SPRUI04F july   2015  – april 2023

 

  1.   Read This First
    1.     About This Manual
    2.     Notational Conventions
    3.     Related Documentation
    4.     Related Documentation From Texas Instruments
    5.     Trademarks
  2. 1 Introduction to the Software Development Tools
    1. 1.1 Software Development Tools Overview
    2. 1.2 Compiler Interface
    3. 1.3 ANSI/ISO Standard
    4. 1.4 Output Files
    5. 1.5 Utilities
  3. 2 Getting Started with the Code Generation Tools
    1. 2.1 How Code Composer Studio Projects Use the Compiler
    2. 2.2 Compiling from the Command Line
  4. 3 Using the C/C++ Compiler
    1. 3.1  About the Compiler
    2. 3.2  Invoking the C/C++ Compiler
    3. 3.3  Changing the Compiler's Behavior with Options
      1. 3.3.1  Linker Options
      2. 3.3.2  Frequently Used Options
      3. 3.3.3  Miscellaneous Useful Options
      4. 3.3.4  Run-Time Model Options
      5. 3.3.5  Selecting Target CPU Version (--silicon_version Option)
      6. 3.3.6  Symbolic Debugging and Profiling Options
      7. 3.3.7  Specifying Filenames
      8. 3.3.8  Changing How the Compiler Interprets Filenames
      9. 3.3.9  Changing How the Compiler Processes C Files
      10. 3.3.10 Changing How the Compiler Interprets and Names Extensions
      11. 3.3.11 Specifying Directories
      12. 3.3.12 Assembler Options
    4. 3.4  Controlling the Compiler Through Environment Variables
      1. 3.4.1 Setting Default Compiler Options (C6X_C_OPTION)
      2. 3.4.2 Naming One or More Alternate Directories (C6X_C_DIR)
    5. 3.5  Controlling the Preprocessor
      1. 3.5.1  Predefined Macro Names
      2. 3.5.2  The Search Path for #include Files
        1. 3.5.2.1 Adding a Directory to the #include File Search Path (--include_path Option)
      3. 3.5.3  Support for the #warning and #warn Directives
      4. 3.5.4  Generating a Preprocessed Listing File (--preproc_only Option)
      5. 3.5.5  Continuing Compilation After Preprocessing (--preproc_with_compile Option)
      6. 3.5.6  Generating a Preprocessed Listing File with Comments (--preproc_with_comment Option)
      7. 3.5.7  Generating Preprocessed Listing with Line-Control Details (--preproc_with_line Option)
      8. 3.5.8  Generating Preprocessed Output for a Make Utility (--preproc_dependency Option)
      9. 3.5.9  Generating a List of Files Included with #include (--preproc_includes Option)
      10. 3.5.10 Generating a List of Macros in a File (--preproc_macros Option)
    6. 3.6  Passing Arguments to main()
    7. 3.7  Understanding Diagnostic Messages
      1. 3.7.1 Controlling Diagnostic Messages
      2. 3.7.2 How You Can Use Diagnostic Suppression Options
    8. 3.8  Other Messages
    9. 3.9  Generating Cross-Reference Listing Information (--gen_cross_reference_listing Option)
    10. 3.10 Generating a Raw Listing File (--gen_preprocessor_listing Option)
    11. 3.11 Using Inline Function Expansion
      1. 3.11.1 Inlining Intrinsic Operators
      2. 3.11.2 Inlining Restrictions
