This document is a user guide to the next-generation NVIDIA Nsight Compute profiling tools. NVIDIA Nsight Compute is an interactive kernel profiler for CUDA applications. It provides detailed performance metrics and API debugging via a user interface and command line tool. In addition, its baseline feature allows users to compare results within the tool. NVIDIA Nsight Compute provides a customizable and data-driven user interface and metric collection and can be extended with analysis scripts for post-processing results.

1. Launch the target application from NVIDIA Nsight Compute

   When starting NVIDIA Nsight Compute, the Welcome Page will appear. Click on Quick Launch to open the Connection dialog. If the Connection dialog doesn't appear, you can open it using the Connect button from the main toolbar, as long as you are not currently connected. Select your target platform on the left-hand side and your connection target (machine) from the Connection drop down. If you have your local target platform selected, localhost will become available as a connection. Use + to add a new connection target. Then, fill in the launch details and select Launch. In the Activity panel, select the Interactive Profile activity to initiate a session that allows controlling the execution of the target application and selecting the kernels of interest interactively. Press Launch to start the session.

   
   
   

   

   
   
   
2. Launch the target application with tools instrumentation from the command line
   The ncu can act as a simple wrapper that forces the target application to load the necessary libraries for tools instrumentation. The parameter --mode=launch specifies that the target application should be launched and suspended before the first instrumented API call. That way the application waits until we connect with the UI.

   $ ncu --mode=launch CuVectorAddDrv.exe

3. Launch NVIDIA Nsight Compute and connect to target application
   
   
   

   

   
   
   

   Select the target machine at the top of the dialog to connect and update the list of attachable applications. By default, localhost is pre-selected if the target matches your current local platform. Select the Attach tab and the target application of interest and press Attach. Once connected, the layout of NVIDIA Nsight Compute changes into stepping mode that allows you to control the execution of any calls into the instrumented API. When connected, the API Stream window indicates that the target application waits before the very first API call.

   
   
   

   

   
   
   
4. Control application execution

   Use the API Stream window to step the calls into the instrumented API. The dropdown at the top allows switching between different CPU threads of the application. Step In (F11), Step Over (F10), and Step Out (Shift + F11) are available from the Debug menu or the corresponding toolbar buttons. While stepping, function return values and function parameters are captured.

   
   
   

   

   
   
   

   Use Resume (F5) and Pause to allow the program to run freely. Freeze control is available to define the behavior of threads currently not in focus, i.e. selected in the thread drop down. By default, the API Stream stops on any API call that returns an error code. This can be toggled in the Debug menu by Break On API Error.

5. Isolate a kernel launch

   To quickly isolate a kernel launch for profiling, use the Next API Launch button in the toolbar of the API Stream window to jump to the next kernel launch. The execution will stop before the kernel launch is executed.

   
   
   

   

   
   
   
6. Profile a kernel launch

   Once the execution of the target application is suspended at a kernel launch, additional actions become available in the UI. These actions are either available from the menu or from the toolbar. Please note that the actions are disabled, if the API stream is not at a qualifying state (not at a kernel launch or launching on an unsupported GPU). To profile, press Profile and wait until the result is shown in the Profiler Report. Profiling progress is reported in the lower right corner status bar.

   Instead of manually selecting Profile, it is also possible to enable Auto Profile from the Profile menu. If enabled, each kernel matching the current kernel filter (if any) will be profiled using the current section configuration. This is especially useful if an application is to be profiled unattended, or the number of kernel launches to be profiled is very large. Sections can be enabled or disabled using the Sections/Rules Info tool window.

   Profile Series allows to configure the collection of a set of profile results at once. Each result in the set is profiled with varying parameters. Series are useful to investigate the behavior of a kernel across a large set of parameters without the need to recompile and rerun the application many times.

For a detailed description of the options available in this activity, see Interactive Profile Activity.

1. Launch the target application from NVIDIA Nsight Compute

   When starting NVIDIA Nsight Compute, the Welcome Page will appear. Click on Quick Launch to open the Connection dialog. If the Connection dialog doesn’t appear, you can open it using the Connect button from the main toolbar, as long as you are not currently connected. Select your target platform on the left-hand side and your localhost from the Connection drop down. Then, fill in the launch details and select Launch. In the Activity panel, select the Profile activity to initiate a session that pre-configures the profile session and launches the command line profiler to collect the data. Provide the Output File name to enable starting the session with the Launch button.

