5 :Version: 1 as of 2016-11-08
14 XRay is a function call tracing system which combines compiler-inserted
15 instrumentation points and a runtime library that can dynamically enable and
16 disable the instrumentation.
18 More high level information about XRay can be found in the `XRay whitepaper`_.
20 This document describes how to use XRay as implemented in LLVM.
25 XRay consists of three main parts:
27 - Compiler-inserted instrumentation points.
28 - A runtime library for enabling/disabling tracing at runtime.
29 - A suite of tools for analysing the traces.
31 **NOTE:** As of July 25, 2018 , XRay is only available for the following
32 architectures running Linux: x86_64, arm7 (no thumb), aarch64, powerpc64le,
33 mips, mipsel, mips64, mips64el, NetBSD: x86_64, FreeBSD: x86_64 and
36 The compiler-inserted instrumentation points come in the form of nop-sleds in
37 the final generated binary, and an ELF section named ``xray_instr_map`` which
38 contains entries pointing to these instrumentation points. The runtime library
39 relies on being able to access the entries of the ``xray_instr_map``, and
40 overwrite the instrumentation points at runtime.
45 You can use XRay in a couple of ways:
47 - Instrumenting your C/C++/Objective-C/Objective-C++ application.
48 - Generating LLVM IR with the correct function attributes.
50 The rest of this section covers these main ways and later on how to customize
51 what XRay does in an XRay-instrumented binary.
53 Instrumenting your C/C++/Objective-C Application
54 ------------------------------------------------
56 The easiest way of getting XRay instrumentation for your application is by
57 enabling the ``-fxray-instrument`` flag in your clang invocation.
63 clang -fxray-instrument ...
65 By default, functions that have at least 200 instructions (or contain a loop) will
66 get XRay instrumentation points. You can tweak that number through the
67 ``-fxray-instruction-threshold=`` flag:
71 clang -fxray-instrument -fxray-instruction-threshold=1 ...
73 The loop detection can be disabled with ``-fxray-ignore-loops`` to use only the
74 instruction threshold. You can also specifically instrument functions in your
75 binary to either always or never be instrumented using source-level attributes.
76 You can do it using the GCC-style attributes or C++11-style attributes.
80 [[clang::xray_always_instrument]] void always_instrumented();
82 [[clang::xray_never_instrument]] void never_instrumented();
84 void alt_always_instrumented() __attribute__((xray_always_instrument));
86 void alt_never_instrumented() __attribute__((xray_never_instrument));
88 When linking a binary, you can either manually link in the `XRay Runtime
89 Library`_ or use ``clang`` to link it in automatically with the
90 ``-fxray-instrument`` flag. Alternatively, you can statically link-in the XRay
91 runtime library from compiler-rt -- those archive files will take the name of
92 `libclang_rt.xray-{arch}` where `{arch}` is the mnemonic supported by clang
95 LLVM Function Attribute
96 -----------------------
98 If you're using LLVM IR directly, you can add the ``function-instrument``
99 string attribute to your functions, to get the similar effect that the
100 C/C++/Objective-C source-level attributes would get:
104 define i32 @always_instrument() uwtable "function-instrument"="xray-always" {
108 define i32 @never_instrument() uwtable "function-instrument"="xray-never" {
112 You can also set the ``xray-instruction-threshold`` attribute and provide a
113 numeric string value for how many instructions should be in the function before
114 it gets instrumented.
118 define i32 @maybe_instrument() uwtable "xray-instruction-threshold"="2" {
125 Attributes can be imbued through the use of special case files instead of
126 adding them to the original source files. You can use this to mark certain
127 functions and classes to be never, always, or instrumented with first-argument
128 logging from a file. The file's format is described below:
132 # Comments are supported
134 fun:always_instrument
135 fun:log_arg1=arg1 # Log the first argument for the function
140 These files can be provided through the ``-fxray-attr-list=`` flag to clang.
141 You may have multiple files loaded through multiple instances of the flag.
146 The XRay Runtime Library is part of the compiler-rt project, which implements
147 the runtime components that perform the patching and unpatching of inserted
148 instrumentation points. When you use ``clang`` to link your binaries and the
149 ``-fxray-instrument`` flag, it will automatically link in the XRay runtime.
151 The default implementation of the XRay runtime will enable XRay instrumentation
152 before ``main`` starts, which works for applications that have a short
153 lifetime. This implementation also records all function entry and exit events
154 which may result in a lot of records in the resulting trace.
156 Also by default the filename of the XRay trace is ``xray-log.XXXXXX`` where the
157 ``XXXXXX`` part is randomly generated.
