Parallel Programming with Intel Parallel Studio XE -- Contents -- Foreword -- Introduction -- Part I: An Introduction to Parallelism -- Chapter 1: Parallelism Today -- The Arrival of Parallelism -- The Power Density Race -- The Emergence of Multi-Core and Many-Core Computing -- The Top Six Challenges -- Legacy Code -- Tools -- Education -- Fear of Many-Core Computing -- Maintainability -- Return on Investment -- Parallelism and the Programmer -- Types of Parallelism -- Intel's Family of Parallel Models -- Cilk Plus and Threading Building Blocks -- Domain-Specific Libraries -- Established Standards -- Research and Development -- Choosing the Right Parallel Constructs -- High-Level vs. Low-Level Constructs -- Data Parallelism vs. General Parallelism -- Examples of Mixing and Matching Parallel Constructs -- Parallel Programming Errors -- Data Races -- Determinacy Races -- Deadlocks -- Poor Load Balancing -- Threading/Tasking Overhead -- Synchronization Overhead -- Memory Errors -- Speedup and Scalability -- Calculating Speedup -- Predicting Scalability -- Parallelism and Real-Time Systems -- Hard and Soft Real-Time -- A Hard Real-Time Example using RTX -- Advice for Real-Time Programmers -- Summary -- Chapter 2: An Overview of Parallel Studio XE -- Why Parallel Studio XE? -- What's in Parallel Studio XE? -- Intel Parallel Studio XE -- Intel Parallel Advisor -- The Advisor Workflow -- Surveying the Site -- Annotating Code -- Checking Suitability -- Checking Correctness -- Replacing Annotations -- Intel Parallel Composer XE -- Intel C/C++ Optimizing Compiler -- Profile-Guided Optimization -- Cilk Plus -- OpenMP -- Intel Threading Building Blocks -- Intel Integrated Performance Primitives -- An Application Example -- IPP and Threading -- Intel Parallel Debugger Extension -- Intel Debugger -- Math Kernel Library.

VTune Amplifier XE -- Hotspot Analysis -- Concurrency Analysis -- Locks and Waits Analysis -- Dissassembly Source View -- Parallel Inspector XE -- Predefined Analysis Types -- Errors and Warnings -- Static Security Analysis -- Different Approaches to Using Parallel Studio XE -- Summary -- Chapter 3: Parallel Studio XE for the Impatient -- The Four-Step Methodology -- Example 1: Working with Cilk Plus -- Obtaining a Suitable Serial Program -- Running the Serial Example Program -- Creating the Project -- Running the Serial Version of the Code -- Step 1: Analyze the Serial Program -- Using Intel Parallel Amplifier XE for Hotspot Analysis -- Step 2: Implement Parallelism using Cilk Plus -- Step 3: Debug and Check for Errors -- Checking for Errors -- Narrowing the Scope of the Shared Variables -- Adding Cilk Plus Reducers -- Running the Corrected Application -- Step 4: Tune the Cilk Plus Program -- Example 2: Working with OpenMP -- Step 1: Analyze the Serial Program -- Step 2: Implement Parallelism using OpenMP -- Step 3: Debug and Check for Errors -- Making the Shared Variables Private -- Adding a Reduction Clause -- Step 4: Tune the OpenMP Program -- Improving the Load Balancing -- Summary -- Part II Using Parallel Studio XE -- Chapter 4: Producing Optimized Code -- Introduction -- The Example Application -- Optimizing Code in Seven Steps -- Using the Compiler's Reporting Features -- Step 1: Build with Optimizations Disabled -- Step 2: Use General Optimizations -- Using the General Options on the Example Application -- Generating Optimization Reports Using /Qopt-report -- Step 3: Use Processor-Specific Optimizations -- What Is Auto-Vectorization? -- Auto-Vectorization Guidelines -- Turning On Auto-Vectorization -- Enhancing Auto-Vectorization -- Building for Non-Intel CPUs.