   
   
   

   

   
   
   
2. Additional Launch Options

   For more details on these options, see Command Line Options. The options are grouped into tabs: The Filter tab exposes the options to specify which kernels should be profiled. Options include the kernel regex filter, the number of launches to skip, and the total number of launch to profile. The Section tab allows you to select which sections should be collected for each kernel launch. The Sampling tab allows you to configure sampling options for each kernel launch. The Other tab includes the option to collect NVTX information or custom metrics via the --metrics option.

   The Section tab allows you to select which sections should be collected for each kernel launch. Hover over a section to see its description as a tool-tip. To change the sections that are enabled by default, use the Sections/Rules Info tool window.

   
   
   

   

   
   
   

For a detailed description of the options available in this activity, see Profile Activity.

1. Navigate the report

   The profile report comes up by default on the Details page. You can switch between different Report Pages of the report with the dropdown labeled Page on the top-left of the report. A report can contain any number of results from kernel launches. The Launch dropdown allows switching between the different results in a report.

   
   
   

   

   
   
   
2. Diffing multiple results

   On the Details page, press the button Add Baseline to promote the current result in focus to become the baseline all other results from this report and any other report opened in the same instance of NVIDIA Nsight Compute gets compared to. If a baseline is set, every element on the Details page shows two values: The current value of the result in focus and the corresponding value of the baseline or the percentage of change from the corresponding baseline value.

   
   
   

   

   
   
   

   Use the Clear Baselines entry from the dropdown button, the Profile menu or the corresponding toolbar button to remove all baselines. For more information see Baselines.

3. Executing rules

   On the Details page some sections may provide rules. Press the Apply button to execute an individual rule. The Apply Rules button on the top executes all available rules for the current result in focus. Rules can be user-defined too. For more information see the Customization Guide.

   
   
   

   

   
   
   

Remote devices that support SSH can also be configured as a target in the Connection Dialog. To configure a remote device, ensure an SSH-capable Target Platform is selected, then press the + button. The following configuration dialog will be presented.

NVIDIA Nsight Compute supports both password and private key authentication methods. In this dialog, select the authentication method and enter the following information:

In addition to keyfiles specified by path and plain password authentication, NVIDIA Nsight Compute supports keyboard-interactive authentication, standard keyfile path searching and SSH agents.

When all information is entered, click the Add button to make use of this new connection.

When a remote connection is selected in the Connection Dialog, the Application Executable file browser will browse the remote file system using the configured SSH connection, allowing the user to select the target application on the remote device.

The progress of each of these steps is presented in the Progress Log.

Progress Log

Note that once either activity type has been launched remotely, the tools necessary for further profiling sessions can be found in the Deployment Directory on the remote device.

On Linux and Mac host platforms, NVIDIA Nsight Compute supports SSH remote profiling on target machines which are not directly addressable from the machine the UI is running on through the ProxyJump and ProxyCommand SSH options.

These options can be used to specify intermediate hosts to connect to or actual commands to run to obtain a socket connected to the SSH server on the target host and can be added to your SSH configuration file.

Note that for both options, NVIDIA Nsight Compute runs external commands and does not implement any mechanism to authenticate to the intermediate hosts using the credentials entered in the Connection Dialog. These credentials will only be used to authenticate to the final target in the chain of machines.

When using the ProxyJump option NVIDIA Nsight Compute uses the OpenSSH client to establish the connection to the intermediate hosts. This means that in order to use ProxyJump or ProxyCommand, a version of OpenSSH supporting these options must be installed on the host machine.

A common way to authenticate to the intermediate hosts in this case is to use a SSH agent and have it hold the private keys used for authentication.

Since the OpenSSH SSH client is used, you can also use the SSH askpass mechanism to handle these authentications in an interactive manner.

It might happen on slow networks that connections used for remote profiling through SSH time out. If this is the case, the ConnectTimeout option can be used to set the desired timeout value.

