1llvm-profdata - Profile data tool 2================================= 3 4.. program:: llvm-profdata 5 6SYNOPSIS 7-------- 8 9:program:`llvm-profdata` *command* [*args...*] 10 11DESCRIPTION 12----------- 13 14The :program:`llvm-profdata` tool is a small utility for working with profile 15data files. 16 17COMMANDS 18-------- 19 20* :ref:`merge <profdata-merge>` 21* :ref:`show <profdata-show>` 22* :ref:`overlap <profdata-overlap>` 23 24.. program:: llvm-profdata merge 25 26.. _profdata-merge: 27 28MERGE 29----- 30 31SYNOPSIS 32^^^^^^^^ 33 34:program:`llvm-profdata merge` [*options*] [*filename...*] 35 36DESCRIPTION 37^^^^^^^^^^^ 38 39:program:`llvm-profdata merge` takes several profile data files 40generated by PGO instrumentation and merges them together into a single 41indexed profile data file. 42 43By default profile data is merged without modification. This means that the 44relative importance of each input file is proportional to the number of samples 45or counts it contains. In general, the input from a longer training run will be 46interpreted as relatively more important than a shorter run. Depending on the 47nature of the training runs it may be useful to adjust the weight given to each 48input file by using the ``-weighted-input`` option. 49 50Profiles passed in via ``-weighted-input``, ``-input-files``, or via positional 51arguments are processed once for each time they are seen. 52 53 54OPTIONS 55^^^^^^^ 56 57.. option:: --help 58 59 Print a summary of command line options. 60 61.. option:: --output=<output>, -o 62 63 Specify the output file name. *Output* cannot be ``-`` as the resulting 64 indexed profile data can't be written to standard output. 65 66.. option:: --weighted-input=<weight,filename> 67 68 Specify an input file name along with a weight. The profile counts of the 69 supplied ``filename`` will be scaled (multiplied) by the supplied 70 ``weight``, where ``weight`` is a decimal integer >= 1. 71 Input files specified without using this option are assigned a default 72 weight of 1. Examples are shown below. 73 74.. option:: --input-files=<path>, -f 75 76 Specify a file which contains a list of files to merge. The entries in this 77 file are newline-separated. Lines starting with '#' are skipped. Entries may 78 be of the form <filename> or <weight>,<filename>. 79 80.. option:: --remapping-file=<path>, -r 81 82 Specify a file which contains a remapping from symbol names in the input 83 profile to the symbol names that should be used in the output profile. The 84 file should consist of lines of the form ``<input-symbol> <output-symbol>``. 85 Blank lines and lines starting with ``#`` are skipped. 86 87 The :doc:`llvm-cxxmap <llvm-cxxmap>` tool can be used to generate the symbol 88 remapping file. 89 90.. option:: --instr (default) 91 92 Specify that the input profile is an instrumentation-based profile. 93 94.. option:: --sample 95 96 Specify that the input profile is a sample-based profile. 97 98 The format of the generated file can be generated in one of three ways: 99 100 .. option:: --binary (default) 101 102 Emit the profile using a binary encoding. For instrumentation-based profile 103 the output format is the indexed binary format. 104 105 .. option:: --extbinary 106 107 Emit the profile using an extensible binary encoding. This option can only 108 be used with sample-based profile. The extensible binary encoding can be 109 more compact with compression enabled and can be loaded faster than the 110 default binary encoding. 111 112 .. option:: --text 113 114 Emit the profile in text mode. This option can also be used with both 115 sample-based and instrumentation-based profile. When this option is used 116 the profile will be dumped in the text format that is parsable by the profile 117 reader. 118 119 .. option:: --gcc 120 121 Emit the profile using GCC's gcov format (Not yet supported). 122 123.. option:: --sparse[=true|false] 124 125 Do not emit function records with 0 execution count. Can only be used in 126 conjunction with -instr. Defaults to false, since it can inhibit compiler 127 optimization during PGO. 128 129.. option:: --num-threads=<N>, -j 130 131 Use N threads to perform profile merging. When N=0, llvm-profdata auto-detects 132 an appropriate number of threads to use. This is the default. 