      3. 3.11.3 Unguarded Definition-Controlled Inlining
        1. 3.11.3.1 Using the Inline Keyword
      4. 3.11.4 Guarded Inlining and the _INLINE Preprocessor Symbol
        1. 3.11.4.1 Header File string.h
        2. 3.11.4.2 Library Definition File
    12. 3.12 Interrupt Flexibility Options (--interrupt_threshold Option)
    13. 3.13 Using Interlist
    14. 3.14 Generating and Using Performance Advice
    15. 3.15 About the Application Binary Interface
    16. 3.16 Enabling Entry Hook and Exit Hook Functions
  5. 4 Optimizing Your Code
    1. 4.1  Invoking Optimization
    2. 4.2  Controlling Code Size Versus Speed
    3. 4.3  Performing File-Level Optimization (--opt_level=3 option)
      1. 4.3.1 Creating an Optimization Information File (--gen_opt_info Option)
    4. 4.4  Program-Level Optimization (--program_level_compile and --opt_level=3 options)
      1. 4.4.1 Controlling Program-Level Optimization (--call_assumptions Option)
      2. 4.4.2 Optimization Considerations When Mixing C/C++ and Assembly
    5. 4.5  Automatic Inline Expansion (--auto_inline Option)
    6. 4.6  Optimizing Software Pipelining
      1. 4.6.1 Turn Off Software Pipelining (--disable_software_pipeline Option)
      2. 4.6.2 Software Pipelining Information
        1. 4.6.2.1 Software Pipelining Information
        2. 4.6.2.2 Software Pipelining Information Terms
        3. 4.6.2.3 Loop Disqualified for Software Pipelining Messages
        4. 4.6.2.4 Pipeline Failure Messages
        5. 4.6.2.5 Register Usage Table Generated by the --debug_software_pipeline Option
      3. 4.6.3 Collapsing Prologs and Epilogs for Improved Performance and Code Size
        1. 4.6.3.1 Speculative Execution
        2. 4.6.3.2 Selecting the Best Threshold Value
    7. 4.7  Redundant Loops
    8. 4.8  Utilizing the Loop Buffer Using SPLOOP
    9. 4.9  Reducing Code Size (--opt_for_space (or -ms) Option)
    10. 4.10 Using Feedback Directed Optimization
      1. 4.10.1 Feedback Directed Optimization
        1. 4.10.1.1 Phase 1 -- Collect Program Profile Information
        2. 4.10.1.2 Phase 2 -- Use Application Profile Information for Optimization
        3. 4.10.1.3 Generating and Using Profile Information
        4. 4.10.1.4 Example Use of Feedback Directed Optimization
        5. 4.10.1.5 The .ppdata Section
        6. 4.10.1.6 Feedback Directed Optimization and Code Size Tune
        7. 4.10.1.7 Instrumented Program Execution Overhead
        8. 4.10.1.8 Invalid Profile Data
      2. 4.10.2 Profile Data Decoder
      3. 4.10.3 Feedback Directed Optimization API
      4. 4.10.4 Feedback Directed Optimization Summary
    11. 4.11 Using Profile Information to Get Better Program Cache Layout and Analyze Code Coverage
      1. 4.11.1 Background and Motivation
      2. 4.11.2 Code Coverage
        1. 4.11.2.1 Phase1 -- Collect Program Profile Information
        2. 4.11.2.2 Phase 2 -- Generate Code Coverage Reports
      3. 4.11.3 What Performance Improvements Can You Expect to See?