159 These options can be controlled through the ``XRAY_OPTIONS`` environment
160 variable during program run-time, where we list down the options and their
163 +-------------------+-----------------+---------------+------------------------+
164 | Option | Type | Default | Description |
165 +===================+=================+===============+========================+
166 | patch_premain | ``bool`` | ``false`` | Whether to patch |
167 | | | | instrumentation points |
168 | | | | before main. |
169 +-------------------+-----------------+---------------+------------------------+
170 | xray_mode | ``const char*`` | ``""`` | Default mode to |
171 | | | | install and initialize |
172 | | | | before ``main``. |
173 +-------------------+-----------------+---------------+------------------------+
174 | xray_logfile_base | ``const char*`` | ``xray-log.`` | Filename base for the |
175 | | | | XRay logfile. |
176 +-------------------+-----------------+---------------+------------------------+
177 | verbosity | ``int`` | ``0`` | Runtime verbosity |
179 +-------------------+-----------------+---------------+------------------------+
182 In addition to environment variable, you can also provide your own definition of
183 ``const char *__xray_default_options(void)`` function, which returns the option
184 strings. This method effectively provides default options during program build
185 time. For example, you can create an additional source file (e.g. ``xray-opt.c``
186 ) with the following ``__xray_default_options`` definition:
190 __attribute__((xray_never_instrument))
191 const char *__xray_default_options() {
192 return "patch_premain=true,xray_mode=xray-basic";
195 And link it with the program you want to instrument:
199 clang -fxray-instrument prog.c xray-opt.c ...
201 The instrumented binary will use 'patch_premain=true,xray_mode=xray-basic' by
202 default even without setting ``XRAY_OPTIONS``.
204 Note that you still can override options designated by ``__xray_default_options``
205 using ``XRAY_OPTIONS`` during run-time.
207 If you choose to not use the default logging implementation that comes with the
208 XRay runtime and/or control when/how the XRay instrumentation runs, you may use
209 the XRay APIs directly for doing so. To do this, you'll need to include the
210 ``xray_log_interface.h`` from the compiler-rt ``xray`` directory. The important API
211 functions we list below:
213 - ``__xray_log_register_mode(...)``: Register a logging implementation against
214 a string Mode identifier. The implementation is an instance of
215 ``XRayLogImpl`` defined in ``xray/xray_log_interface.h``.
216 - ``__xray_log_select_mode(...)``: Select the mode to install, associated with
217 a string Mode identifier. Only implementations registered with
218 ``__xray_log_register_mode(...)`` can be chosen with this function.
219 - ``__xray_log_init_mode(...)``: This function allows for initializing and
220 re-initializing an installed logging implementation. See
221 ``xray/xray_log_interface.h`` for details, part of the XRay compiler-rt
224 Once a logging implementation has been initialized, it can be "stopped" by
225 finalizing the implementation through the ``__xray_log_finalize()`` function.
226 The finalization routine is the opposite of the initialization. When finalized,
227 an implementation's data can be cleared out through the
228 ``__xray_log_flushLog()`` function. For implementations that support in-memory
229 processing, these should register an iterator function to provide access to the
230 data via the ``__xray_log_set_buffer_iterator(...)`` which allows code calling
231 the ``__xray_log_process_buffers(...)`` function to deal with the data in
234 All of this is better explained in the ``xray/xray_log_interface.h`` header.
239 XRay supports a basic logging mode which will trace the application's
240 execution, and periodically append to a single log. This mode can be
241 installed/enabled by setting ``xray_mode=xray-basic`` in the ``XRAY_OPTIONS``
242 environment variable. Combined with ``patch_premain=true`` this can allow for
243 tracing applications from start to end.
245 Like all the other modes installed through ``__xray_log_select_mode(...)``, the
246 implementation can be configured through the ``__xray_log_init_mode(...)``
247 function, providing the mode string and the flag options. Basic-mode specific
248 defaults can be provided in the ``XRAY_BASIC_OPTIONS`` environment variable.
250 Flight Data Recorder Mode
251 -------------------------
253 XRay supports a logging mode which allows the application to only capture a
254 fixed amount of memory's worth of events. Flight Data Recorder (FDR) mode works
255 very much like a plane's "black box" which keeps recording data to memory in a
256 fixed-size circular queue of buffers, and have the data available
257 programmatically until the buffers are finalized and flushed. To use FDR mode
258 on your application, you may set the ``xray_mode`` variable to ``xray-fdr`` in
259 the ``XRAY_OPTIONS`` environment variable. Additional options to the FDR mode
260 implementation can be provided in the ``XRAY_FDR_OPTIONS`` environment
261 variable. Programmatic configuration can be done by calling
262 ``__xray_log_init_mode("xray-fdr", <configuration string>)`` once it has been
265 When the buffers are flushed to disk, the result is a binary trace format
266 described by `XRay FDR format <XRayFDRFormat.html>`_
268 When FDR mode is on, it will keep writing and recycling memory buffers until
269 the logging implementation is finalized -- at which point it can be flushed and
270 re-initialised later. To do this programmatically, we follow the workflow
275 // Patch the sleds, if we haven't yet.