A known limitation of the remote profiling through SSH is that problems may arise if NVIDIA Nsight Compute tries to do remote profiling through SSH by connecting to the same machine it is running on. In this case, the workaround is to do local profiling through localhost.

For more information about available options for the OpenSSH client and the ecosystem of tools it can be used with for authentication refer to the official manual pages.

The Interactive Profile activity allows you to initiate a session that controls the execution of the target application, similar to a debugger. You can step API calls and workloads (CUDA kernels), pause and resume, and interactively select the kernels of interest and which metrics to collect.

This activity does currently not support profiling or attaching to child processes.

• Enable CPU Call Stack

  Collect the CPU-sided Call Stack at the location of each profiled kernel launch.

• Enable NVTX Support

  Collect NVTX information provided by the application or its libraries. Required to support stepping to specific NVTX contexts.

• Disable Profiling Start/Stop

  Ignore calls to cu(da)ProfilerStart or cu(da)ProfilerStop made by the application.

• Enable Profiling From Start

  Enables profiling from the application start. Disabling this is useful if the application calls cu(da)ProfilerStart and kernels before the first call to this API should not be profiled. Note that disabling this does not prevent you from manually profiling kernels.

• Cache Control

  Control the behavior of the GPU caches during profiling. Allowed values: For Flush All, all GPU caches are flushed before each kernel replay iteration during profiling. While metric values in the execution environment of the application might be slightly different without invalidating the caches, this mode offers the most reproducible metric results across the replay passes and also across multiple runs of the target application.

  For Flush None, no GPU caches are flushed during profiling. This can improve performance and better replicates the application behavior if only a single kernel replay pass is necessary for metric collection. However, some metric results will vary depending on prior GPU work, and between replay iterations. This can lead to inconsistent and out-of-bounds metric values.

• Clock Control

  Control the behavior of the GPU clocks during profiling. Allowed values: For Base, GPC and memory clocks are locked to their respective base frequency during profiling. This has no impact on thermal throttling. For None, no GPC or memory frequencies are changed during profiling.

• Import Source

  Enables permanently importing available source files into the report. Source information must be embedded in the executable, e.g. via the -lineinfo compiler option. Imported files are used in the CUDA-C view on the Source Page.

The Profile activity provides a traditional, pre-configurable profiler. After configuring which kernels to profile, which metrics to collect, etc, the application is run under the profiler without interactive control. The activity completes once the application terminates. For applications that normally do not terminate on their own, e.g. interactive user interfaces, you can cancel the activity once all expected kernels are profiled.

This activity does not support attaching to processes previously launched via NVIDIA Nsight Compute. These processes will be shown grayed out in the Attach tab.

• Output File

  Path to report file where the collected profile should be stored. If not present, the report extension .ncu-rep is added automatically. The placeholder %i is supported for the filename component. It is replaced by a sequentially increasing number to create a unique filename. This maps to the --export command line option.

• Force Overwrite

  If set, existing report file are overwritten. This maps to the --force-overwrite command line option.

• Target Processes

  Select the processes you want to profile. In mode Application Only, only the root application process is profiled. In mode all, the root application process and all its child processes are profiled. This maps to the --target-processes command line option.

• Replay Mode

  Select the method for replaying kernel launches multiple times. In mode Kernel, individual kernel launches are replayed transparently during the single execution of the target application. In mode Application, the entire target application is relaunched multiple times. In each iteration, additional data for the target kernel launches is collected. Application replay requires the program execution to be deterministic. This maps to the --replay-mode command line option. See the Kernel Profiling Guide for more details on the replay modes.

• Additional Options

  All remaining options map to their command line profiler equivalents. See the Command Line Options for details.

Entries in the connection dialog are saved as part of the current project. When working in a custom project, simply close the project to reset the dialog.

When not working in a custom project, entries are stored as part of the default project. You can delete all information from the default project by closing NVIDIA Nsight Compute and then deleting the project file from disk.

