Perf Tool For Mac
There are several graphical applications and command-line tools available for gathering performance metrics. Some performance tools are installed with Xcode. Others are available for download from the Apple Developer website.
Perf Tool For Mac
Instruments is not the only analysis tool that can gather data about the performance of your app. Other analysis tools are geared towards finding specific types of performance problems. Table 3-1 lists the analysis tools installed with Xcode and available for download.
Gathers GPU-related performance data, including data related to VRAM usage, video bus traffic, and hardware stalls among others. You can use this information to identify the cause of temporary slowdowns or sporadic hesitations in your OpenGL application.
Displays common usage statistics relating to memory and CPU usage for the currently running processes. You can also initiate the sampling of a process from this application. This tool provides information that is similar to that of the top tool.
The CHUD Tools include additional applications for doing hardware and low-level software analysis in OS X. (These tools are not able to analyze hardware running iOS.) Table 3-3 lists the tools that are part of this package, which is available on the downloads page of the Apple Developer website.
A command-line tool that analyzes TT6E instruction traces and presents detailed analyses and histograms. You can use this tool to detect bad instruction sequences, such as misaligned operands, data dependency stalls, and spilled loads.
Table 3-4 lists some additional command-line tools that you can use to monitor and analyze performance in OS X. (These tools are not able to monitor and analyze applications running in iOS.) These tools are located in the /usr/bin/ directory and must be run from a command-line prompt. Most are installed along with the Xcode Tools. For information about these tools, see OS X Man Pages.
I like to use the "Time Profiler" which will show you how much time your application is its various parts during execution. I haven't used perf myself, but from talks/videos that I've seen this seems to be the most common use.
This is a quick post that details how to run perf (also known as perf_events) on an OSX machine. Perf is a powerful Linux tool - it can instrument CPU performance and is capable of lightweight profiling. You can get more context about what perf can do here.
TLDR; If you need a quick solution, try using the Nginx-based Docker image with instructions here. The rest of the post just describes how to create a Node-based image and install perf.
On Linux, you may get mangled function names in profiling results. This indicates that Perf tool you are using is compiled without access to demangling functions, or it is disabled. In this case, try another version of Perf.
Additionally, you might be fine profiling the BSD based unix layer with the supplied command line tools. It really depends on which operations or aspect of system performance you wish to measure or debug.
To the specific topic of page faults, you can get cumulative statistics from top and vm_stat and then dig into specific processes using either dtrace or Instruments - which is a free tool part of Xcode on the app store and has a graphical interface for profiling the system calls per process or as a system.
The Microsoft 365 network connectivity test tool is located at It's an adjunct tool to the network assessment and network insights available in the Microsoft 365 admin center under the Health Connectivity menu.
Network insights in the Microsoft 365 Admin Center are based on regular in-product measurements for your Microsoft 365 tenant, aggregated each day. In comparison, network insights from the Microsoft 365 network connectivity test are run locally in the tool.
In-product testing is limited, and running tests local to the user collects more data resulting in deeper insights. Network insights in the Microsoft 365 Admin Center will show that there's a networking problem at a specific office location. The Microsoft 365 connectivity test can help to identify the root cause of that problem and provide a targeted performance improvement action.
When you click the Run test button, we show the running test page and identify the office location. You can type in your location by city, state, and country or choose to have it detected for you. If you detect the office location, the tool requests the latitude and longitude from the web browser and limits the accuracy to 300 meters by 300 meters before use. It's not necessary to identify the location more accurately than the building to measure network performance.
Network latency between the user office location and the Exchange Online service is compared to other Microsoft 365 customers in the same metro area. A network insight is shown if 10% or more of customers in the same metro area have better performance. This means their users will have better performance in the Microsoft 365 user interface.
Connectivity may be blocked by a proxy server, a firewall, or another network security device on the enterprise network perimeter. Connectivity to TCP port 80 is tested with an HTTP request and connectivity to TCP port 443 is tested with an HTTPS request. If there's no response the FQDN is marked as a failure. If there's an HTTP response code 407 the FQDN is marked as a failure. If there's an HTTP response code 403 then we check the Server attribute of the response and if it appears to be a proxy server we mark this as a failure. You can simulate the tests we perform with the Windows command-line tool curl.exe.
We provide a command line executable that can be used by your remote deployment and execution tools and run the same tests as are available in the Microsoft 365 network connectivity test tool web site.
The first time you launch the executable you will be prompted to accept the end user license agreement (EULA) before testing is performed. If you have already read and accepted the EULA you can create an empty file called Microsoft-365-Network-Connectivity-Test-EULA-accepted.txt in the current working directory for the executable process when it is launched. To accept the EULA you can type 'y' and press enter in the command line window when prompted.
Output of results are written to a JSON file in a folder called TestResults which is created in the current working directory of the process unless it already exists. The filename format for the output is connectivity_test_result_YYYY-MM-DD-HH-MM-SS.json. The results are in JSON nodes that match the output shown on the web page for the Microsoft 365 network connectivity test tool web site. A new result file is created each time you run it and the standalone executable does not upload results to your Microsoft tenant for viewing in the Admin Center Network Connectivity pages. Front door codes, longitudes, and latitudes are not included in the result file.
The commandline tool uses Windows Location Services to find the users City State Country information for determining some distances. If Windows Location Services is disabled in the control panel then user location based assessments will be blank. In Windows Settings "Location services" must be on and "Let desktop apps access your location" must also be on.
The commandline tool will attempt to install the .NET Framework if it is not already installed. It will also download the main testing executable from the Microsoft 365 network connectivity test tool and launch that.
Network Connectivity Scan Report (JSON)This file provides the raw output test results from the command-line version of the Microsoft 365 network connectivity test tool (MicrosoftConnectivityTest.exe).
A best Microsoft 365 service front door (formerly known as an optimal service front door) is one that is closest to your network egress, generally in your city or metro area. Use the Microsoft 365 network performance tool to determine location of your in-use Microsoft 365 service front door and the best service front door(s). If the tool determines your in-use front door is one of the best ones, then you should expect great connectivity into Microsoft's global network.
This is the wiki page for the Linux perf command, also called perf_events. perf is powerful: it can instrument CPU performance counters, tracepoints, kprobes, and uprobes (dynamic tracing). It is capable of lightweight profiling. It is also included in the Linux kernel, under tools/perf, and is frequently updated and enhanced.
Performance counters are CPU hardware registers that count hardware events such as instructions executed, cache-misses suffered, or branches mispredicted. They form a basis for profiling applications to trace dynamic control flow and identify hotspots. perf provides rich generalized abstractions over hardware specific capabilities. Among others, it provides per task, per CPU and per-workload counters, sampling on top of these and source code event annotation.
Tracepoints are instrumentation points placed at logical locations in code, such as for system calls, TCP/IP events, file system operations, etc. These have negligible overhead when not in use, and can be enabled by the perf command to collect information including timestamps and stack traces. perf can also dynamically create tracepoints using the kprobes and uprobes frameworks, for kernel and userspace dynamic tracing. The possibilities with these are endless.
Before diving too deeply into perf or eBPF to look for performance problems, it helps to know where to start. Netflix has an excellent post called Linux Performance Analysis in 60,000 Milliseconds which recommends some good tools. I recommend that you install these inside of your Docker containers to help with performance analysis. In my case, running pidstat and mpstat was sufficient to discover the cause of my slow builds.
I also installed tmux so I could run the command causing the problem in one pane, and run perf record in the other pane. An alternative would be to run docker exec into your running container, which would work well for something like a server.