133 134.. option:: --failure-mode=[any|all] 135 136 Set the failure mode. There are two options: 'any' causes the merge command to 137 fail if any profiles are invalid, and 'all' causes the merge command to fail 138 only if all profiles are invalid. If 'all' is set, information from any 139 invalid profiles is excluded from the final merged product. The default 140 failure mode is 'any'. 141 142.. option:: --prof-sym-list=<path> 143 144 Specify a file which contains a list of symbols to generate profile symbol 145 list in the profile. This option can only be used with sample-based profile 146 in extbinary format. The entries in this file are newline-separated. 147 148.. option:: --compress-all-sections=[true|false] 149 150 Compress all sections when writing the profile. This option can only be used 151 with sample-based profile in extbinary format. 152 153.. option:: --use-md5=[true|false] 154 155 Use MD5 to represent string in name table when writing the profile. 156 This option can only be used with sample-based profile in extbinary format. 157 158.. option:: --gen-partial-profile=[true|false] 159 160 Mark the profile to be a partial profile which only provides partial profile 161 coverage for the optimized target. This option can only be used with 162 sample-based profile in extbinary format. 163 164.. option:: --supplement-instr-with-sample=<file> 165 166 Supplement an instrumentation profile with sample profile. The sample profile 167 is the input of the flag. Output will be in instrumentation format (only works 168 with -instr). 169 170.. option:: --zero-counter-threshold=<float> 171 172 For the function which is cold in instr profile but hot in sample profile, if 173 the ratio of the number of zero counters divided by the total number of 174 counters is above the threshold, the profile of the function will be regarded 175 as being harmful for performance and will be dropped. 176 177.. option:: --instr-prof-cold-threshold=<int> 178 179 User specified cold threshold for instr profile which will override the cold 180 threshold got from profile summary. 181 182.. option:: --suppl-min-size-threshold=<int> 183 184 If the size of a function is smaller than the threshold, assume it can be 185 inlined by PGO early inliner and it will not be adjusted based on sample 186 profile. 187 188.. option:: --debug-info=<path> 189 190 Specify the executable or ``.dSYM`` that contains debug info for the raw profile. 191 When ``-debug-info-correlate`` was used for instrumentation, use this option 192 to correlate the raw profile. 193 194 195EXAMPLES 196^^^^^^^^ 197Basic Usage 198+++++++++++ 199Merge three profiles: 200 201:: 202 203 llvm-profdata merge foo.profdata bar.profdata baz.profdata -output merged.profdata 204 205Weighted Input 206++++++++++++++ 207The input file ``foo.profdata`` is especially important, multiply its counts by 10: 208 209:: 210 211 llvm-profdata merge --weighted-input=10,foo.profdata bar.profdata baz.profdata --output merged.profdata 212 213Exactly equivalent to the previous invocation (explicit form; useful for programmatic invocation): 214 215:: 216 217 llvm-profdata merge --weighted-input=10,foo.profdata --weighted-input=1,bar.profdata --weighted-input=1,baz.profdata --output merged.profdata 218 219.. program:: llvm-profdata show 220 221.. _profdata-show: 222 223SHOW 224---- 225 226SYNOPSIS 227^^^^^^^^ 228 229:program:`llvm-profdata show` [*options*] [*filename*] 230 231DESCRIPTION 232^^^^^^^^^^^ 233 234:program:`llvm-profdata show` takes a profile data file and displays the 235information about the profile counters for this file and 236for any of the specified function(s). 237 238If *filename* is omitted or is ``-``, then **llvm-profdata show** reads its 239input from standard input. 240 241OPTIONS 242^^^^^^^ 243 244.. option:: --all-functions 245 246 Print details for every function. 247 248.. option:: --counts 249 250 Print the counter values for the displayed functions. 251 252.. option:: --function=<string> 253 254 Print details for a function if the function's name contains the given string. 255 256.. option:: --help 257 258 Print a summary of command line options. 