        1. 4.11.3.1 Evaluating L1P Cache Performance
      4. 4.11.4 Program Cache Layout Related Features and Capabilities
        1. 4.11.4.1 Path Profiler
        2. 4.11.4.2 Analysis Options
        3. 4.11.4.3 Environment Variables
        4. 4.11.4.4 Program Cache Layout Tool, clt6x
        5. 4.11.4.5 Linker
        6. 4.11.4.6 Linker Command File Operator unordered()
      5. 4.11.5 Program Instruction Cache Layout Development Flow
        1. 4.11.5.1 Gather Dynamic Profile Information
        2. 4.11.5.2 Generate Preferred Function Order from Dynamic Profile Information
        3. 4.11.5.3 Utilize Preferred Function Order in Re-Build of Application
      6. 4.11.6 Comma-Separated Values (CSV) Files with Weighted Call Graph (WCG) Information
      7. 4.11.7 Linker Command File Operator - unordered()
        1. 4.11.7.1 Output Section for unordered() Operator
        2.       124
        3. 4.11.7.2 Generated Linker Map File for
        4. 4.11.7.3 About Dot (.) Expressions in the Presence of unordered()
          1. 4.11.7.3.1 Respecting Position of a . Expression
          2.        128
        5. 4.11.7.4 GROUPs and UNIONs
          1. 4.11.7.4.1 Applying unordered() to GROUPs
        6.       131
      8. 4.11.8 Things to be Aware of
    12. 4.12 Indicating Whether Certain Aliasing Techniques Are Used
      1. 4.12.1 Use the --aliased_variables Option When Certain Aliases are Used
      2. 4.12.2 Use the --no_bad_aliases Option to Indicate That These Techniques Are Not Used
      3. 4.12.3 Using the --no_bad_aliases Option With the Assembly Optimizer
    13. 4.13 Prevent Reordering of Associative Floating-Point Operations
    14. 4.14 Use Caution With asm Statements in Optimized Code
    15. 4.15 Using Performance Advice to Optimize Your Code
      1. 4.15.1  Advice #27000
      2. 4.15.2  Advice #27001 Increase Optimization Level
      3. 4.15.3  Advice #27002 Do not turn off software pipelining
      4. 4.15.4  Advice #27003 Avoid compiling with debug options
      5. 4.15.5  Advice #27004 No Performance Advice generated
      6. 4.15.6  Advice #30000 Prevent Loop Disqualification due to call
      7. 4.15.7  Advice #30001 Prevent Loop Disqualification due to rts-call
      8. 4.15.8  Advice #30002 Prevent Loop Disqualification due to asm statement
      9. 4.15.9  Advice #30003 Prevent Loop Disqualification due to complex condition
      10. 4.15.10 Advice #30004 Prevent Loop Disqualification due to switch statement
      11. 4.15.11 Advice #30005 Prevent Loop Disqualification due to arithmetic operation
      12. 4.15.12 Advice #30006 Prevent Loop Disqualification due to call(2)
      13. 4.15.13 Advice #30007 Prevent Loop Disqualification due to rts-call(2)
      14. 4.15.14 Advice #30008 Improve Loop; Qualify with restrict
      15. 4.15.15 Advice #30009 Improve Loop; Add MUST_ITERATE pragma
      16. 4.15.16 Advice #30010 Improve Loop; Add MUST_ITERATE pragma(2)
      17. 4.15.17 Advice #30011 Improve Loop; Add _nasssert()
    16. 4.16 Using the Interlist Feature With Optimization
    17. 4.17 Debugging and Profiling Optimized Code
      1. 4.17.1 Profiling Optimized Code
    18. 4.18 What Kind of Optimization Is Being Performed?
      1. 4.18.1  Cost-Based Register Allocation
      2. 4.18.2  Alias Disambiguation
      3. 4.18.3  Branch Optimizations and Control-Flow Simplification
      4. 4.18.4  Data Flow Optimizations
      5. 4.18.5  Expression Simplification
      6. 4.18.6  Inline Expansion of Functions
      7. 4.18.7  Function Symbol Aliasing
      8. 4.18.8  Induction Variables and Strength Reduction
      9. 4.18.9  Loop-Invariant Code Motion
      10. 4.18.10 Loop Rotation
      11. 4.18.11 Vectorization (SIMD)
      12. 4.18.12 Instruction Scheduling
      13. 4.18.13 Register Variables
