Adjust function counts based on basic blocks execution count
460 - `--force-inline=<func1,func2,func3,...>`
462 List of functions to always consider for inlining
464 - `--frame-opt=<value>`
466 Optimize stack frame accesses
467 - `none`: do not perform frame optimization
468 - `hot`: perform FOP on hot functions
469 - `all`: perform FOP on all functions
471 - `--frame-opt-rm-stores`
473 Apply additional analysis to remove stores (experimental)
475 - `--function-order=<string>`
477 File containing an ordered list of functions to use for function reordering
479 - `--generate-function-order=<string>`
481 File to dump the ordered list of functions to use for function reordering
483 - `--generate-link-sections=<string>`
485 Generate a list of function sections in a format suitable for inclusion in a
490 Share stubs across functions
494 Automatically put hot code on 2MB page(s) (hugify) at runtime. No manual call
495 to hugify is needed in the binary (which is what --hot-text relies on).
499 Fold functions with identical code
503 Alias for --indirect-call-promotion
505 - `--icp-calls-remaining-percent-threshold=<uint>`
507 The percentage threshold against remaining unpromoted indirect call count for
508 the promotion for calls
512 Alias for --indirect-call-promotion-calls-topn
514 - `--icp-calls-total-percent-threshold=<uint>`
516 The percentage threshold against total count for the promotion for calls
518 - `--icp-eliminate-loads`
520 Enable load elimination using memory profiling data when performing ICP
522 - `--icp-funcs=<func1,func2,func3,...>`
524 List of functions to enable ICP for
528 Only promote call targets eligible for inlining
530 - `--icp-jt-remaining-percent-threshold=<uint>`
532 The percentage threshold against remaining unpromoted indirect call count for
533 the promotion for jump tables
537 Alias for --icp-jump-tables-targets
541 Alias for --indirect-call-promotion-jump-tables-topn
543 - `--icp-jt-total-percent-threshold=<uint>`
545 The percentage threshold against total count for the promotion for jump tables
547 - `--icp-jump-tables-targets`
549 For jump tables, optimize indirect jmp targets instead of indices
551 - `--icp-mp-threshold`
553 Alias for --indirect-call-promotion-mispredict-threshold
555 - `--icp-old-code-sequence`
557 Use old code sequence for promoted calls
559 - `--icp-top-callsites=<uint>`
561 Optimize hottest calls until at least this percentage of all indirect calls
562 frequency is covered. 0 = all callsites
566 Alias for --indirect-call-promotion-topn
570 Alias for --indirect-call-promotion-use-mispredicts
572 - `--indirect-call-promotion=<value>`
574 Indirect call promotion
575 - `none`: do not perform indirect call promotion
576 - `calls`: perform ICP on indirect calls
577 - `jump-tables`: perform ICP on jump tables
578 - `all`: perform ICP on calls and jump tables
580 - `--indirect-call-promotion-calls-topn=<uint>`
582 Limit number of targets to consider when doing indirect call promotion on
585 - `--indirect-call-promotion-jump-tables-topn=<uint>`
587 Limit number of targets to consider when doing indirect call promotion on jump
590 - `--indirect-call-promotion-topn=<uint>`
592 Limit number of targets to consider when doing indirect call promotion. 0 = no
595 - `--indirect-call-promotion-use-mispredicts`
597 Use misprediction frequency for determining whether or not ICP should be
598 applied at a callsite. The -indirect-call-promotion-mispredict-threshold
599 value will be used by this heuristic
601 - `--infer-fall-throughs`
603 Infer execution count for fall-through blocks
605 - `--infer-stale-profile`
607 Infer counts from stale profile data.
615 Adjust function profile after inlining
617 - `--inline-limit=<uint>`
619 Maximum number of call sites to inline
621 - `--inline-max-iters=<uint>`
623 Maximum number of inline iterations
627 Inline memcpy using 'rep movsb' instruction (X86-only)
629 - `--inline-small-functions`
631 Inline functions if increase in size is less than defined by -inline-small-
634 - `--inline-small-functions-bytes=<uint>`
636 Max number of bytes for the function to be considered small for inlining
641 Instrument code to generate accurate profile data
643 - `--iterative-guess`
645 In non-LBR mode, guess edge counts using iterative technique
647 - `--jt-footprint-optimize-for-icache`
649 With jt-footprint-reduction, only process PIC jumptables and turn off other
650 transformations that increase code size
652 - `--jt-footprint-reduction`
654 Make jump tables size smaller at the cost of using more instructions at jump
657 - `--jump-tables=<value>`
659 Jump tables support (default=basic)
660 - `none`: do not optimize functions with jump tables
661 - `basic`: optimize functions with jump tables
662 - `move`: move jump tables to a separate section
663 - `split`: split jump tables section into hot and cold based on function
665 - `aggressive`: aggressively split jump tables section based on usage of the
670 Keep no-op instructions. By default they are removed.
672 - `--lite-threshold-count=<uint>`
674 Similar to '-lite-threshold-pct' but specify threshold using absolute function
675 call count. I.e. limit processing to functions executed at least the specified
678 - `--lite-threshold-pct=<uint>`
680 Threshold (in percent) for selecting functions to process in lite mode. Higher
681 threshold means fewer functions to process. E.g threshold of 90 means only top
682 10 percent of functions with profile will be processed.