• File
  • New Project Create new profiling Projects with the New Project Dialog
  • Open Project Open an existing profiling project
  • Recent Projects Open an existing profiling project from the list of recently used projects
  • Save Project Save the current profiling project
  • Save Project As Save the current profiling project with a new filename
  • Close Project Close the current profiling project
  • New File Create a new file
  • Open File Open an existing file
  • Save Save the current file
  • Save As Save a copy of the current file with a different name or type or in a different location
  • Save All Files Save all open files
  • Close Close the current file
  • Close All Files Close all open files
  • Recent Files Open an existing file from the list of recently used files
  • Exit Exit Nsight Compute
• Connection
  • Connect Open the Connection Dialog to launch or attach to a target application. Disabled when already connected.
  • Disconnect Disconnect from the current target application, allows the application to continue normally and potentially re-attach.
  • Terminate Disconnect from and terminate the current target application immediately.
• Debug
  • Pause Pause the target application at the next intercepted API call or launch.
  • Resume Resume the target application.
  • Step In Step into the current API call or launch to the next nested call, if any, or the subsequent API call, otherwise.
  • Step Over Step over the current API call or launch and suspend at the next, non-nested API call or launch.
  • Step Out Step out of the current nested API call or launch to the next, non-parent API call or launch one level above.
  • Freeze API

    When disabled, all CPU threads are enabled and continue to run during stepping or resume, and all threads stop as soon as at least one thread arrives at the next API call or launch. This also means that during stepping or resume the currently selected thread might change as the old selected thread makes no forward progress and the API Stream automatically switches to the thread with a new API call or launch. When enabled, only the currently selected CPU thread is enabled. All other threads are disabled and blocked.

    Stepping now completes if the current thread arrives at the next API call or launch. The selected thread never changes. However, if the selected thread does not call any further API calls or waits at a barrier for another thread to make progress, stepping may not complete and hang indefinitely. In this case, pause, select another thread, and continue stepping until the original thread is unblocked. In this mode, only the selected thread will ever make forward progress.

  • Break On API Error When enabled, during resume or stepping, execution is suspended as soon as an API call returns an error code.
  • Run to Next Kernel See API Stream tool window.
  • Run to Next API Call See API Stream tool window.
  • Run to Next Range Start See API Stream tool window.
  • Run to Next Range End See API Stream tool window.
  • API Statistics Opens the API Statistics tool window
  • API Stream Opens the API Stream tool window
  • Resources Opens the Resources tool window
  • NVTX Opens the NVTX tool window
• Profile
  • Profile Kernel When suspended at a kernel launch, select the profile using the current configuration.
  • Profile Series When suspended at a kernel launch, open the Profile Series configuration dialog to setup and collect a series of profile results.
  • Auto Profile Enable or disable auto profiling. If enabled, each kernel matching the current kernel filter (if any) will be profiled using the current section configuration.
  • Clear Baselines Clear all current baselines.
  • Import Source Permanently import resolved source files into the report. Existing content may be overwritten.
  • Section/Rules Info Opens the Sections/Rules Info tool window.
• Tools
  • Project Explorer Opens the Project Explorer tool window.
  • Output Messages Opens the Output Messages tool window.
  • Options Opens the Options dialog.
• Window
  • Save Window Layout Allows you to specify a name for the current layout. The layouts are saved to a Layouts folder in the documents directory as named ".nvlayout" files.
  • Apply Window Layout Once you have saved a layout, you can restore them by using the "Apply Window Layout" menu entry. Simply select the entry from sub-menu you want to apply.
  • Manage Window Layout Allows you to delete or rename old layouts.
  • Restore Default Layout Restore views to their original size and position.
  • Show Welcome Page Opens the Welcome Page.
• Help
  • Documentation Opens the latest documentation for NVIDIA Nsight Compute online.
  • Documentation (local) Opens the local HTML documentation for NVIDIA Nsight Compute that has shipped with the tool.
  • Check For Updates Checks online if a newer version of NVIDIA Nsight Compute is available for download.
  • Reset Application Data Reset all NVIDIA Nsight Compute configuration data saved on disk, including option settings, default paths, recent project references etc. This will not delete saved reports.
  • Send Feedback Opens a dialog that allows you to send bug reports and suggestions for features. Optionally, the feedback includes basic system information, screenshots, or additional files (such as profile reports).
  • About Opens the About dialog with information about the version of NVIDIA Nsight Compute.