      14. 4.18.14 Register Tracking/Targeting
      15. 4.18.15 Software Pipelining
  6. 5 Using the Assembly Optimizer
    1. 5.1 Code Development Flow to Increase Performance
    2. 5.2 About the Assembly Optimizer
    3. 5.3 What You Need to Know to Write Linear Assembly
      1. 5.3.1 Linear Assembly Source Statement Format
      2. 5.3.2 Register Specification for Linear Assembly
        1. 5.3.2.1 Linear Assembly Code for Computing a Dot Product
        2.       183
        3. 5.3.2.2 C Code for Computing a Dot Product
        4.       185
        5. 5.3.2.3 Specifying a Register Pair
        6.       187
        7. 5.3.2.4 Specifying a Register Quad (C6600 Only)
        8.       189
      3. 5.3.3 Functional Unit Specification for Linear Assembly
      4. 5.3.4 Using Linear Assembly Source Comments
        1. 5.3.4.1 Lmac Function Code Showing Comments
      5. 5.3.5 Assembly File Retains Your Symbolic Register Names
    4. 5.4 Assembly Optimizer Directives
      1.      .call
      2.      .circ
      3.      .cproc/.endproc
      4.      .map
      5.      .mdep
      6.      .mptr
      7.      .no_mdep
      8.      .pref
      9.      .proc/.endproc
      10.      .reg
      11.      .rega/.regb
      12.      .reserve
      13.      .return
      14.      .trip
      15.      .volatile
      16. 5.4.1 Instructions That Are Not Allowed in Procedures
    5. 5.5 Avoiding Memory Bank Conflicts With the Assembly Optimizer
      1. 5.5.1 Preventing Memory Bank Conflicts
        1. 5.5.1.1 Load and Store Instructions That Specify Memory Bank Information
      2. 5.5.2 A Dot Product Example That Avoids Memory Bank Conflicts
        1. 5.5.2.1 C Code for Dot Product
        2. 5.5.2.2 Linear Assembly for Dot Product
        3. 5.5.2.3 Dot Product Software-Pipelined Kernel
        4.       218
        5. 5.5.2.4 Dot Product From Unrolled to Prevent Memory Bank Conflicts
        6.       220
        7. 5.5.2.5 Unrolled Dot Product Kernel From
        8.       222
      3. 5.5.3 Memory Bank Conflicts for Indexed Pointers
        1. 5.5.3.1 Using .mptr for Indexed Pointers
      4. 5.5.4 Memory Bank Conflict Algorithm
    6. 5.6 Memory Alias Disambiguation
      1. 5.6.1 How the Assembly Optimizer Handles Memory References (Default)
      2. 5.6.2 Using the --no_bad_aliases Option to Handle Memory References
      3. 5.6.3 Using the .no_mdep Directive
      4. 5.6.4 Using the .mdep Directive to Identify Specific Memory Dependencies
        1. 5.6.4.1 Annotating a Memory Reference
        2.       232
        3. 5.6.4.2 Software Pipeline Using .mdep ld1, st1
        4.       234
        5. 5.6.4.3 Software Pipeline Using .mdep st1, ld1 and .mdep ld1, st1
        6.       236
      5. 5.6.5 Memory Alias Examples
  7. 6 Linking C/C++ Code
    1. 6.1 Invoking the Linker Through the Compiler (-z Option)
      1. 6.1.1 Invoking the Linker Separately
      2. 6.1.2 Invoking the Linker as Part of the Compile Step
      3. 6.1.3 Disabling the Linker (--compile_only Compiler Option)
    2. 6.2 Linker Code Optimizations
      1. 6.2.1 Conditional Linking
      2. 6.2.2 Generating Function Subsections (--gen_func_subsections Compiler Option)
      3. 6.2.3 Generating Aggregate Data Subsections (--gen_data_subsections Compiler Option)
    3. 6.3 Controlling the Linking Process
      1. 6.3.1 Including the Run-Time-Support Library
        1. 6.3.1.1 Automatic Run-Time-Support Library Selection
          1. 6.3.1.1.1 Using the --issue_remarks Option
        2. 6.3.1.2 Manual Run-Time-Support Library Selection
        3. 6.3.1.3 Library Order for Searching for Symbols
      2. 6.3.2 Run-Time Initialization
      3. 6.3.3 Global Object Constructors
      4. 6.3.4 Specifying the Type of Global Variable Initialization
