In C28x CPU-based projects, 8-bit data types are not supported. char is 16 bits wide and the types uint8_t and int8_t are not defined by the C28x compiler. C2000Ware remaps these to uint16_t and int16_t data types. For more details on the data types, see the TMS320C28x Optimizing C/C++ Compiler v22.6.0.LTS User's Guide.

However, the C28x compiler provides intrinsic __byte() for byte accesses. For more details, see the https://software-dl.ti.com/ccs/esd/documents/c2000_byte-accesses-with-the-c28x-cpu.html.

• While porting an application from an 8-bit addressable architecture, such as Arm to C28x, you may find an increase in memory requirement due to this difference.

  Example:

  struct
  {
     uint8_t a;
     uint8_t b;
     uint16_t c;
  } myStruct; 

  In an Arm device, the size of myStruct would be 8 + 8 + 16 = 32 bits. Whereas, in C28x, the size would be 16 + 16 + 16 = 48 bits.

• Since the int8 or char data types in C28x is 16 bit wide, the compiler will not do a wraparound at 0xFF while performing arithmetic or shift operations.

  Example:

  uint8_t a = 0xFF;
  a += 1;
  if (a == 0)
  {
       //Condition is true for Arm and false for C28x
  }

• While converting larger data types to smaller ones or vise versa, care should be taken to properly typecast, use masks, and be aware of the device endianness. The C28x-based devices are little endian devices.
  
• Usage of unions needs to be revisited.