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=output
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 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=path
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=path
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:: -text
106
107 Emit the profile in text mode. This option can also be used with both
108 sample-based and instrumentation-based profile. When this option is used
109 the profile will be dumped in the text format that is parsable by the profile
110 reader.
111
112 .. option:: -gcc
113
114 Emit the profile using GCC's gcov format (Not yet supported).
115
116.. option:: -sparse[=true|false]
117
118 Do not emit function records with 0 execution count. Can only be used in
119 conjunction with -instr. Defaults to false, since it can inhibit compiler
120 optimization during PGO.
121
122.. option:: -num-threads=N, -j=N
123
124 Use N threads to perform profile merging. When N=0, llvm-profdata auto-detects
125 an appropriate number of threads to use. This is the default.
126
127EXAMPLES
128^^^^^^^^
129Basic Usage
130+++++++++++
131Merge three profiles:
132
133::
134
135    llvm-profdata merge foo.profdata bar.profdata baz.profdata -output merged.profdata
136
137Weighted Input
138++++++++++++++
139The input file `foo.profdata` is especially important, multiply its counts by 10:
140
141::
142
143    llvm-profdata merge -weighted-input=10,foo.profdata bar.profdata baz.profdata -output merged.profdata
144
145Exactly equivalent to the previous invocation (explicit form; useful for programmatic invocation):
146
147::
148
149    llvm-profdata merge -weighted-input=10,foo.profdata -weighted-input=1,bar.profdata -weighted-input=1,baz.profdata -output merged.profdata
150
151.. program:: llvm-profdata show
152
153.. _profdata-show:
154
155SHOW
156----
157
158SYNOPSIS
159^^^^^^^^
160
161:program:`llvm-profdata show` [*options*] [*filename*]
162
163DESCRIPTION
164^^^^^^^^^^^
165
166:program:`llvm-profdata show` takes a profile data file and displays the
167information about the profile counters for this file and
168for any of the specified function(s).
169
170If *filename* is omitted or is ``-``, then **llvm-profdata show** reads its
171input from standard input.
172
173OPTIONS
174^^^^^^^
175
176.. option:: -all-functions
177
178 Print details for every function.
179
180.. option:: -counts
181
182 Print the counter values for the displayed functions.
183
184.. option:: -function=string
185
186 Print details for a function if the function's name contains the given string.
187
188.. option:: -help
189
190 Print a summary of command line options.
191
192.. option:: -output=output, -o=output
193
194 Specify the output file name.  If *output* is ``-`` or it isn't specified,
195 then the output is sent to standard output.
196
197.. option:: -instr (default)
198
199 Specify that the input profile is an instrumentation-based profile.
200
201.. option:: -text
202
203 Instruct the profile dumper to show profile counts in the text format of the
204 instrumentation-based profile data representation. By default, the profile
205 information is dumped in a more human readable form (also in text) with
206 annotations.
207
208.. option:: -topn=n
209
210 Instruct the profile dumper to show the top ``n`` functions with the
211 hottest basic blocks in the summary section. By default, the topn functions
212 are not dumped.
213
214.. option:: -sample
215
216 Specify that the input profile is a sample-based profile.
217
218.. option:: -memop-sizes
219
220 Show the profiled sizes of the memory intrinsic calls for shown functions.
221
222.. option:: -value-cutoff=n
223
224 Show only those functions whose max count values are greater or equal to ``n``.
225 By default, the value-cutoff is set to 0.
226
227.. option:: -list-below-cutoff
228
229 Only output names of functions whose max count value are below the cutoff
230 value.
231
232.. option:: -showcs
233
234 Only show context sensitive profile counts. The default is to filter all
235 context sensitive profile counts.
236
237.. program:: llvm-profdata overlap
238
239.. _profdata-overlap:
240
241OVERLAP
242-------
243
244SYNOPSIS
245^^^^^^^^
246
247:program:`llvm-profdata overlap` [*options*] [*base profile file*] [*test profile file*]
248
249DESCRIPTION
250^^^^^^^^^^^
251
252:program:`llvm-profdata overlap` takes two profile data files and displays the
253*overlap* of counter distribution between the whole files and between any of the
254specified functions.
255
256In this command, *overlap* is defined as follows:
257Suppose *base profile file* has the following counts:
258{c1_1, c1_2, ..., c1_n, c1_u_1, c2_u_2, ..., c2_u_s},
259and *test profile file* has
260{c2_1, c2_2, ..., c2_n, c2_v_1, c2_v_2, ..., c2_v_t}.
261Here c{1|2}_i (i = 1 .. n) are matched counters and c1_u_i (i = 1 .. s) and
262c2_v_i (i = 1 .. v) are unmatched counters (or counters only existing in)
263*base profile file* and *test profile file*, respectively.
264Let sum_1 = c1_1 + c1_2 +  ... + c1_n +  c1_u_1 + c2_u_2 + ... + c2_u_s, and
265sum_2 = c2_1 + c2_2 + ... + c2_n + c2_v_1 + c2_v_2 + ... + c2_v_t.
266*overlap* = min(c1_1/sum_1, c2_1/sum_2) + min(c1_2/sum_1, c2_2/sum_2) + ...
267+ min(c1_n/sum_1, c2_n/sum_2).
268
269The result overlap distribution is a percentage number, ranging from 0.0% to
270100.0%, where 0.0% means there is no overlap and 100.0% means a perfect
271overlap.
272
273Here is an example, if *base profile file* has counts of {400, 600}, and
274*test profile file* has matched counts of {60000, 40000}. The *overlap* is 80%.
275
276OPTIONS
277^^^^^^^
278
279.. option:: -function=string
280
281 Print details for a function if the function's name contains the given string.
282
283.. option:: -help
284
285 Print a summary of command line options.
286
287.. option:: -o=output or -o output
288
289 Specify the output file name.  If *output* is ``-`` or it isn't specified,
290 then the output is sent to standard output.
291
292.. option:: -value-cutoff=n
293
294 Show only those functions whose max count values are greater or equal to ``n``.
295 By default, the value-cutoff is set to max of unsigned long long.
296
297.. option:: -cs
298
299 Only show overlap for the context sensitive profile counts. The default is to show
300 non-context sensitive profile counts.
301
302EXIT STATUS
303-----------
304
305:program:`llvm-profdata` returns 1 if the command is omitted or is invalid,
306if it cannot read input files, or if there is a mismatch between their data.
307