The main toolbar shows commonly used operations from the main menu. See Main Menu for their description.

Status banners are used to display important messages, such as profiler errors. The message can be dismissed by clicking the 'X' button. The number of banners shown at the same time is limited and old messages can get dismissed automatically if new ones appear. Use the Output Messages window to see the complete message history.

The API Statistics window is available when NVIDIA Nsight Compute is connected to a target application. It opens by default as soon as the connection is established. It can be re-opened using Debug > API Statistics from the main menu.

Whenever the target application is suspended, it shows a summary of tracked API calls with some statistical information, such as the number of calls, their total, average, minimum and maximum duration. Note that this view cannot be used as a replacement for Nsight Systems when trying to optimize CPU performance of your application.

The Reset button deletes all statistics collected to the current point and starts a new collection. Use the Export to CSV button to export the current statistics to a CSV file.

The API Stream window is available when NVIDIA Nsight Compute is connected to a target application. It opens by default as soon as the connection is established. It can be re-opened using Debug > API Stream from the main menu.

Whenever the target application is suspended, the window shows the history of API calls and traced kernel launches. The currently suspended API call or kernel launch (activity) is marked with a yellow arrow. If the suspension is at a subcall, the parent call is marked with a green arrow. The API call or kernel is suspended before being executed.

For each activity, further information is shown such as the kernel name or the function parameters (Func Parameters) and return value (Func Return). Note that the function return value will only become available once you step out or over the API call.

Use the Current Thread dropdown to switch between the active threads. The dropdown shows the thread ID followed by the current API name. One of several options can be chosen in the trigger dropdown, which are executed by the adjacent >> button. Run to Next Kernel resumes execution until the next kernel launch is found in any enabled thread. Run to Next API Call resumes execution until the next API call matching Next Trigger is found in any enabled thread. Run to Next Range Start resumes execution until the next start of an active profiler range is found. Profiler ranges are defined by using the cu(da)ProfilerStart/Stop API calls. Run to Next Range Stop resumes execution until the next stop of an active profiler range is found. The API Level dropdown changes which API levels are shown in the stream. The Export to CSV button exports the currently visible stream to a CSV file.

The Baselines tool window can be opened by clicking the Baselines entry in the Profile menu. It provides a centralized place from which to manage configured baselines. (Refer to Baselines, for information on how to create baselines from profile results.)

The baseline visibility can be controlled by clicking on the check box in a table row. When the check box is checked, the baseline will be visible in the summary header as well as all graphs in all sections. When unchecked the baseline will be hidden and will not contribute to metric difference calculations.

The baseline color can be changed by double-clicking on the color swatch in the table row. The color dialog which is opened provides the ability to choose an arbitrary color as well as offers a palette of predefined colors associated with the stock baseline color rotation.

The baseline name can be changed by double-clicking on the Name column in the table row. The name must not be empty and must be less than the Maximum Baseline Name Length as specified in the options dialog.

The z-order of a selected baseline can be changed by clicking the Move Baseline Up and Move Baseline Down buttons in the tool bar. When a baseline is moved up or down its new position will be reflected in the report header as well as in each graph. Currently, only one baseline may be moved at a time.

The selected baselines may be removed by clicking on the Clear Selected Baselines button in the tool bar. All baselines can be removed at once by clicking on the Clear All Baselines button, from either the global tool bar or the tool window tool bar.

The configured baselines can be saved to a file by clicking on the Save Baselines button in the tool bar. By default baseline files use the .ncu-bln extension. Baseline files can be opened locally and/or shared with other users.

Baseline information can be loaded by clicking on the Load Baselines button in the tool bar. When a baseline file is loaded, currently configured baselines will be replaced. A dialog will be presented to the user to confirm this operation when necessary.

The NVTX window is available when NVIDIA Nsight Compute is connected to a target application. If closed, it can be re-opened using Debug > NVTX from the main menu. Whenever the target application is suspended, the window shows the state of all active NVTX domains and ranges in the currently selected thread. Note that NVTX information is only tracked if the launching command line profiler instance was started with --nvtx or NVTX was enabled in the NVIDIA Nsight Compute launch dialog.

