SLAA534A Application note

SLAA534A June 2013 – June 2020

14.1 Copy Table Format

A copy table has the following format:

        typedef struct
        {
           uint16      rec_size;
           uint16      num_recs;
           COPY_RECORD recs[num_recs];
        } COPY_TABLE;

rec_size is a 16-bit unsigned integer that specifies the size in bytes of each copy record in the table.

num_recs is a 16-bit unsigned integer that specifies the number of copy records in the table.

The remainder of the table consists of a vector of copy records. The format of the COPY_RECORD structure depends on the code and data model being used.

For the small data model and small code model:

        typedef struct
        {
           void  * load_addr; /* 16-bit pointer */
           void  * run_addr;  /* 16-bit pointer */
           uint16  size;
        } COPY_RECORD;

For the small data model and large code model:

        typedef struct
        {
           uint32  load_addr; /* 32-bit storage for data or code pointer */
           uint32  run_addr;  /* 32-bit storage for data or code pointer */
           uint32  size;
        } COPY_RECORD;

For the large (or restricted) data model and large code model:

        typedef struct
        {
           void  * load_addr; /* 20-bit pointer */
           void  * run_addr;  /* 20-bit pointer */
           uint32  size;
        } COPY_RECORD;

The load_addr field is the address of the source data in offline storage.

The run_addr field is the destination address to which the data will be copied.

The size field is overloaded:

If the size is zero, the load data is compressed. The source data has a format-specific encoding that implies its size. In this case, the first byte of the source data encodes the compression format. The format is encoded as an index into the handler table, which is a table of pointers to handler routines for each format in use.
If the size is non-zero, the source data is the exact image of the data to copy; in other words, it is not compressed. The copy-in operation is to simply copy sizebytes from the load address to the run address.

The rest of the source data is format-specific. The copy-in routine reads the first byte of the source data to determine its format/index, uses that value to index into the handler table, and invokes the handler to finish decompressing and copying the data.

The handler table has the following format:

Figure 14-1 Handler Table Format

The copy-in routine references the table via special linker-defined symbols as shown. The assignment of handler indexes is not fixed; the linker reassigns indices for each application depending on what decompression routines are needed for that application. The handler table is generated into the .cinit section of the executable file.

The run-time support library in the TI toolchain contains handler functions for all the supported compression formats. The first argument to the handler function is the address pointing to the byte after the 8-bit index. The second argument is the destination address.

Reference Implementation of Copy-In Function provides a reference implementation of the copy_in function:

Reference Implementation of Copy-In Function

typedef void (*handler_fptr)(const unsigned char *src, unsigned char *dst);
extern int __TI_Handler_Table_Base;
void copy_in(COPY_TABLE *tp)
{
   unsigned short i;
   for (i = 0; i < tp->num_recs; i++)
   {
      COPY_RECORD crp = tp->recs[i];
      const unsigned char *ld_addr = (const unsigned char *)crp.load_addr;
      unsigned char       *rn_addr = (unsigned char *)crp.run_addr;
      if (crp.size)       // not compressed, just copy the data.
         memcpy(rn_addr, ld_addr, crp.size);
      else                // invoke decompression routine
      {
         unsigned char index = *ld_addr++;
         handler_fptr  hndl  = ((handler_fptr *)(__TI_Handler_Table_Base))[index];
        (*hndl)(ld_addr, rn_addr);
      }
   }
}