SPRAC71B February   2019  – October 2023

 

  1.   1
  2. 1 Introduction
    1. 1.1  ABIs for the C28x
    2. 1.2  Scope
    3. 1.3  ABI Variants
    4. 1.4  Toolchains and Interoperability
    5. 1.5  Libraries
    6. 1.6  Types of Object Files
    7. 1.7  Segments
    8. 1.8  C28x Architecture Overview
    9. 1.9  C28x Memory Models
    10. 1.10 Reference Documents
    11. 1.11 Code Fragment Notation
  3. 2 Data Representation
    1. 2.1 Basic Types
    2. 2.2 Data in Registers
    3. 2.3 Data in Memory
    4. 2.4 Pointer Types
    5. 2.5 Complex Types
    6. 2.6 Structures and Unions
    7. 2.7 Arrays
    8. 2.8 Bit Fields
      1. 2.8.1 Volatile Bit Fields
    9. 2.9 Enumeration Types
  4. 3 Calling Conventions
    1. 3.1 Call and Return
      1. 3.1.1 Call Instructions
        1. 3.1.1.1 Indirect Calls
        2. 3.1.1.2 Direct Calls
      2. 3.1.2 Return Instruction
      3. 3.1.3 Pipeline Conventions
      4. 3.1.4 Weak Functions
    2. 3.2 Register Conventions
      1. 3.2.1 Argument Registers
      2. 3.2.2 Callee-Saved Registers
    3. 3.3 Argument Passing
      1. 3.3.1 Passing 16-Bit Arguments
      2. 3.3.2 Passing Longer Arguments
      3. 3.3.3 C++ Argument Passing
      4. 3.3.4 Passing Structs and Unions
      5. 3.3.5 Stack Layout of Arguments Not Passed in Registers
      6. 3.3.6 Frame Pointer
    4. 3.4 Return Values
    5. 3.5 Structures and Unions Passed and Returned by Reference
    6. 3.6 Conventions for Compiler Helper Functions
    7. 3.7 Prolog and Epilog Helper Functions
    8. 3.8 Scratch Registers for Functions Already Seen
    9. 3.9 Interrupt Functions
  5. 4 Data Allocation and Addressing
    1. 4.1 Data Sections and Segments
    2. 4.2 Data Blocking
    3. 4.3 Addressing Modes
    4. 4.4 Allocation and Addressing of Static Data
      1. 4.4.1 Addressing Methods for Static Data
      2. 4.4.2 Placement Conventions for Static Data
        1. 4.4.2.1 Abstract Conventions for Addressing
      3. 4.4.3 Initialization of Static Data
    5. 4.5 Automatic Variables
    6. 4.6 Frame Layout
      1. 4.6.1 Stack Alignment
      2. 4.6.2 Register Save Order
    7. 4.7 Heap-Allocated Objects
  6. 5 Code Allocation and Addressing
    1. 5.1 Computing the Address of a Code Label
    2. 5.2 Calls
      1. 5.2.1 Direct Call
      2. 5.2.2 Far Call Trampoline
      3. 5.2.3 Indirect Calls
  7. 6 Helper Function API
    1. 6.1 Floating-Point Behavior
    2. 6.2 C Helper Function API
    3. 6.3 Floating-Point Helper Functions for C99
  8. 7 Standard C Library API
    1. 7.1  About Standard C Libraries
    2. 7.2  Reserved Symbols
    3. 7.3  <assert.h> Implementation
    4. 7.4  <complex.h> Implementation
    5. 7.5  <ctype.h> Implementation
    6. 7.6  <errno.h> Implementation
    7. 7.7  <float.h> Implementation
    8. 7.8  <inttypes.h> Implementation
    9. 7.9  <iso646.h> Implementation
    10. 7.10 <limits.h> Implementation
    11. 7.11 <locale.h> Implementation
    12. 7.12 <math.h> Implementation
    13. 7.13 <setjmp.h> Implementation
    14. 7.14 <signal.h> Implementation
    15. 7.15 <stdarg.h> Implementation
    16. 7.16 <stdbool.h> Implementation
    17. 7.17 <stddef.h> Implementation
    18. 7.18 <stdint.h> Implementation
    19. 7.19 <stdio.h> Implementation
    20. 7.20 <stdlib.h> Implementation
    21. 7.21 <string.h> Implementation
    22. 7.22 <tgmath.h> Implementation
    23. 7.23 <time.h> Implementation
    24. 7.24 <wchar.h> Implementation
    25. 7.25 <wctype.h> Implementation
  9. 8 C++ ABI
    1. 8.1  Limits (GC++ABI 1.2)
    2. 8.2  Export Template (GC++ABI 1.4.2)
    3. 8.3  Data Layout (GC++ABI Chapter 2)
    4. 8.4  Initialization Guard Variables (GC++ABI 2.8)
    5. 8.5  Constructor Return Value (GC++ABI 3.1.5)
    6. 8.6  One-Time Construction API (GC++ABI 3.3.2)
    7. 8.7  Controlling Object Construction Order (GC++ ABI 3.3.4)
    8. 8.8  Demangler API (GC++ABI 3.4)
    9. 8.9  Static Data (GC++ ABI 5.2.2)
    10. 8.10 Virtual Tables and the Key function (GC++ABI 5.2.3)
    11. 8.11 Unwind Table Location (GC++ABI 5.3)
  10. 9 Exception Handling
    1. 9.1  Overview
    2. 9.2  PREL31 Encoding
    3. 9.3  The Exception Index Table (EXIDX)
      1. 9.3.1 Pointer to Out-of-Line EXTAB Entry
      2. 9.3.2 EXIDX_CANTUNWIND
      3. 9.3.3 Inlined EXTAB Entry
    4. 9.4  The Exception Handling Instruction Table (EXTAB)
      1. 9.4.1 EXTAB Generic Model
      2. 9.4.2 EXTAB Compact Model
      3. 9.4.3 Personality Routines
    5. 9.5  Unwinding Instructions
      1. 9.5.1 Common Sequence
      2. 9.5.2 Byte-Encoded Unwinding Instructions
    6. 9.6  Descriptors
      1. 9.6.1 Encoding of Type Identifiers
      2. 9.6.2 Scope
      3. 9.6.3 Cleanup Descriptor
      4. 9.6.4 Catch Descriptor
      5. 9.6.5 Function Exception Specification (FESPEC) Descriptor
    7. 9.7  Special Sections
    8. 9.8  Interaction With Non-C++ Code
      1. 9.8.1 Automatic EXIDX Entry Generation
      2. 9.8.2 Hand-Coded Assembly Functions
    9. 9.9  Interaction With System Features
      1. 9.9.1 Shared Libraries
      2. 9.9.2 Overlays
      3. 9.9.3 Interrupts
    10. 9.10 Assembly Language Operators in the TI Toolchain
  11. 10DWARF
    1. 10.1 DWARF Register Names
    2. 10.2 Call Frame Information
    3. 10.3 Vendor Names
    4. 10.4 Vendor Extensions
  12. 11ELF Object Files (Processor Supplement)
    1. 11.1 Registered Vendor Names
    2. 11.2 ELF Header
    3. 11.3 Sections
      1. 11.3.1 Section Indexes
      2. 11.3.2 Section Types
      3. 11.3.3 Extended Section Header Attributes
      4. 11.3.4 Subsections
      5. 11.3.5 Special Sections
      6. 11.3.6 Section Alignment
    4. 11.4 Symbol Table
      1. 11.4.1 Symbol Types
      2. 11.4.2 Common Block Symbols
      3. 11.4.3 Symbol Names
      4. 11.4.4 Reserved Symbol Names
      5. 11.4.5 Mapping Symbols
    5. 11.5 Relocation
      1. 11.5.1 Relocation Types
        1. 11.5.1.1 Absolute Relocations
        2. 11.5.1.2 PC-Relative Relocations
        3. 11.5.1.3 Relocations in Data Sections
        4. 11.5.1.4 Relocations for C28x Instructions
        5. 11.5.1.5 Other Relocation Types
      2. 11.5.2 Relocation Operations
      3. 11.5.3 Relocation of Unresolved Weak References
  13. 12ELF Program Loading and Linking (Processor Supplement)
    1. 12.1 Program Header
      1. 12.1.1 Base Address
      2. 12.1.2 Segment Contents
      3. 12.1.3 Thread-Local Storage
    2. 12.2 Program Loading
  14. 13Build Attributes
    1. 13.1 About Build Attributes
    2. 13.2 C28x ABI Build Attribute Subsection
    3. 13.3 Build Attribute Tags
  15. 14Copy Tables and Variable Initialization
    1. 14.1 About Copy Tables
    2. 14.2 Copy Table Format
    3. 14.3 Compressed Data Formats
      1. 14.3.1 RLE
      2. 14.3.2 LZSS Format
    4. 14.4 Variable Initialization
  16. 15Revision History
    1.     188

6.1 Floating-Point Behavior

Floating-point behavior varies by device and by toolchain and is therefore difficult to standardize. The goal of the ABI is to provide a basis for conformance to the C, C99, and C++ standards. Of these C99 is the best-specified with respect to floating-point. Appendix F of the C99 standard defines floating-point behavior of the C language behavior in terms of the IEEE floating-point standard (ISO IEC 60559:1989, previously designated as ANSI/IEEE 754−1985).

The C28x ABI specifies that the helper functions in this section that operate on floating-point values must conform to the behavior specified by Appendix F of the C99 standard.

C99 allows customization of, and access to, the floating-point behavioral environment though the <fenv.h> header file. For purposes of standardizing the behavior of the helper functions, the ABI specifies them to operate in accordance with a basic default environment, with the following properties:

  • The rounding mode is round to nearest. Dynamic rounding precision modes are not supported.
  • No floating-point exceptions are supported.
  • Inputs that represent Signaling NaNs behave like Quiet NaNs.
  • The helper functions support only the behavior under the FENV_ACCESS off state. That is, the program is assumed to execute in non-stop mode and assumed not to access the floating-point environment.

A toolchain is free to implement more complete floating-point support, using its own library. Users who invoke toolchain-specific floating-point support may be required to link using that toolchain's library (in addition to an ABI-conforming helper function library).