1.. SPDX-License-Identifier: BSD-3-Clause 2 Copyright(c) 2010-2014 Intel Corporation. 3 4Profile Your Application 5======================== 6 7The following sections describe methods of profiling DPDK applications on 8different architectures. 9 10 11Profiling on x86 12---------------- 13 14Intel processors provide performance counters to monitor events. 15Some tools provided by Intel, such as Intel® VTune™ Amplifier, can be used 16to profile and benchmark an application. 17See the *VTune Performance Analyzer Essentials* publication from Intel Press for more information. 18 19For a DPDK application, this can be done in a Linux* application environment only. 20 21The main situations that should be monitored through event counters are: 22 23* Cache misses 24 25* Branch mis-predicts 26 27* DTLB misses 28 29* Long latency instructions and exceptions 30 31Refer to the 32`Intel Performance Analysis Guide <http://software.intel.com/sites/products/collateral/hpc/vtune/performance_analysis_guide.pdf>`_ 33for details about application profiling. 34 35 36Profiling with VTune 37~~~~~~~~~~~~~~~~~~~~ 38 39To allow VTune attaching to the DPDK application, reconfigure a DPDK build 40folder by passing ``-Dc_args=-DRTE_ETHDEV_PROFILE_WITH_VTUNE`` meson option 41and recompile the DPDK: 42 43.. code-block:: console 44 45 meson setup build 46 meson configure build -Dc_args=-DRTE_ETHDEV_PROFILE_WITH_VTUNE 47 ninja -C build 48 49 50Profiling on ARM64 51------------------ 52 53Using Linux perf 54~~~~~~~~~~~~~~~~ 55 56The ARM64 architecture provide performance counters to monitor events. The 57Linux ``perf`` tool can be used to profile and benchmark an application. In 58addition to the standard events, ``perf`` can be used to profile arm64 59specific PMU (Performance Monitor Unit) events through raw events (``-e`` 60``-rXX``). 61 62For more details refer to the 63`ARM64 specific PMU events enumeration <http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.100095_0002_04_en/way1382543438508.html>`_. 64 65 66Low-resolution generic counter 67~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 68 69The default ``cntvct_el0`` based ``rte_rdtsc()`` provides a portable means to 70get a wall clock counter in user space. Typically it runs at a lower clock frequency than the CPU clock frequency. 71Cycles counted using this method should be scaled to CPU clock frequency. 72 73 74High-resolution cycle counter 75~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 76 77The alternative method to enable ``rte_rdtsc()`` for a high resolution wall 78clock counter is through the ARMv8 PMU subsystem. The PMU cycle counter runs 79at CPU frequency. However, access to the PMU cycle counter from user space is 80not enabled by default in the arm64 linux kernel. It is possible to enable 81cycle counter for user space access by configuring the PMU from the privileged 82mode (kernel space). 83 84By default the ``rte_rdtsc()`` implementation uses a portable ``cntvct_el0`` 85scheme. 86 87The example below shows the steps to configure the PMU based cycle counter on 88an ARMv8 machine. 89 90.. code-block:: console 91 92 git clone https://github.com/jerinjacobk/armv8_pmu_cycle_counter_el0 93 cd armv8_pmu_cycle_counter_el0 94 make 95 sudo insmod pmu_el0_cycle_counter.ko 96 97Please refer to :doc:`../linux_gsg/build_dpdk` for generic details on compiling DPDK with meson. 98 99In order to enable ``PMU`` based ``rte_rdtsc()``, user needs to configure the 100build with ``-Dc_args='-DRTE_ARM_EAL_RDTSC_USE_PMU'``. 101 102Example: 103 104.. code-block:: console 105 106 meson setup --cross config/arm/arm64_armv8_linux_gcc -Dc_args='-DRTE_ARM_EAL_RDTSC_USE_PMU' build 107 108.. warning:: 109 110 The PMU based scheme is useful for high accuracy performance profiling with 111 ``rte_rdtsc()``. However, this method can not be used in conjunction with 112 Linux userspace profiling tools like ``perf`` as this scheme alters the PMU 113 registers state. 114