259 260.. option:: --output=<output>, -o 261 262 Specify the output file name. If *output* is ``-`` or it isn't specified, 263 then the output is sent to standard output. 264 265.. option:: --instr (default) 266 267 Specify that the input profile is an instrumentation-based profile. 268 269.. option:: --text 270 271 Instruct the profile dumper to show profile counts in the text format of the 272 instrumentation-based profile data representation. By default, the profile 273 information is dumped in a more human readable form (also in text) with 274 annotations. 275 276.. option:: --topn=<n> 277 278 Instruct the profile dumper to show the top ``n`` functions with the 279 hottest basic blocks in the summary section. By default, the topn functions 280 are not dumped. 281 282.. option:: --sample 283 284 Specify that the input profile is a sample-based profile. 285 286.. option:: --memop-sizes 287 288 Show the profiled sizes of the memory intrinsic calls for shown functions. 289 290.. option:: --value-cutoff=<n> 291 292 Show only those functions whose max count values are greater or equal to ``n``. 293 By default, the value-cutoff is set to 0. 294 295.. option:: --list-below-cutoff 296 297 Only output names of functions whose max count value are below the cutoff 298 value. 299 300.. option:: --showcs 301 302 Only show context sensitive profile counts. The default is to filter all 303 context sensitive profile counts. 304 305.. option:: --show-prof-sym-list=[true|false] 306 307 Show profile symbol list if it exists in the profile. This option is only 308 meaningful for sample-based profile in extbinary format. 309 310.. option:: --show-sec-info-only=[true|false] 311 312 Show basic information about each section in the profile. This option is 313 only meaningful for sample-based profile in extbinary format. 314 315.. program:: llvm-profdata overlap 316 317.. _profdata-overlap: 318 319OVERLAP 320------- 321 322SYNOPSIS 323^^^^^^^^ 324 325:program:`llvm-profdata overlap` [*options*] [*base profile file*] [*test profile file*] 326 327DESCRIPTION 328^^^^^^^^^^^ 329 330:program:`llvm-profdata overlap` takes two profile data files and displays the 331*overlap* of counter distribution between the whole files and between any of the 332specified functions. 333 334In this command, *overlap* is defined as follows: 335Suppose *base profile file* has the following counts: 336{c1_1, c1_2, ..., c1_n, c1_u_1, c2_u_2, ..., c2_u_s}, 337and *test profile file* has 338{c2_1, c2_2, ..., c2_n, c2_v_1, c2_v_2, ..., c2_v_t}. 339Here c{1|2}_i (i = 1 .. n) are matched counters and c1_u_i (i = 1 .. s) and 340c2_v_i (i = 1 .. v) are unmatched counters (or counters only existing in) 341*base profile file* and *test profile file*, respectively. 342Let sum_1 = c1_1 + c1_2 + ... + c1_n + c1_u_1 + c2_u_2 + ... + c2_u_s, and 343sum_2 = c2_1 + c2_2 + ... + c2_n + c2_v_1 + c2_v_2 + ... + c2_v_t. 344*overlap* = min(c1_1/sum_1, c2_1/sum_2) + min(c1_2/sum_1, c2_2/sum_2) + ... 345+ min(c1_n/sum_1, c2_n/sum_2). 346 347The result overlap distribution is a percentage number, ranging from 0.0% to 348100.0%, where 0.0% means there is no overlap and 100.0% means a perfect 349overlap. 350 351Here is an example, if *base profile file* has counts of {400, 600}, and 352*test profile file* has matched counts of {60000, 40000}. The *overlap* is 80%. 353 354OPTIONS 355^^^^^^^ 356 357.. option:: --function=<string> 358 359 Print details for a function if the function's name contains the given string. 360 361.. option:: --help 362 363 Print a summary of command line options. 364 365.. option:: --output=<output>, -o 366 367 Specify the output file name. If *output* is ``-`` or it isn't specified, 368 then the output is sent to standard output. 369 370.. option:: --value-cutoff=<n> 371 372 Show only those functions whose max count values are greater or equal to ``n``. 373 By default, the value-cutoff is set to max of unsigned long long. 374 375.. option:: --cs 376 377 Only show overlap for the context sensitive profile counts. The default is to show 378 non-context sensitive profile counts. 379 380EXIT STATUS 381----------- 382 383:program:`llvm-profdata` returns 1 if the command is omitted or is invalid, 384if it cannot read input files, or if there is a mismatch between their data. 385