684 - `--match-with-call-graph`
686 Match functions with call graph
688 - `--memcpy1-spec=<func1,func2:cs1:cs2,func3:cs1,...>`
690 List of functions with call sites for which to specialize memcpy() for size 1
692 - `--min-branch-clusters`
694 Use a modified clustering algorithm geared towards minimizing branches
696 - `--name-similarity-function-matching-threshold=<uint>`
698 Match functions using namespace and edit distance.
702 Disable all inlining (overrides other inlining options)
706 Do not scan cold functions for external references (may result in slower binary)
708 - `--peepholes=<value>`
710 Enable peephole optimizations
711 - `none`: disable peepholes
712 - `double-jumps`: remove double jumps when able
713 - `tailcall-traps`: insert tail call traps
714 - `useless-branches`: remove useless conditional branches
715 - `all`: enable all peephole optimizations
719 Optimize PLT calls (requires linking with -znow)
720 - `none`: do not optimize PLT calls
721 - `hot`: optimize executed (hot) PLT calls
722 - `all`: optimize all PLT calls
724 - `--preserve-blocks-alignment`
726 Try to preserve basic block alignment
728 - `--profile-ignore-hash`
730 Ignore hash while reading function profile
732 - `--profile-use-dfs`
734 Use DFS order for YAML profile
738 Reassign registers so as to avoid using REX prefixes in hot code
740 - `--reorder-blocks=<value>`
742 Change layout of basic blocks in a function
743 - `none`: do not reorder basic blocks
744 - `reverse`: layout blocks in reverse order
745 - `normal`: perform optimal layout based on profile
746 - `branch-predictor`: perform optimal layout prioritizing branch predictions
747 - `cache`: perform optimal layout prioritizing I-cache behavior
748 - `cache+`: perform layout optimizing I-cache behavior
749 - `ext-tsp`: perform layout optimizing I-cache behavior
750 - `cluster-shuffle`: perform random layout of clusters
752 - `--reorder-data=<section1,section2,section3,...>`
754 List of sections to reorder
756 - `--reorder-data-algo=<value>`
758 Algorithm used to reorder data sections
759 - `count`: sort hot data by read counts
760 - `funcs`: sort hot data by hot function usage and count
762 - `--reorder-data-inplace`
764 Reorder data sections in place
766 - `--reorder-data-max-bytes=<uint>`
768 Maximum number of bytes to reorder
770 - `--reorder-data-max-symbols=<uint>`
772 Maximum number of symbols to reorder
774 - `--reorder-functions=<value>`
776 Reorder and cluster functions (works only with relocations)
777 - `none`: do not reorder functions
778 - `exec-count`: order by execution count
779 - `hfsort`: use hfsort algorithm
780 - `hfsort+`: use cache-directed sort
781 - `cdsort`: use cache-directed sort
782 - `pettis-hansen`: use Pettis-Hansen algorithm
783 - `random`: reorder functions randomly
784 - `user`: use function order specified by -function-order
786 - `--reorder-functions-use-hot-size`
788 Use a function's hot size when doing clustering
790 - `--report-bad-layout=<uint>`
792 Print top <uint> functions with suboptimal code layout on input
796 Print the list of functions with stale profile
798 - `--runtime-hugify-lib=<string>`
800 Specify file name of the runtime hugify library
802 - `--runtime-instrumentation-lib=<string>`
804 Specify file name of the runtime instrumentation library
806 - `--sctc-mode=<value>`
808 Mode for simplify conditional tail calls
809 - `always`: always perform sctc
810 - `preserve`: only perform sctc when branch direction is preserved
811 - `heuristic`: use branch prediction data to control sctc
813 - `--sequential-disassembly`
815 Performs disassembly sequentially
817 - `--shrink-wrapping-threshold=<uint>`
819 Percentage of prologue execution count to use as threshold when evaluating
820 whether a block is cold enough to be profitable to move eligible spills there
822 - `--simplify-conditional-tail-calls`
824 Simplify conditional tail calls by removing unnecessary jumps
826 - `--simplify-rodata-loads`
828 Simplify loads from read-only sections by replacing the memory operand with
829 the constant found in the corresponding section
831 - `--split-align-threshold=<uint>`
833 When deciding to split a function, apply this alignment while doing the size
834 comparison (see -split-threshold). Default value: 2.