Use the Current Thread dropdown in the API Stream window to change the currently selected thread. NVIDIA Nsight Compute supports NVTX named resources, such as threads, CUDA devices, CUDA contexts, etc. If a resource is named using NVTX, the appropriate UI elements will be updated.

The Resources window is available when NVIDIA Nsight Compute is connected to a target application. It shows information about the currently known resources, such as CUDA devices, CUDA streams or kernels. The window is updated every time the target application is suspended. If closed, it can be re-opened using Debug > Resources from the main menu.

Using the dropdown on the top, different views can be selected, where each view is specific to one kind of resource (context, stream, kernel, …). The Filter edit allows you to create filter expressions using the column headers of the currently selected resource.

The resource table shows all information for each resource instance. Each instance has a unique ID, the API Call ID when this resource was created, its handle, associated handles, and further parameters. When a resource is destroyed, it is removed from its table.

The Sections/Rules Info window can be opened from the main menu using Profile > Sections/Rules Info. It tracks all section sets, sections and rules currently loaded in NVIDIA Nsight Compute, independent from a specific connection or report. The directory to load those files from can be configured in the Profile options dialog. It is used to inspect available sets, sections and rules, as well as to configure which should be collected, and which rules should be applied. The window has two views, which can be selected using the dropdown in its header.

The Section Sets view shows all available section sets. Each set is associated with a number of sections. You can choose a set appropriate to the level of detail for which you want to collect performance metrics. Sets which collect more detailed information normally incur higher runtime overhead during profiling.

When enabling a set in this view, the associated sections are enabled in the Sections/Rules view. When disabling a set in this view, the associated sections in the Sections/Rules view are disabled. If no set is enabled, or if sections are manually enabled/disabled in the Sections/Rules view, the <custom> entry is marked active to represent that no section set is currently enabled. Note that the default set is enabled by default.

Whenever a kernel is profiled manually, or when auto-profiling is enabled, only sections enabled in the Sections/Rules view are collected. Similarly, whenever rules are applied, only rules enabled in this view are active.

The enabled states of sections and rules are persisted across NVIDIA Nsight Compute launches. The Reload button reloads all sections and rules from disk again. If a new section or rule is found, it will be enabled if possible. If any errors occur while loading a rule, they will be listed in an extra entry with a warning icon and a description of the error.

Use the Enable All and Disable All checkboxes to enable or disable all sections and rules at once. The Filter text box can be used to filter what is currently shown in the view. It does not alter activation of any entry.

The table shows sections and rules with their activation status, their relationship and further parameters, such as associated metrics or the original file on disk. Rules associated with a section are shown as children of their section entry. Rules independent of any section are shown under an additional Independent Rules entry.

Double-clicking an entry in the table's Filename column opens this file as a document. It can be edited and saved directly in NVIDIA Nsight Compute. After editing the file, Reload must be selected to apply those changes.

When a section or rule file is modified, the entry in the State column will show User Modified to reflect that it has been modified from its default state. When a User Modified row is selected, the Restore button will be enabled. Clicking the Restore button will restore the entry to its default state and automatically Reload the sections and rules.

Similarly, when a stock section or rule file is removed from the configured Sections Directory (specified in the Profile options dialog), the State column will show User Deleted. User Deleted files can also be restored using the Restore button.

Section and rule files that are created by the user (and not shipped with NVIDIA Nsight Compute) will show up as User Created in the state column.

See the Sections and Rules for the list of default sections for NVIDIA Nsight Compute.

The Page dropdown can be used to switch between the available report pages, which are explained in detail in the next section.

Profiler report header

The Launch dropdown can be used to switch between all collected kernel launches. The information displayed in each page commonly represents the selected launch instance. On some pages (e.g. Raw), information for all launches is shown and the selected instance is highlighted. You can type in this dropdown to quickly filter and find a kernel launch.

The Apply Filters button open the filter dialog. You can use more than one filter to narrow down your results. On the filter dialog, enter your filter parameters and press OK button. The Launch dropdown will be filtered accordingly. Select the arrow dropdown to access the Clear Filters button, which removes all filters.

