Mastering Performance Analysis With Perf

Mastering Performance Analysis With Perf
Published 10/2024
MP4 | Video: h264, 1920x1080 | Audio: AAC, 44.1 KHz
Language: English | Size: 2.07 GB | Duration: 6h 18m

Deep Dive into Linux Performance Profiling and Analysis

What you'll learn

Understand the basics of perf - the official Linux performance tool

Learn microarchitectural performance analysis

Profile system and application performance

Create custom performance analyses

Requirements

Familiar with Linux operating systems

No performance engineering experience needed

Description

Start with the fundamentals of performance analysis using perf stat, where you'll learn how to monitor key performance metrics like CPU utilization, cache performance, and context switches. These are essential to a high-level understanding of your application and system performance.From there, you'll advance to Top-Down Microarchitecture Analysis (TMA), a method that helps you drill down into CPU inefficiencies by categorizing execution into retiring, bad speculation, backend bound, and frontend bound. This approach will allow you to precisely identify CPU performance issues.Next, you'll dive into detailed profiling with tools like perf record, perf report, and perf top to analyze workloads and uncover performance bottlenecks. You'll learn techniques like flame graphs for visualizing code execution, off-CPU analysis to detect wait times and latencies, and real-time monitoring techniques to address critical performance issues as they arise.Finally, you'll learn dynamic instrumentation with perf probe, which allows you to insert probes into running code without needing to restart applications. Combined with perf script, you'll learn how to write custom scripts that track specific events, offering deeper insights into application behavior and enabling more tailored performance analyses.By the end of this course, you'll have the skills to identify performance issues and optimize the performance of your applications and systems with confidence.

Overview

Section 1: Get Started

Lecture 1 What You'll Learn

Lecture 2 Setup and Installation

Section 2: Basic Performance Analysis

Lecture 3 Overview

Lecture 4 What Is Software Performance?

Lecture 5 Introduction to perf stat

Lecture 6 Matrix Multiplication

Lecture 7 Monitored Targets

Lecture 8 Performance Events

Lecture 9 Counting Hardware Events

Lecture 10 CPU Utilization

Lecture 11 CPU Benchmark

Lecture 12 Context Switch and CPU Migration

Lecture 13 Page Faults

Lecture 14 Hardware Events

Lecture 15 Symbolic Event Names

Lecture 16 Measuring Energy Consumption

Lecture 17 Summary

Section 3: Microarchitectural Performance Analysis

Lecture 18 Overview

Lecture 19 Understanding the Top-Down Microarchitecture Analysis Method

Lecture 20 Exploring the TMA Metric Hierarchy

Lecture 21 Implementing TMA Metrics with perf stat

Lecture 22 Vector Instructions (Retiring)

Lecture 23 Branch Misprediction (Bad Speculation)

Lecture 24 Memory Bound Issue (Backend Bound)

Lecture 25 False Sharing (Backend Bound)

Lecture 26 Instruction Cache (Frontend Bound)

Lecture 27 Summary

Section 4: Software Performance Profiling

Lecture 28 Overview

Lecture 29 Software Performance Profiling

Lecture 30 Performance Profiling with perf record and perf report

Lecture 31 Real-time Monitoring with perf top

Lecture 32 Flame Graphs

Lecture 33 Workload Characteristics Analysis on Memcached

Lecture 34 Custom Flame Graphs

Lecture 35 Tracepoints

Lecture 36 Off-CPU Analysis

Lecture 37 Summary

Section 5: Dynamic Instrumentation and Scripting

Lecture 38 Overview

Lecture 39 Dynamic Instrumentation with perf probe

Lecture 40 Catching Crash Points

Lecture 41 Debugging Memcached

Lecture 42 Introduction to perf script

Lecture 43 Function Latency Analysis

Lecture 44 Amdahl's Law

Lecture 45 Summary

Section 6: Bonus

Lecture 46 A Practical Approach

Software engineers working on Linux-based systems need to understand how their applications perform,Performance engineers or anyone specifically focused on the performance of software and systems