838 Outline as many cold basic blocks as possible
842 Split C++ exception handling code
844 - `--split-functions`
846 Split functions into fragments
848 - `--split-strategy=<value>`
850 Strategy used to partition blocks into fragments
851 - `profile2`: split each function into a hot and cold fragment using profiling
853 - `cdsplit`: split each function into a hot, warm, and cold fragment using
854 profiling information
855 - `random2`: split each function into a hot and cold fragment at a randomly
856 chosen split point (ignoring any available profiling information)
857 - `randomN`: split each function into N fragments at a randomly chosen split
858 points (ignoring any available profiling information)
859 - `all`: split all basic blocks of each function into fragments such that each
860 fragment contains exactly a single basic block
862 - `--split-threshold=<uint>`
864 Split function only if its main size is reduced by more than given amount of
865 bytes. Default value: 0, i.e. split iff the size is reduced. Note that on some
866 architectures the size can increase after splitting.
868 - `--stale-matching-max-func-size=<uint>`
870 The maximum size of a function to consider for inference.
872 - `--stale-matching-min-matched-block=<uint>`
874 Percentage threshold of matched basic blocks at which stale profile inference
877 - `--stale-threshold=<uint>`
879 Maximum percentage of stale functions to tolerate (default: 100)
883 Turn on the stoke analysis
887 Strip 'repz' prefix from 'repz retq' sequence (on by default)
889 - `--tail-duplication=<value>`
891 Duplicate unconditional branches that cross a cache line
892 - `none`: do not apply
893 - `aggressive`: aggressive strategy
894 - `moderate`: moderate strategy
895 - `cache`: cache-aware duplication strategy
897 - `--tsp-threshold=<uint>`
899 Maximum number of hot basic blocks in a function for which to use a precise
900 TSP solution while re-ordering basic blocks
902 - `--use-aggr-reg-reassign`
904 Use register liveness analysis to try to find more opportunities for -reg-
905 reassign optimization
907 - `--use-compact-aligner`
909 Use compact approach for aligning functions
911 - `--use-edge-counts`
913 Use edge count data when doing clustering
917 Verify the CFG after every pass
919 - `--x86-align-branch-boundary-hot-only`
921 Only apply branch boundary alignment in hot code
923 - `--x86-strip-redundant-address-size`
925 Remove redundant Address-Size override prefix
927 ### BOLT options in relocation mode:
929 - `--align-macro-fusion=<value>`
931 Fix instruction alignment for macro-fusion (x86 relocation mode)
932 - `none`: do not insert alignment no-ops for macro-fusion
933 - `hot`: only insert alignment no-ops on hot execution paths (default)
934 - `all`: always align instructions to allow macro-fusion
936 ### BOLT instrumentation options:
938 `llvm-bolt <executable> -instrument [-o outputfile] <instrumented-executable>`
940 - `--conservative-instrumentation`
942 Disable instrumentation optimizations that sacrifice profile accuracy (for
943 debugging, default: false)