Filter Dialog

The Add Baseline button promotes the current result in focus to become the baseline of all other results from this report and any other report opened in the same instance of NVIDIA Nsight Compute. Select the arrow dropdown to access the Clear Baselines button, which removes all currently active baselines.

The Apply Rules button applies all rules available for this report. If rules had been applied previously, those results will be replaced. By default, rules are applied immediately once the kernel launch has been profiled. This can be changed in the options under Tools > Options > Profile > Report UI > Apply Applicable Rules Automatically.

Note that not all functions are available on all pages.

Information about the selected kernel is shown as Current. [+] and [-] buttons can be used to show or hide the section body content. The visibility of the output of the rules can be toggled with the r button. The info toggle button i changes the section description's visibility.

Use the Page dropdown in the header to switch between the report pages.

Numeric metric values are shown in various places in the report, including the header and tables and charts on most pages. NVIDIA Nsight Compute supports various ways to display those metrics and their values.

When available and applicable to the UI component, metrics are shown along with their unit. This is to make it apparent if a metric represents cycles, threads, bytes/s, and so on. The unit will normally be shown in rectangular brackets, e.g. Metric Name [bytes] 128.

By default, units are scaled automatically so that metric values are shown with a reasonable order of magnitude. Units are scaled using their SI-factors, i.e. byte-based units are scaled using a factor of 1000 and the prefixes K, M, G, etc. Time-based units are also scaled using a factor of 1000, with the prefixes n, u and m. This scaling can be disabled in the Profile options.

Metrics which could not be collected are shown as n/a and assigned a warning icon. If the metric floating point value is out of the regular range (i.e. nan (Not a number) or inf (infinite)), they are also assigned a warning icon. The exception are metrics for which these values are expected and which are white-listed internally.

The tables show the occupancy, as well as the number of active threads, warps, and thread blocks per multiprocessor, and the maximum number of active blocks on the GPU.

Tables

The graphs show the occupancy for your chosen block size as a blue circle, and for all other possible block sizes as a line graph.

Graphs

The GPU Data shows the properties of all supported devices.

Graphs

Connection properties are grouped into Target Connection Options and Host Connection Properties.

• New Project

  Creates a new project. The project must be given a name, which will also be used for the project file. You can select the location where the project file should be saved on disk. Select whether a new directory with the project name should be created in that location.

The Project Explorer window allows you to inspect and manage the current project. It shows the project name as well as all Items (profile reports and other files) associated with it. Right-click on any entry to see further actions, such as adding, removing or grouping items. Type in the Search project toolbar at the top to filter the currently shown entries.

Project Explorer

NVIDIA Nsight Compute does not support tracing GPU or API activities on an accurate timeline. This functionality is covered by NVIDIA Nsight Systems. In the Interactive Profile Activity, the API Stream tool window provides a stream of recent API calls on each thread. However, since all tracked API calls are serialized by default, it does not collect accurate timestamps.

Instead of sessions, NVIDIA Nsight Compute uses Projects to launch and gather connection details and collected reports.

• Executable and Import Sessions

  Use the Project Explorer or the Main Menu to create a new project. Reports collected from the command line, i.e. using NVIDIA Nsight Compute CLI, can be opened directly using the main menu. In addition, you can use the Project Explorer to associate existing reports as well as any other artifacts such as executables, notes, etc., with the project. Note that those associations are only references; in other words, moving or deleting the project file on disk will not update its artifacts.

  nvprof or command-line profiler output files, as well as Visual Profiler sessions, cannot be imported into NVIDIA Nsight Compute.

Since trace analysis is now covered by Nsight Systems, NVIDIA Nsight Compute does not provide views of the application timeline. The API Stream tool window does show a per-thread stream of the last captured CUDA API calls. However, those are serialized and do not maintain runtime concurrency or provide accurate timing information.

• Guided Analysis

  All trace-based analysis is now covered by NVIDIA Nsight Systems. This means that NVIDIA Nsight Compute does not include analysis regarding concurrent CUDA streams or (for example) UVM events. For per-kernel analysis, NVIDIA Nsight Compute provides recommendations based on collected performance data on the Details Page. These rules currently require you to collect the required metrics via their sections up front, and do not support partial on-demand profiling.