Determining That Auto-Vectorization Has Happened -- When Auto-Vectorization Fails -- Helping the Compiler to Vectorize -- Step 4: Add Interprocedural Optimization -- Adding Interprocedural Optimization to the Example Application -- The Impact of Interprocedural Optimization on Auto-Vectorization -- Step 5: Use Profile-Guided Optimization -- Benefits of Profile-Guided Optimization -- The Profile-Guided Optimization Steps -- The Results -- Step 6: Tune Auto-Vectorization -- Activating Guided Auto-Parallelization -- An Example Session -- More on Auto-Vectorization -- Building Applications to Run on More Than One Type of CPU -- Additional Ways to Insert Vectorization -- Using Cilk Plus Array Notation -- Manual CPU Dispatch: Rolling Your Own CPU-Specific Code -- Source Code -- Summary -- Chapter 5: Writing Secure Code -- A Simple Security Flaw Example -- Understanding Static Security Analysis -- False Positives -- Static Security Analysis Workflow -- Conducting a Static Security Analysis -- Investigating the Results of the Analysis -- Working with Problem States -- The Build Specification -- Creating a Build Specification File by Injection -- Utility Options -- The Directory Structure of the Results -- Using Static Security Analysis in a QA Environment -- Regression Testing -- Metrics Tracking -- Source Code -- Summary -- Chapter 6: Where to Parallelize -- Different Ways of Profiling -- The Example Application -- Hotspot Analysis Using the Intel Compiler -- Profiling Steps -- An Example Session -- Overhead Introduced by Profiling -- Hotspot Analysis Using the Auto-Parallelizer -- Profiling Steps -- An Example Session -- Programming Guidelines for Auto-Parallelism -- Additional Options -- Helping the Compiler to Auto-Parallelize -- Hotspot Analysis with Amplifier XE -- Conducting a Default Analysis.

Finding the Right Loop to Parallelize -- Large or Long-Running Applications -- Reducing the Size of Data Collected -- Using the Pause and Resume APIs -- Source Code -- Summary -- Chapter 7: Implementing Parallelism -- C or C++, That Is the Question -- Taking a Simple Approach -- The Beauty of Lambda Functions -- Parallelizing Loops -- The for Loop -- The Cilk Plus cilk_for Loop -- The OpenMP for Loop -- The TBB for Loop -- Nested for Loops -- The for Loop with Reduction -- Cilk Plus Reduction -- OpenMP Reduction -- TBB Reduction -- The while Loop -- Cilk Plus -- OpenMP -- TBB -- Parallelizing Sections and Functions -- The Serial Version -- Cilk Plus -- OpenMP -- TBB -- Parallelizing Recursive Functions -- The Serial Version -- Cilk Plus -- OpenMP -- TBB -- Parallelizing Pipelined Applications -- Parallel Pipelined Patterns -- The Serial Version -- OpenMP -- TBB -- Parallelizing Linked Lists -- Serial Iteration of the Linked List -- Parallel Iteration of the Linked List -- Source Code -- Summary -- Chapter 8: Checking for Errors -- Parallel Inspector XE Analysis Types -- Detecting Threading Errors -- Types of Threading Problems -- Thread Information -- Potential Privacy Infringement -- Data Races -- Deadlocks -- An Example Application Involving Deadlocks -- Detecting Deadlocks -- Detecting Data Races -- Running the Threaded Program -- First Results of the Analysis -- Controlling the Right Level of Detail -- Testing All the Code Paths -- Avoiding Being Overwhelmed by the Amount of Data -- Using Suppression Files -- Fixing Data Races -- Using Cilk Plus -- Cilk Plus Reducers -- Cilk Plus Holders -- Using OpenMP -- Using Locks -- Using Critical Sections -- Using Atomic Operations -- Using a reduction Clause -- Using TBB -- Detecting Memory Errors -- Types of Memory Errors.

An Example Application for Memory Analysis -- Creating a Custom Analysis -- The Source Code -- Summary -- Chapter 9: Tuning Parallel Applications -- Introduction -- Defining a Baseline -- Ensuring Consistency -- Measuring the Performance Improvements -- Measuring the Baseline Using the Amplifier XE Command Line -- Identifying Concurrency Hotspots -- Thread Concurrency and CPU Usage -- Identifying Hotspots in the Code -- Analyzing the Timeline -- Questions to Answer -- Fixing the Critical Section Hotspot -- Analyzing an Algorithm -- Conducting Further Analysis and Tuning -- Using Other Viewpoints -- Using Locks and Waits Analysis -- Other Analysis Types -- Using the Intel Software Autotuning Tool -- Source Code -- Summary -- Chapter 10: Parallel Advisor-Driven Design -- Using Parallel Advisor -- Understanding the Advisor Workflow -- Finding Documentation -- Getting Started with the NQueens Example Program -- Surveying the Site -- Running a Survey Analysis -- The Survey Report -- Finding Candidate Parallel Regions -- The Survey Source Window -- How Survey Analysis Works -- Annotating Your Code -- Site Annotations -- Lock Annotations -- Adding Annotations -- Checking Suitability -- Running a Suitability Analysis -- The Suitability Report -- Parallel Choices -- Using the Suitability Report -- How Suitability Analysis Works -- Checking for Correctness -- Running a Correctness Analysis -- The Correctness Report -- The Correctness Source Window -- Understanding Common Problems -- Using the Correctness Report -- Correctness Analysis Limitation -- How Correctness Analysis Works -- Replacing Annotations -- The Summary Report -- Common Mappings -- Summary -- Chapter 11: Debugging Parallel Applications -- Introduction to the Intel Debugger -- The Parallel Debugger Workflow.

Using the Intel Debugger to Detect Data Races.