      5. 6.3.5 Specifying Where to Allocate Sections in Memory
      6. 6.3.6 A Sample Linker Command File
  8. 7 C/C++ Language Implementation
    1. 7.1  Characteristics of TMS320C6000 C
      1. 7.1.1 Implementation-Defined Behavior
    2. 7.2  Characteristics of TMS320C6000 C++
    3. 7.3  Data Types
      1. 7.3.1 Size of Enum Types
      2. 7.3.2 Vector Data Types
    4. 7.4  File Encodings and Character Sets
    5. 7.5  Keywords
      1. 7.5.1 The complex Keyword
      2. 7.5.2 The const Keyword
      3. 7.5.3 The __cregister Keyword
        1. 7.5.3.1 Define and Use Control Registers
      4. 7.5.4 The __interrupt Keyword
      5. 7.5.5 The __near and __far Keywords
        1. 7.5.5.1 Near and Far Data Objects
        2. 7.5.5.2 Near and Far Function Calls
      6. 7.5.6 The restrict Keyword
      7. 7.5.7 The volatile Keyword
    6. 7.6  C++ Exception Handling
    7. 7.7  Register Variables and Parameters
    8. 7.8  The __asm Statement
    9. 7.9  Pragma Directives
      1. 7.9.1  The CALLS Pragma
      2. 7.9.2  The CODE_ALIGN Pragma
      3. 7.9.3  The CODE_SECTION Pragma
      4. 7.9.4  The DATA_ALIGN Pragma
      5. 7.9.5  The DATA_MEM_BANK Pragma
        1. 7.9.5.1 Using the DATA_MEM_BANK Pragma
      6. 7.9.6  The DATA_SECTION Pragma
        1. 7.9.6.1 Using the DATA_SECTION Pragma C Source File
        2. 7.9.6.2 Using the DATA_SECTION Pragma C++ Source File
        3. 7.9.6.3 Using the DATA_SECTION Pragma Assembly Source File
      7. 7.9.7  The Diagnostic Message Pragmas
      8. 7.9.8  The FORCEINLINE Pragma
      9. 7.9.9  The FORCEINLINE_RECURSIVE Pragma
      10. 7.9.10 The FUNC_ALWAYS_INLINE Pragma
      11. 7.9.11 The FUNC_CANNOT_INLINE Pragma
      12. 7.9.12 The FUNC_EXT_CALLED Pragma
      13. 7.9.13 The FUNC_INTERRUPT_THRESHOLD Pragma
      14. 7.9.14 The FUNC_IS_PURE Pragma
      15. 7.9.15 The FUNC_IS_SYSTEM Pragma
      16. 7.9.16 The FUNC_NEVER_RETURNS Pragma
      17. 7.9.17 The FUNC_NO_GLOBAL_ASG Pragma
      18. 7.9.18 The FUNC_NO_IND_ASG Pragma
      19. 7.9.19 The FUNCTION_OPTIONS Pragma
      20. 7.9.20 The INTERRUPT Pragma
      21. 7.9.21 The LOCATION Pragma
      22. 7.9.22 The MUST_ITERATE Pragma
        1. 7.9.22.1 The MUST_ITERATE Pragma Syntax
        2. 7.9.22.2 Using MUST_ITERATE to Expand Compiler Knowledge of Loops
      23. 7.9.23 The NMI_INTERRUPT Pragma
      24. 7.9.24 The NOINIT and PERSISTENT Pragmas
      25. 7.9.25 The NOINLINE Pragma
      26. 7.9.26 The NO_HOOKS Pragma
      27. 7.9.27 The once Pragma
      28. 7.9.28 The pack Pragma
      29. 7.9.29 The PROB_ITERATE Pragma
      30. 7.9.30 The RETAIN Pragma
      31. 7.9.31 The SET_CODE_SECTION and SET_DATA_SECTION Pragmas
      32. 7.9.32 The STRUCT_ALIGN Pragma
      33. 7.9.33 The UNROLL Pragma
    10. 7.10 The _Pragma Operator
    11. 7.11 Application Binary Interface
    12. 7.12 Object File Symbol Naming Conventions (Linknames)
    13. 7.13 Changing the ANSI/ISO C/C++ Language Mode
      1. 7.13.1 C99 Support (--c99)
      2. 7.13.2 C11 Support (--c11)
      3. 7.13.3 Strict ANSI Mode and Relaxed ANSI Mode (--strict_ansi and --relaxed_ansi)
    14. 7.14 GNU and Clang Language Extensions
      1. 7.14.1 Extensions
      2. 7.14.2 Function Attributes
      3. 7.14.3 For Loop Attributes
      4. 7.14.4 Variable Attributes
      5. 7.14.5 Type Attributes
      6. 7.14.6 Built-In Functions
    15. 7.15 Operations and Functions for Vector Data Types
      1. 7.15.1 Vector Literals and Concatenation
      2. 7.15.2 Unary and Binary Operators for Vectors
      3. 7.15.3 Swizzle Operators for Vectors
      4. 7.15.4 Conversion Functions for Vectors
      5. 7.15.5 Re-Interpretation Functions for Vectors
      6. 7.15.6 Using printf() with Vectors
      7. 7.15.7 Built-In Vector Functions