945 - `--instrument-calls`
947 Record profile for inter-function control flow activity (default: true)
949 - `--instrument-hot-only`
951 Only insert instrumentation on hot functions (needs profile, default: false)
953 - `--instrumentation-binpath=<string>`
955 Path to instrumented binary in case if /proc/self/map_files is not accessible
956 due to access restriction issues
958 - `--instrumentation-file=<string>`
960 File name where instrumented profile will be saved (default: /tmp/prof.fdata)
962 - `--instrumentation-file-append-pid`
964 Append PID to saved profile file name (default: false)
966 - `--instrumentation-no-counters-clear`
968 Don't clear counters across dumps (use with instrumentation-sleep-time option)
970 - `--instrumentation-sleep-time=<uint>`
972 Interval between profile writes (default: 0 = write only at program end).
973 This is useful for service workloads when you want to dump profile every X
974 minutes or if you are killing the program and the profile is not being dumped
977 - `--instrumentation-wait-forks`
979 Wait until all forks of instrumented process will finish (use with
980 instrumentation-sleep-time option)
982 ### BOLT printing options:
986 Print aliases when printing objects
990 Print functions after each stage
994 Print functions after CFG construction
996 - `--print-debug-info`
998 Print debug info when printing functions
1002 Print function after disassembly
1004 - `--print-dyno-opcode-stats=<uint>`
1006 Print per instruction opcode dyno stats and the functionnames:BB offsets of
1007 the nth highest execution counts
1009 - `--print-dyno-stats-only`
1011 While printing functions output dyno-stats and skip instructions
1013 - `--print-exceptions`
1015 Print exception handling data
1019 Print global symbols after disassembly
1021 - `--print-jump-tables`
1027 Print loop related information
1029 - `--print-mem-data`
1031 Print memory data annotations when printing functions
1033 - `--print-normalized`
1035 Print functions after CFG is normalized
1037 - `--print-only=<func1,func2,func3,...>`
1039 List of functions to print
1043 Print ORC unwind information for instructions
1047 Print functions after attaching profile
1049 - `--print-profile-stats`
1051 Print profile quality/bias analysis
1053 - `--print-pseudo-probes=<value>`
1055 Print pseudo probe info
1056 - `decode`: decode probes section from binary
1057 - `address_conversion`: update address2ProbesMap with output block address
1058 - `encoded_probes`: display the encoded probes in binary section
1059 - `all`: enable all debugging printout
1061 - `--print-relocations`
1063 Print relocations when printing functions/objects
1065 - `--print-reordered-data`
1067 Print section contents after reordering
1069 - `--print-retpoline-insertion`
1071 Print functions after retpoline insertion pass
1075 Print all SDT markers
1077 - `--print-sections`
1079 Print all registered sections
1083 Print names of functions with unknown control flow
1087 Print time spent constructing binary functions
1091 Print time spent in rewriting passes
1093 - `--print-after-branch-fixup`
1095 Print function after fixing local branches
1097 - `--print-after-jt-footprint-reduction`
1099 Print function after jt-footprint-reduction pass
1101 - `--print-after-lowering`
1103 Print function after instruction lowering
1105 - `--print-cache-metrics`
1107 Calculate and print various metrics for instruction cache
1109 - `--print-clusters`
1113 - `--print-estimate-edge-counts`
1115 Print function after edge counts are set for no-LBR profile
1117 - `--print-finalized`
1119 Print function after CFG is finalized
1121 - `--print-fix-relaxations`
1123 Print functions after fix relaxations pass
1125 - `--print-fix-riscv-calls`
1127 Print functions after fix RISCV calls pass
1131 Print functions after frame optimizer pass
1133 - `--print-function-statistics=<uint>`
1135 Print statistics about basic block ordering
1139 Print functions after ICF optimization
1143 Print functions after indirect call promotion
1147 Print functions after inlining optimization
1149 - `--print-large-functions`
1151 Print functions that could not be overwritten due to excessive size
1155 Print functions after longjmp pass
1157 - `--print-optimize-bodyless`
1159 Print functions after bodyless optimization
1161 - `--print-output-address-range`
1163 Print output address range for each basic block in the function
1164 whenBinaryFunction::print is called
1166 - `--print-peepholes`
1168 Print functions after peephole optimization
1172 Print functions after PLT optimization
1174 - `--print-regreassign`
1176 Print functions after regreassign pass
1178 - `--print-reordered`
1180 Print functions after layout optimization
1182 - `--print-reordered-functions`
1184 Print functions after clustering
1188 Print functions after conditional tail call simplification
1190 - `--print-simplify-rodata-loads`
1192 Print functions after simplification of RO data loads
1194 - `--print-sorted-by=<value>`
1196 Print functions sorted by order of dyno stats
1197 - `executed-forward-branches`: executed forward branches
1198 - `taken-forward-branches`: taken forward branches
1199 - `executed-backward-branches`: executed backward branches
1200 - `taken-backward-branches`: taken backward branches
1201 - `executed-unconditional-branches`: executed unconditional branches
1202 - `all-function-calls`: all function calls
1203 - `indirect-calls`: indirect calls
1204 - `PLT-calls`: PLT calls
1205 - `executed-instructions`: executed instructions
1206 - `executed-load-instructions`: executed load instructions
1207 - `executed-store-instructions`: executed store instructions
1208 - `taken-jump-table-branches`: taken jump table branches
1209 - `taken-unknown-indirect-branches`: taken unknown indirect branches
1210 - `total-branches`: total branches
1211 - `taken-branches`: taken branches
1212 - `non-taken-conditional-branches`: non-taken conditional branches
1213 - `taken-conditional-branches`: taken conditional branches
1214 - `all-conditional-branches`: all conditional branches
1215 - `linker-inserted-veneer-calls`: linker-inserted veneer calls
1216 - `all`: sorted by all names
1218 - `--print-sorted-by-order=<value>`
1220 Use ascending or descending order when printing functions ordered by dyno stats
1224 Print functions after code splitting
1228 Print functions after stoke analysis
1232 Print functions after unreachable code elimination
1234 - `--print-veneer-elimination`
1236 Print functions after veneer elimination pass
1240 Print time spent in each optimization
1242 - `--print-all-options`
1244 Print all option values after command line parsing
1248 Print non-default options after command line parsing