1 .. SPDX-License-Identifier: BSD-3-Clause
2 Copyright(c) 2010-2014 Intel Corporation.
4 Profile Your Application
5 ========================
7 The following sections describe methods of profiling DPDK applications on
8 different architectures.
14 Intel processors provide performance counters to monitor events.
15 Some tools provided by Intel, such as Intel® VTune™ Amplifier, can be used
16 to profile and benchmark an application.
17 See the *VTune Performance Analyzer Essentials* publication from Intel Press for more information.
19 For a DPDK application, this can be done in a Linux* application environment only.
21 The main situations that should be monitored through event counters are:
29 * Long latency instructions and exceptions
32 `Intel Performance Analysis Guide <http://software.intel.com/sites/products/collateral/hpc/vtune/performance_analysis_guide.pdf>`_
33 for details about application profiling.
39 Iterations that yielded no RX packets (empty cycles, wasted iterations) can
40 be analyzed using VTune Amplifier. This profiling employs the
41 `Instrumentation and Tracing Technology (ITT) API
42 <https://software.intel.com/en-us/node/544195>`_
43 feature of VTune Amplifier and requires only reconfiguring the DPDK library,
44 no changes in a DPDK application are needed.
46 To trace wasted iterations on RX queues, first reconfigure DPDK with
47 ``CONFIG_RTE_ETHDEV_RXTX_CALLBACKS`` and
48 ``CONFIG_RTE_ETHDEV_PROFILE_ITT_WASTED_RX_ITERATIONS`` enabled.
50 Then rebuild DPDK, specifying paths to the ITT header and library, which can
51 be found in any VTune Amplifier distribution in the *include* and *lib*
52 directories respectively:
54 .. code-block:: console
56 make EXTRA_CFLAGS=-I<path to ittnotify.h> \
57 EXTRA_LDLIBS="-L<path to libittnotify.a> -littnotify"
59 Finally, to see wasted iterations in your performance analysis results,
60 select the *"Analyze user tasks, events, and counters"* checkbox in the
61 *"Analysis Type"* tab when configuring analysis via VTune Amplifier GUI.
62 Alternatively, when running VTune Amplifier via command line, specify
63 ``-knob enable-user-tasks=true`` option.
65 Collected regions of wasted iterations will be marked on VTune Amplifier's
66 timeline as ITT tasks. These ITT tasks have predefined names, containing
67 Ethernet device and RX queue identifiers.
76 The ARM64 architecture provide performance counters to monitor events. The
77 Linux ``perf`` tool can be used to profile and benchmark an application. In
78 addition to the standard events, ``perf`` can be used to profile arm64
79 specific PMU (Performance Monitor Unit) events through raw events (``-e``
82 For more derails refer to the
83 `ARM64 specific PMU events enumeration <http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.100095_0002_04_en/way1382543438508.html>`_.
86 High-resolution cycle counter
87 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
89 The default ``cntvct_el0`` based ``rte_rdtsc()`` provides a portable means to
90 get a wall clock counter in user space. Typically it runs at <= 100MHz.
92 The alternative method to enable ``rte_rdtsc()`` for a high resolution wall
93 clock counter is through the armv8 PMU subsystem. The PMU cycle counter runs
94 at CPU frequency. However, access to the PMU cycle counter from user space is
95 not enabled by default in the arm64 linux kernel. It is possible to enable
96 cycle counter for user space access by configuring the PMU from the privileged
99 By default the ``rte_rdtsc()`` implementation uses a portable ``cntvct_el0``
100 scheme. Application can choose the PMU based implementation with
101 ``CONFIG_RTE_ARM_EAL_RDTSC_USE_PMU``.
103 The example below shows the steps to configure the PMU based cycle counter on
106 .. code-block:: console
108 git clone https://github.com/jerinjacobk/armv8_pmu_cycle_counter_el0
109 cd armv8_pmu_cycle_counter_el0
111 sudo insmod pmu_el0_cycle_counter.ko
113 make config T=arm64-armv8a-linuxapp-gcc
114 echo "CONFIG_RTE_ARM_EAL_RDTSC_USE_PMU=y" >> build/.config
119 The PMU based scheme is useful for high accuracy performance profiling with
120 ``rte_rdtsc()``. However, this method can not be used in conjunction with
121 Linux userspace profiling tools like ``perf`` as this scheme alters the PMU