Enhance Data Analysis in VTune Profiler with Custom Data Collectors

Table of Contents

• Introduction
• Configuring the Custom Data Collector
• Starting the Data Collection
• Integrating Custom Data into VTune Profiler Results
• Viewing Custom Data in the Timeline
• Anatomy of Custom Data Files
• Generating Timestamps for Custom Data
• Writing Data Types to the Custom Data File
• Adding Custom Data to VTune Profiler Without Interaction
• Interval Custom Data
• Conclusion

Introduction

Welcome to this Tutorial on using Intel V2 Profiler with a custom data collector to enable customized analysis. In this tutorial, we will walk you through the process of configuring the custom data collector, starting the data collection, integrating custom data into VTune Profiler results, and more. By the end of this tutorial, you will have a solid understanding of how to use custom data in VTune Profiler to enhance your data analysis.

Configuring the Custom Data Collector

To begin, we will first configure the custom data collector. This involves selecting an application, process ID, or enabling system-wide collection. We will also explore the advanced options, specifically the custom collector text box. In this example, we will be using a batch file, but other options such as compiled programs or Python scripts are also possible. The batch file will be called by VTune Profiler, setting environment variables including a collect command argument to start, stop, pause, or Resume collection. We will focus on implementing the start and stop options in the batch file.

Starting the Data Collection

Once the custom data collector is configured, we can start the data collection. When the VTune Profiler data collection is initiated, an output window will appear for the application and each custom data collector instance. We will see how to start multiple instances of the custom data collector, each collecting different metrics.

Integrating Custom Data into VTune Profiler Results

After the data collection is completed, the custom data collector script is called to stop each data collector instance. VTune Profiler seamlessly integrates the custom data into the analysis results during the finalization phase of the collection. We will explore the integration of custom data and how it adds valuable context to the analysis.

Viewing Custom Data in the Timeline

To analyze the custom data, we will switch to the bottom-up tab in VTune Profiler. Here, we can view the custom data in the timeline. We will examine how different custom data counters are reported as counter rates or Instant values. Interacting with the custom data in the timeline is similar to other metrics, such as hovering over the timeline to see the custom data values at specific times.

Anatomy of Custom Data Files

Now, let's take a closer look at the custom data files themselves. The file names can start with any user-defined STRING but must end with the keyword "hostname" followed by the hostname of the system under test and the ".csv" extension. The header of the custom data file indicates the metric name and its data type. We will explore examples of custom data file names and their content.

Generating Timestamps for Custom Data

Each data line in the custom data file must start with a timestamp to map it to the other data collected at that time. We will provide an example implementation of generating timestamps for each data line using time and get system time calls.

Writing Data Types to the Custom Data File

Custom data files can include different data types such as instantaneous values and counter rate values. We will demonstrate an example implementation of writing these discrete data types to the custom data file.

Adding Custom Data to VTune Profiler Without Interaction

In previously demonstrated interactions, we showed how custom data can be collected through the VTune Profiler. However, there may be cases where custom data has been collected separately. In this tutorial, we will demonstrate how to add custom data files directly to the VTune Profiler results data directory. We will then instruct VTune Profiler to re-resolve the data, including the newly added custom data files.

Interval Custom Data

VTune Profiler also supports interval custom data. After finalization, the interval custom data appears in the frame rate swim lane of the timeline. We will explore how to view and analyze overlapping interval custom data in the timeline.

Conclusion

In conclusion, this tutorial has provided a comprehensive guide on using custom data in VTune Profiler. We have covered configuring the custom data collector, starting the data collection, integrating custom data into results, viewing custom data in the timeline, generating timestamps, writing data types, adding custom data without interaction, and interval custom data. By leveraging custom data, you can gain additional context and enhance your data analysis in VTune Profiler.

Highlights

• Learn how to configure the custom data collector in VTune Profiler.
• Explore different options for starting the data collection.
• Understand how custom data is integrated into VTune Profiler results.
• View and analyze custom data in the timeline.
• Gain insights into the anatomy of custom data files.
• Implement timestamp generation for custom data.
• Write different data types to the custom data file.
• Add custom data directly to VTune Profiler results without interaction.
• Discover interval custom data and its analysis in the timeline.

FAQ

Q: Can I use a compiled program instead of a batch file for the custom data collector? A: Yes, you can use either a compiled program or a Python script as the custom data collector.

Q: How can I interact with the custom data in the timeline? A: Similar to other metrics, you can hover over the timeline to see the custom data values at specific times.

Q: What is the naming convention for custom data files? A: Custom data file names can start with any user-defined string but must end with the keyword "hostname" followed by the hostname of the system under test and the ".csv" extension.

Q: Can I add custom data to VTune Profiler results without interacting through the profiler? A: Yes, you can add custom data files directly to the VTune Profiler results data directory and instruct VTune Profiler to re-resolve the data.

Q: Does VTune Profiler support interval custom data? A: Yes, VTune Profiler supports interval custom data, which appears in the frame rate swim lane of the timeline after finalization.