CCS/TMS320VC5402: cast from float to short

Glen Wardrop said:

I can see compiler is calling function F$$FTOI  followed by SSBX SXM 

What I have been unable to find is documentation on either  of these functions..

The function F$$FTOI is called by the compiler to convert a 32-bit float to a 16-bit integer (or short).  The source code to this function is in the compiler package, in the file f_ftoi.asm.  That file is located in the zip file compiler_install_root/lib/rtssrc.zip.  

The instruction SSBX SXM sets the sign extension mode bit in a status register.  More details can be found in the C54x Mnemonic Instruction Set manual.

Glen Wardrop said:

It appears that data that exceeds short integer range may be clipped which would be great if I knew for sure it will always happen.

It does.  The comments in f_ftoi.asm say as much.  

Glen Wardrop said:

Original C code is equivalent to....

To make it compile I changed your C code to ...

short doit (const float value)
{
   return (short) (value * 32768.0F);
}

When I compile this code, I do not get a SSBX instruction.  I doubt this is relevant to this problem.  But I thought I would point it out anyway.

Thanks and regards,

-George