276 auto patch_status = __xray_patch();
278 // Maybe handle the patch_status errors.
280 // When we want to flush the log, we need to finalize it first, to give
281 // threads a chance to return buffers to the queue.
282 auto finalize_status = __xray_log_finalize();
283 if (finalize_status != XRAY_LOG_FINALIZED) {
284 // maybe retry, or bail out.
287 // At this point, we are sure that the log is finalized, so we may try
289 auto flush_status = __xray_log_flushLog();
290 if (flush_status != XRAY_LOG_FLUSHED) {
291 // maybe retry, or bail out.
294 The default settings for the FDR mode implementation will create logs named
295 similarly to the basic log implementation, but will have a different log
296 format. All the trace analysis tools (and the trace reading library) will
297 support all versions of the FDR mode format as we add more functionality and
298 record types in the future.
300 **NOTE:** We do not promise perpetual support for when we update the log
301 versions we support going forward. Deprecation of the formats will be
302 announced and discussed on the developers mailing list.
307 We currently have the beginnings of a trace analysis tool in LLVM, which can be
308 found in the ``tools/llvm-xray`` directory. The ``llvm-xray`` tool currently
309 supports the following subcommands:
311 - ``extract``: Extract the instrumentation map from a binary, and return it as
313 - ``account``: Performs basic function call accounting statistics with various
314 options for sorting, and output formats (supports CSV, YAML, and
315 console-friendly TEXT).
316 - ``convert``: Converts an XRay log file from one format to another. We can
317 convert from binary XRay traces (both basic and FDR mode) to YAML,
318 `flame-graph <https://github.com/brendangregg/FlameGraph>`_ friendly text
319 formats, as well as `Chrome Trace Viewer (catapult)
320 <https://github.com/catapult-project/catapult>` formats.
321 - ``graph``: Generates a DOT graph of the function call relationships between
322 functions found in an XRay trace.
323 - ``stack``: Reconstructs function call stacks from a timeline of function
324 calls in an XRay trace.
326 These subcommands use various library components found as part of the XRay
327 libraries, distributed with the LLVM distribution. These are:
329 - ``llvm/XRay/Trace.h`` : A trace reading library for conveniently loading
330 an XRay trace of supported forms, into a convenient in-memory representation.
331 All the analysis tools that deal with traces use this implementation.
332 - ``llvm/XRay/Graph.h`` : A semi-generic graph type used by the graph
333 subcommand to conveniently represent a function call graph with statistics
334 associated with edges and vertices.
335 - ``llvm/XRay/InstrumentationMap.h``: A convenient tool for analyzing the
336 instrumentation map in XRay-instrumented object files and binaries. The
337 ``extract`` and ``stack`` subcommands uses this particular library.
340 Minimizing Binary Size
341 ----------------------
343 XRay supports several different instrumentation points including ``function-entry``,
344 ``function-exit``, ``custom``, and ``typed`` points. These can be enabled individually
345 using the ``-fxray-instrumentation-bundle=`` flag. For example if you only wanted to
346 instrument function entry and custom points you could specify:
350 clang -fxray-instrument -fxray-instrumentation-bundle=function-entry,custom ...
352 This will omit the other sled types entirely, reducing the binary size. You can also
353 instrument just a sampled subset of functions using instrumentation groups.
354 For example, to instrument only a quarter of available functions invoke:
358 clang -fxray-instrument -fxray-function-groups=4
360 A subset will be chosen arbitrarily based on a hash of the function name. To sample a
361 different subset you can specify ``-fxray-selected-function-group=`` with a group number
362 in the range of 0 to ``xray-function-groups`` - 1. Together these options could be used
363 to produce multiple binaries with different instrumented subsets. If all you need is
364 runtime control over which functions are being traced at any given time it is better
365 to selectively patch and unpatch the individual functions you need using the XRay
366 Runtime Library's ``__xray_patch_function()`` method.
371 There are a number of ongoing efforts for expanding the toolset building around
372 the XRay instrumentation system.
377 - Work is in progress to integrate with or develop tools to visualize findings
378 from an XRay trace. Particularly, the ``stack`` tool is being expanded to
379 output formats that allow graphing and exploring the duration of time in each
381 - With a large instrumented binary, the size of generated XRay traces can
382 quickly become unwieldy. We are working on integrating pruning techniques and
383 heuristics for the analysis tools to sift through the traces and surface only
384 relevant information.
389 We're looking forward to contributions to port XRay to more architectures and
394 .. _`XRay whitepaper`: http://research.google.com/pubs/pub45287.html