  9. 8 Run-Time Environment
    1. 8.1  Memory Model
      1. 8.1.1 Sections
      2. 8.1.2 C/C++ System Stack
      3. 8.1.3 Dynamic Memory Allocation
      4. 8.1.4 Data Memory Models
        1. 8.1.4.1 Determining the Data Address Model
        2. 8.1.4.2 DP-Relative Vs. Absolute Addressing
        3. 8.1.4.3 Const Objects as Far
      5. 8.1.5 Trampoline Generation for Function Calls
      6. 8.1.6 Position Independent Data
    2. 8.2  Object Representation
      1. 8.2.1 Data Type Storage
        1. 8.2.1.1 char and short Data Types (signed and unsigned)
        2. 8.2.1.2 enum, int, and long Data Types (signed and unsigned)
        3. 8.2.1.3 float Data Type
        4. 8.2.1.4 The __int40_t Data Type (signed and unsigned)
        5. 8.2.1.5 long long Data Types (signed and unsigned)
        6. 8.2.1.6 double and long double Data Types
        7. 8.2.1.7 Pointer to Data Member Types
        8. 8.2.1.8 Pointer to Member Function Types
        9. 8.2.1.9 Structures and Arrays
      2. 8.2.2 Bit Fields
      3. 8.2.3 Character String Constants
      4.      366
    3. 8.3  Register Conventions
    4. 8.4  Function Structure and Calling Conventions
      1. 8.4.1 How a Function Makes a Call
      2. 8.4.2 How a Called Function Responds
      3. 8.4.3 Accessing Arguments and Local Variables
    5. 8.5  Accessing Linker Symbols in C and C++
    6. 8.6  Interfacing C and C++ With Assembly Language
      1. 8.6.1  Using Assembly Language Modules With C/C++ Code
      2. 8.6.2  Accessing Assembly Language Functions From C/C++
        1. 8.6.2.1 Calling an Assembly Language Function From a C/C++ Program
        2. 8.6.2.2 Assembly Language Program Called by
        3.       378
      3. 8.6.3  Accessing Assembly Language Variables From C/C++
        1. 8.6.3.1 Accessing Assembly Language Global Variables
          1. 8.6.3.1.1 Assembly Language Variable Program
          2. 8.6.3.1.2 C Program to Access Assembly Language From
        2.       383
        3. 8.6.3.2 Accessing Assembly Language Constants
          1. 8.6.3.2.1 Accessing an Assembly Language Constant From C
          2. 8.6.3.2.2 Assembly Language Program for
          3.        387
      4. 8.6.4  Sharing C/C++ Header Files With Assembly Source
      5. 8.6.5  Using Inline Assembly Language
      6. 8.6.6  Using Intrinsics to Access Assembly Language Statements
      7. 8.6.7  The __x128_t Container Type
        1. 8.6.7.1 The __x128_t Container Type
        2.       393
      8. 8.6.8  The __float2_t Container Type
      9. 8.6.9  Using Intrinsics for Interrupt Control and Atomic Sections
      10. 8.6.10 Using Unaligned Data and 64-Bit Values
        1. 8.6.10.1 Using the _mem8 Intrinsic
      11. 8.6.11 Using MUST_ITERATE and _nassert to Enable SIMD and Expand Compiler Knowledge of Loops
      12. 8.6.12 Methods to Align Data
        1. 8.6.12.1 Base Address of an Array
        2. 8.6.12.2 Offset from the Base of an Array
        3. 8.6.12.3 Dynamic Memory Allocation
        4. 8.6.12.4 Member of a Structure or Class
          1. 8.6.12.4.1 An Array in a Structure
          2. 8.6.12.4.2 An Array in a Class
          3.        406
      13. 8.6.13 SAT Bit Side Effects
      14. 8.6.14 IRP and AMR Conventions
      15. 8.6.15 Floating Point and Saturation Control Register Side Effects
    7. 8.7  Interrupt Handling
      1. 8.7.1 Saving the SGIE Bit
      2. 8.7.2 Saving Registers During Interrupts
      3. 8.7.3 Using C/C++ Interrupt Routines
      4. 8.7.4 Using Assembly Language Interrupt Routines
    8. 8.8  Run-Time-Support Arithmetic Routines
    9. 8.9  System Initialization
      1. 8.9.1 Boot Hook Functions for System Pre-Initialization
      2. 8.9.2 Automatic Initialization of Variables
        1. 8.9.2.1 Zero Initializing Variables