  To use the rule-based recommendations, enable the respective rules in the Sections/Rules Info. Before profiling, enable Apply Rules in the Profile Options, or click the Apply Rules button in the report afterward.

• Unguided Analysis

  All trace-based analysis is now covered by Nsight Systems. For per-kernel analysis, Python-based rules provide analysis and recommendations. See Guided Analysis above for more details.

• PC Sampling View

  Source-correlated PC sampling information can now be viewed in the Source Page. Aggregated warp states are shown on the Details Page in the Warp State Statistics section.

• Memory Statistics

  Memory Statistics are located on the Details Page. Enable the Memory Workload Analysis sections to collect the respective information.

• NVLink View

  NVLink topology diagram and NVLink property table are located on the Details Page. Enable the NVLink Topology and NVLink Table sections to collect the respective information.

  Refer to the Known Issues section for the limitations related to NVLink.

• Source-Disassembly View

  Source correlated with PTX and SASS disassembly is shown on the Source Page. Which information is available depends on your application's compilation/JIT flags.

• GPU Details View

  NVIDIA Nsight Compute does not automatically collect data for each executed kernel, and it does not collect any data for device-side memory copies. Summary information for all profiled kernel launches is shown on the Summary Page. Comprehensive information on all collected metrics for all profiled kernel launches is shown on the Raw Page.

• CPU Details View

  CPU callstack sampling is now covered by NVIDIA Nsight Systems.

• OpenACC Details View

  OpenACC performance analysis with NVIDIA Nsight Compute is available to limited extent. OpenACC parallel regions are not explicitly recognized, but CUDA kernels generated by the OpenACC compiler can be profiled as regular CUDA kernels. See the NVIDIA Nsight Systems release notes to check its latest support status.

• OpenMP Details View

  OpenMP performance analysis is not supported by NVIDIA Nsight Compute. See the NVIDIA Nsight Systems release notes to check its latest support status.

• Properties View

  NVIDIA Nsight Compute does not collect CUDA API and GPU activities and their properties. Performance data for profiled kernel launches is reported (for example) on the Details Page.

• Console View

  NVIDIA Nsight Compute does not currently collect stdout/stderr application output.

• Settings View

  Application launch settings are specified in the Connection Dialog. For reports collected from the UI, launch settings can be inspected on the Session Page after profiling.

• CPU Source View

  Source for CPU-only APIs is not available. Source for profiled GPU kernel launches is shown on the Source Page.

Please execute ncu-ui with the -h parameter within a shell window to see the currently supported command line arguments for the NVIDIA Nsight Compute UI.

To open a collected profile report with ncu-ui, simply pass the path to the report file as a parameter to the shell command.

NVIDIA Nsight Integration is a Visual Studio extension that allows you to access the power of NVIDIA Nsight Compute from within Visual Studio.

When NVIDIA Nsight Compute is installed along with NVIDIA Nsight Integration, NVIDIA Nsight Compute activities will appear under the NVIDIA 'Nsight' menu in the Visual Studio menu bar. These activities launch NVIDIA Nsight Compute with the current project settings and executable.

NVIDIA Nsight Compute supports profiling of OptiX applications, but with certain restrictions.

• Internal Kernels

  Kernels launched by OptiX that contain no user-defined code are given the generic name NVIDIA internal. These kernels show up on the API Stream in the NVIDIA Nsight Compute UI, and can be profiled in both the UI as well as the NVIDIA Nsight Compute CLI. However, no CUDA-C source, PTX or SASS is available for them.

• User Kernels

  Kernels launched by OptiX can contain user-defined code. OptiX identifies these kernels in the API Stream with a custom name. This name starts with raygen__ (for "ray generation"). These kernels show up on the API Stream and can be profiled in the UI as well as the NVIDIA Nsight Compute CLI. The Source page displays CUDA-C source, PTX and SASS defined by the user. Certain parts of the kernel, including device functions that contain OptiX-internal code, will not be available in the Source page.

• SASS

  When SASS information is available in the profile report, certain instructions might not be available in the Source page and shown as N/A.

The following feature set is supported per OptiX API version: