SPRACN0F October   2021  – March 2023 TMS320F280021 , TMS320F280021-Q1 , TMS320F280023 , TMS320F280023-Q1 , TMS320F280023C , TMS320F280025 , TMS320F280025-Q1 , TMS320F280025C , TMS320F280025C-Q1 , TMS320F280033 , TMS320F280034 , TMS320F280034-Q1 , TMS320F280036-Q1 , TMS320F280036C-Q1 , TMS320F280037 , TMS320F280037-Q1 , TMS320F280037C , TMS320F280037C-Q1 , TMS320F280038-Q1 , TMS320F280038C-Q1 , TMS320F280039 , TMS320F280039-Q1 , TMS320F280039C , TMS320F280039C-Q1 , TMS320F280040-Q1 , TMS320F280040C-Q1 , TMS320F280041 , TMS320F280041-Q1 , TMS320F280041C , TMS320F280041C-Q1 , TMS320F280045 , TMS320F280048-Q1 , TMS320F280048C-Q1 , TMS320F280049 , TMS320F280049-Q1 , TMS320F280049C , TMS320F280049C-Q1 , TMS320F28374D , TMS320F28374S , TMS320F28375D , TMS320F28375S , TMS320F28375S-Q1 , TMS320F28376D , TMS320F28376S , TMS320F28377S , TMS320F28377S-Q1 , TMS320F28378D , TMS320F28378S , TMS320F28379D , TMS320F28379D-Q1 , TMS320F28379S , TMS320F28384D , TMS320F28384S , TMS320F28386D , TMS320F28386S , TMS320F28388D , TMS320F28388S , TMS320F28P650DH , TMS320F28P650DK , TMS320F28P650SH , TMS320F28P650SK , TMS320F28P659DK-Q1

 

  1.    The Essential Guide for Developing With C2000™ Real-Time Microcontrollers
  2.   Trademarks
  3. 1C2000 and Real-Time Control
    1. 1.1 Getting Started Resources
    2. 1.2 Processing
    3. 1.3 Control
    4. 1.4 Sensing
    5. 1.5 Interface
    6. 1.6 Functional Safety
  4. 2Sensing Key Technologies
    1. 2.1 Accurate Digital Domain Representation of Analog Signals
      1. 2.1.1 Value Proposition
      2. 2.1.2 In Depth
      3. 2.1.3 Device List
      4. 2.1.4 Hardware Platforms and Software Examples
      5. 2.1.5 Documentation
    2. 2.2 Optimizing Acquisition Time vs Circuit Complexity for Analog Inputs
      1. 2.2.1 Value Proposition
      2. 2.2.2 In Depth
      3. 2.2.3 Device List
      4. 2.2.4 Hardware Platforms and Software Examples
      5. 2.2.5 Documentation
    3. 2.3 Hardware Based Monitoring of Dual-Thresholds Using a Single Pin Reference
      1. 2.3.1 Value Proposition
      2. 2.3.2 In Depth
      3. 2.3.3 Device List
      4. 2.3.4 Hardware Platforms and Software Examples
      5. 2.3.5 Documentation
    4. 2.4 Resolving Tolerance and Aging Effects During ADC Sampling
      1. 2.4.1 Value Proposition
      2. 2.4.2 In Depth
      3. 2.4.3 Device List
      4. 2.4.4 Hardware Platforms and Software Examples
      5. 2.4.5 Documentation
    5. 2.5 Realizing Rotary Sensing Solutions Using C2000 Configurable Logic Block
      1. 2.5.1 Value Proposition
      2. 2.5.2 In Depth
      3. 2.5.3 Device List
      4. 2.5.4 Hardware Platforms and Software Examples
      5. 2.5.5 Documentation
    6. 2.6 Smart Sensing Across An Isolation Boundary
      1. 2.6.1 Value Proposition
      2. 2.6.2 In Depth
      3. 2.6.3 Device List
      4. 2.6.4 Hardware Platforms and Software Examples
      5. 2.6.5 Documentation
    7. 2.7 Enabling Intra-Period Updates in High Bandwidth Control Topologies
      1. 2.7.1 Value Proposition
      2. 2.7.2 In Depth
      3. 2.7.3 Device List
      4. 2.7.4 Hardware Platforms and Software Examples
      5. 2.7.5 Documentation
    8. 2.8 Accurate Monitoring of Real-Time Control System Events Without the Need for Signal Conditioning
      1. 2.8.1 Value Proposition
      2. 2.8.2 In Depth
      3. 2.8.3 Device List
      4. 2.8.4 Hardware Platforms and Software Examples
      5. 2.8.5 Documentation
  5. 3Processing Key Technologies
    1. 3.1 Accelerated Trigonometric Math Functions
      1. 3.1.1 Value Proposition
      2. 3.1.2 In Depth
      3. 3.1.3 Device List
      4. 3.1.4 Hardware Platforms and Software Examples
      5. 3.1.5 Documentation
    2. 3.2 Fast Onboard Integer Division
      1. 3.2.1 Value Proposition
      2. 3.2.2 In Depth
      3. 3.2.3 Device List
      4. 3.2.4 Hardware Platforms and Software Platforms
      5. 3.2.5 Documentation
    3. 3.3 Hardware Support for Double-Precision Floating-Point Operations
      1. 3.3.1 Value Proposition
      2. 3.3.2 In Depth
      3. 3.3.3 Device List
      4. 3.3.4 Hardware Platforms and Software Examples
      5. 3.3.5 Documentation
    4. 3.4 Increasing Control Loop Bandwidth With An Independent Processing Unit
      1. 3.4.1 Value Proposition
      2. 3.4.2 In Depth
      3. 3.4.3 Device List
      4. 3.4.4 Hardware Platforms and Software Examples
      5. 3.4.5 Documentation
    5. 3.5 Flexible System Interconnect: C2000 X-Bar
      1. 3.5.1 Value Proposition
      2. 3.5.2 In Depth
      3. 3.5.3 Device List
      4. 3.5.4 Hardware Platforms and Software Examples
      5. 3.5.5 Documentation
    6. 3.6 Improving Control Performance With Nonlinear PID Control
      1. 3.6.1 Value Proposition
      2. 3.6.2 In Depth
      3. 3.6.3 Device List
      4. 3.6.4 Hardware Platforms and Software Examples
      5. 3.6.5 Documentation
    7. 3.7 Understanding Flash Memory Performance In Real-Time Control Applications
      1. 3.7.1 Value Proposition
      2. 3.7.2 In Depth
      3. 3.7.3 Device List
      4. 3.7.4 Hardware Platforms and Software Examples
      5. 3.7.5 Documentation
    8. 3.8 Deterministic Program Execution With the C28x DSP Core
      1. 3.8.1 Value Proposition
      2. 3.8.2 In Depth
      3. 3.8.3 Device List
      4. 3.8.4 Hardware Platforms and Software Examples
      5. 3.8.5 Documentation
    9. 3.9 Efficient Live Firmware Updates (LFU) and Firmware Over-The-Air (FOTA) updates
      1. 3.9.1 Value Proposition
      2. 3.9.2 In Depth
      3. 3.9.3 Device List
      4. 3.9.4 Hardware Platforms and Software Examples
      5. 3.9.5 Documentation
  6. 4Control Key Technologies
    1. 4.1 Reducing Limit Cycling in Control Systems With C2000 HRPWMs
      1. 4.1.1 Value Proposition
      2. 4.1.2 In Depth
      3. 4.1.3 Device List
      4. 4.1.4 Hardware Platforms and Software Examples
      5. 4.1.5 Documentation
    2. 4.2 Shoot Through Prevention for Current Control Topologies With Configurable Deadband
      1. 4.2.1 Value Proposition
      2. 4.2.2 In Depth
      3. 4.2.3 Device List
      4. 4.2.4 Documentation
    3. 4.3 On-Chip Hardware Customization Using the C2000 Configurable Logic Block
      1. 4.3.1 Value Proposition
      2. 4.3.2 In Depth
      3. 4.3.3 Device List
      4. 4.3.4 Hardware Platforms and Software Examples
      5. 4.3.5 Documentation
    4. 4.4 Fast Detection of Over and Under Currents and Voltages
      1. 4.4.1 Value Proposition
      2. 4.4.2 In Depth
      3. 4.4.3 Device List
      4. 4.4.4 Hardware Platforms and Software Examples
      5. 4.4.5 Documentation
    5. 4.5 Improving System Power Density With High Resolution Phase Control
      1. 4.5.1 Value Proposition
      2. 4.5.2 In Depth
      3. 4.5.3 Device List
      4. 4.5.4 Hardware Platforms and Software Examples
      5. 4.5.5 Documentation
    6. 4.6 Safe and Optimized PWM Updates in High-Frequency, Multi-Phase and Variable Frequency Topologies
      1. 4.6.1 Value Proposition
      2. 4.6.2 In Depth
      3. 4.6.3 Device List
      4. 4.6.4 Hardware Platforms and Software Examples
      5. 4.6.5 Documentation
    7. 4.7 Solving Event Synchronization Across Multiple Controllers in Decentralized Control Systems
      1. 4.7.1 Value Proposition
      2. 4.7.2 In Depth
      3. 4.7.3 Device List
      4. 4.7.4 Hardware Platforms and Software Examples
      5. 4.7.5 Documentation
  7. 5Interface Key Technologies
    1. 5.1 Direct Host Control of C2000 Peripherals
      1. 5.1.1 Value Proposition
      2. 5.1.2 In Depth
        1. 5.1.2.1 HIC Bridge for FSI Applications
        2. 5.1.2.2 HIC Bridge for Position Encoder Applications Using CLB
      3. 5.1.3 Device List
      4. 5.1.4 Hardware Platforms and Software Examples
      5. 5.1.5 Documentation
    2. 5.2 Securing External Communications and Firmware Updates With an AES Engine
      1. 5.2.1 Value Proposition
      2. 5.2.2 In Depth
      3. 5.2.3 Device List
      4. 5.2.4 Hardware Platforms and Software Examples
      5. 5.2.5 Documentation
    3. 5.3 Distributed Real-Time Control Across an Isolation Boundary
      1. 5.3.1 Value Proposition
      2. 5.3.2 In Depth
      3. 5.3.3 Device List
      4. 5.3.4 Hardware Platforms and Software Examples
      5. 5.3.5 Documentation
    4. 5.4 Custom Tests and Data Pattern Generation Using the Embedded Pattern Generator (EPG)
      1. 5.4.1 Value Proposition
      2. 5.4.2 In Depth
      3. 5.4.3 Device List
      4. 5.4.4 Hardware Platforms and Software Examples
      5. 5.4.5 Documentation
  8. 6Safety Key Technologies
    1. 6.1 Non-Intrusive Run Time Monitoring and Diagnostics as Part of the Control Loop
      1. 6.1.1 Value Proposition
      2. 6.1.2 In Depth
      3. 6.1.3 Device List
      4. 6.1.4 Hardware Platforms and Software Examples
      5. 6.1.5 Documentation
    2. 6.2 Hardware Built-In Self-Test of the C28x CPU
      1. 6.2.1 Value Proposition
      2. 6.2.2 In Depth
      3. 6.2.3 Device List
      4. 6.2.4 Hardware Platforms and Software Examples
      5. 6.2.5 Documentation
    3. 6.3 Zero CPU Overhead Cyclic Redundancy Check for Embedded On-Chip Memories
      1. 6.3.1 Value Proposition
      2. 6.3.2 In Depth
      3. 6.3.3 Device List
      4. 6.3.4 Hardware Platforms and Software Examples
      5. 6.3.5 Documentation
    4. 6.4 Boot Code Authentication Prior To Code Execution
      1. 6.4.1 Value Proposition
      2. 6.4.2 In Depth
      3. 6.4.3 Device List
      4. 6.4.4 Hardware Platforms and Software Examples
        1. 6.4.4.1 Documentation
  9. 7References
    1. 7.1 Device List
    2. 7.2 Hardware/Software Resources
    3. 7.3 Documentation
  10. 8Revision History

3.3.2 In Depth

The FPU64 module is extremely useful when 32-bit precision, such that, single-precision floating point is not sufficient for applications. In many real-time control applications, this is the case, so 64-bit precision (double-precision floating point) is needed. This normally comes at a price since native hardware on the device does not support double-precision floating point operations. Users see a significant increase in CPU cycles when running double-precision floating-point algorithms. By incorporating hardware support, this increase is avoided.

#GUID-BC8C6374-4B4C-4649-B574-974D746FBD8C/TABLE_AZY_HYC_SMB compares the performance of double-precision operations on FPU64 vs FPU (single-precision floating point hardware).

Table 3-3 Cycle Comparison Between FPU64 and FPU32
Floating-Point Operations FPU64 (cycles) --fp_mode=relaxed FPU (cycles) --fp_mode=strict FPU (cycles) --fp_mode=relaxed FPU64 disabled FPU (cycles) --fp_mode=strict FPU64 disabled
32-bit Division 8 234 8 234
64-bit Division 27 27 2222 2222

The results in the above table (profiled with optimization off, code running from RAM) illustrate that the performance of double-precision floating-point division on FPU64 is slightly more expensive than single-precision floating-point division on FPU32. In the single-precision case, when the floating point mode (fp_mode) is set to relaxed, the compiler generates hardware instructions to perform single-precision division, at the slight expense of accuracy. When the floating point mode is set to strict, the compiler does not generate hardware instructions and calls into the RTS library to maintain accuracy, but at the cost of cycles.

In the double-precision case, the floating point mode does not matter, because the FPU64 implementation is accurate. As long as the FPU64 is enabled, the compiler generates hardware instructions supported by the FPU64 to perform double-precision division. If FPU64 is disabled, only then the compiler does not generate FPU64 hardware instructions and calls into the RTS library. 64-bit floating-point division cannot take advantage of 32-bit floating-point hardware, even in relaxed mode, which is why the cycles remain 2222.

Devices with the FPU64 utilize the same registers as the FPU except for the addition of eight floating-point results extension registers for double-precision floating-point operations. The FPU64 enhancements support all existing FPU single-precision floating-point instructions in addition to the 64-bit double-precision floating-point instructions. FPU64 64-bit instructions operate in one to three pipeline cycles, with some instructions also supporting a parallel move operation.

Using the FPU64 is straightforward. Users simply write C code with double-precision floating-point variables and operations, compile it with TI C28x C/C++ Compiler v18.9.0.STS (or later), and with compiler switches --float_support = fpu64. This will generate C28x native double-precision floating-point instructions. It may be helpful to refer to #GUID-BC8C6374-4B4C-4649-B574-974D746FBD8C/GUID-063F2853-70D6-424A-A77B-32A57EB47557 for those unfamiliar with the size of the standard C data types on a C28x CPU. Users who want to write hand-optimized assembly for FPU64 may do so easily as well. There is roughly a one-to-one correspondence between single-precision and double-precision floating point assembly instructions. For the complete instruction set and details, see the document referenced in the documentation section. Software examples are listed in a section below that point to DSP and Math double-precision floating-point hand-optimized assembly routines that users can use for application development.

Table 3-4 Data Type Size C28x vs Arm
Data Type C28x Bit Length (EABI) Arm Bit Length
char 16 8
short 16 16
int 16 32
long 32 32
long long 64 64
float 32 32
double 64(1) 64
long double 64 64
pointer 32 32
(1) C28x COFF treats double as 32-bits

With the advent of auto code generation tools like MATLAB’s Embedded Coder, many customers are migrating to them for auto code generation. Since MATLAB uses double-precision floating-point by default, it becomes easier to port code from a simulation environment to an embedded environment without having to revalidate operating performance of the system due to a reduction in precision. This is another key benefit of using the FPU64.