        2. 8.9.2.2 Direct Initialization
        3. 8.9.2.3 Autoinitialization of Variables at Run Time
        4. 8.9.2.4 Autoinitialization Tables
          1. 8.9.2.4.1 Length Followed by Data Format
          2. 8.9.2.4.2 Zero Initialization Format
          3. 8.9.2.4.3 Run Length Encoded (RLE) Format
          4. 8.9.2.4.4 Lempel-Ziv-Storer-Szymanski Compression (LZSS) Format
          5. 8.9.2.4.5 Sample C Code to Process the C Autoinitialization Table
        5. 8.9.2.5 Initialization of Variables at Load Time
        6. 8.9.2.6 Global Constructors
    10. 8.10 Support for Multi-Threaded Applications
      1. 8.10.1 Compiling with OpenMP
      2. 8.10.2 Multi-Threading Runtime Support
        1. 8.10.2.1 Runtime Thread Safety
        2. 8.10.2.2 Thread Creation, Initialization, and Termination
        3. 8.10.2.3 Thread Local Storage (TLS)
        4. 8.10.2.4 Accessing Shared Data
  10. 9 Using Run-Time-Support Functions and Building Libraries
    1. 9.1 C and C++ Run-Time Support Libraries
      1. 9.1.1 Linking Code With the Object Library
      2. 9.1.2 Header Files
      3. 9.1.3 Modifying a Library Function
      4. 9.1.4 Support for String Handling
      5. 9.1.5 Minimal Support for Internationalization
      6. 9.1.6 Support for Time and Clock Functions
      7. 9.1.7 Allowable Number of Open Files
      8. 9.1.8 Library Naming Conventions
    2. 9.2 The C I/O Functions
      1. 9.2.1 High-Level I/O Functions
        1. 9.2.1.1 Formatting and the Format Conversion Buffer
      2. 9.2.2 Overview of Low-Level I/O Implementation
        1.       open
        2.       close
        3.       read
        4.       write
        5.       lseek
        6.       unlink
        7.       rename
      3. 9.2.3 Device-Driver Level I/O Functions
        1.       DEV_open
        2.       DEV_close
        3.       DEV_read
        4.       DEV_write
        5.       DEV_lseek
        6.       DEV_unlink
        7.       DEV_rename
      4. 9.2.4 Adding a User-Defined Device Driver for C I/O
        1. 9.2.4.1 Mapping Default Streams to Device
      5. 9.2.5 The device Prefix
        1.       add_device
        2.       470
        3. 9.2.5.1 Program for C I/O Device
    3. 9.3 Handling Reentrancy (_register_lock() and _register_unlock() Functions)
    4. 9.4 Library-Build Process
      1. 9.4.1 Required Non-Texas Instruments Software
      2. 9.4.2 Using the Library-Build Process
        1. 9.4.2.1 Automatic Standard Library Rebuilding by the Linker
        2. 9.4.2.2 Invoking mklib Manually
          1. 9.4.2.2.1 Building Standard Libraries
          2. 9.4.2.2.2 Shared or Read-Only Library Directory
          3. 9.4.2.2.3 Building Libraries With Custom Options
          4. 9.4.2.2.4 The mklib Program Option Summary
      3. 9.4.3 Extending mklib
        1. 9.4.3.1 Underlying Mechanism
        2. 9.4.3.2 Libraries From Other Vendors
  11. 10C++ Name Demangler
    1. 10.1 Invoking the C++ Name Demangler
    2. 10.2 Sample Usage of the C++ Name Demangler
  12.   A Glossary
    1.     A.1 Terminology
  13.   B Revision History
  14.   491
  15.   492
  16.   B Earlier Revisions

3.11 Using Inline Function Expansion

When an inline function is called, a copy of the C/C++ source code for the function is inserted at the point of the call. This is known as inline function expansion, commonly called function inlining or just inlining. Inline function expansion can speed up execution by eliminating function call overhead. This is particularly beneficial for very small functions that are called frequently. Function inlining involves a tradeoff between execution speed and code size, because the code is duplicated at each function call site. Large functions that are called in many places are poor candidates for inlining.

Note:

Excessive Inlining Can Degrade Performance: Excessive inlining can make the compiler dramatically slower and degrade the performance of generated code.

 

Function inlining is triggered by the following situations:

  • The use of built-in intrinsic operations. Intrinsic operations look like function calls, and are inlined automatically, even though no function body exists.
  • Use of the inline keyword or the equivalent __inline keyword. Functions declared with the inline keyword may be inlined by the compiler if you set --opt_level=0 or greater. The inline keyword is a suggestion from the programmer to the compiler. Even if your optimization level is high, inlining is still optional for the compiler. The compiler decides whether to inline a function based on the length of the function, the number of times it is called, your --opt_for_speed setting, and any contents of the function that disqualify it from inlining (see Section 3.11.2). Functions can be inlined at --opt_level=0 or above if the function body is visible in the same module or if -pm is also used and the function is visible in one of the modules being compiled. Functions defined as both static and inline are more likely to be inlined.
  • When --opt_level=3 is used, the compiler may automatically inline eligible functions even if they are not declared as inline functions. The same list of decision factors listed for functions explicitly defined with the inline keyword is used. For more about automatic function inlining, see Section 4.5.
  • The pragma FUNC_ALWAYS_INLINE (Section 7.9.10) and the equivalent always_inline attribute (Section 7.14.2) force a function to be inlined (where it is legal to do so) unless --opt_level=off. That is, the pragma FUNC_ALWAYS_INLINE forces function inlining even if the function is not declared as inline and the --opt_level=0 or --opt_level=1.
  • The FORCEINLINE pragma (Section 7.9.8) forces functions to be inlined in the annotated statement. That is, it has no effect on those functions in general, only on function calls in a single statement. The FORCEINLINE_RECURSIVE pragma forces inlining not only of calls visible in the statement, but also in the inlined bodies of calls from that statement.
  • The --disable_inlining option prevents any inlining. The pragma FUNC_CANNOT_INLINE prevents a function from being inlined. The NOINLINE pragma prevents calls within a single statement from being inlined. (NOINLINE is the inverse of the FORCEINLINE pragma.)
Note:

Function Inlining Can Greatly Increase Code Size: Function inlining increases code size, especially inlining a function that is called in a number of places. Function inlining is optimal for functions that are called only from a small number of places and for small functions.

The semantics of the inline keyword in C code follow the C99 standard. The semantics of the inline keyword in C++ code follow the C++ standard.

The inline keyword is supported in all C++ modes, in relaxed ANSI mode for all C standards, and in strict ANSI mode for C99 and C11. It is disabled in strict ANSI mode for C89, because it is a language extension that could conflict with a strictly conforming program. If you want to define inline functions while in strict ANSI C89 mode, use the alternate keyword __inline.

Compiler options that affect inlining are: --opt_level, --auto_inline, --remove_hooks_when_inlining, --opt_for_speed, and --disable_inlining.