Floats conversion

HI Adam,

I'm not sure if I understand what you need, but OMAP-L137 Native support for IEEE 754 Single Precision and Double Precision floating point instructions...

Hi Mariana,

I have got float data in file created on PC computer (not text). I think float bit organisation for PC platform is diferent to DSP processor. In IEEE 754 standard 31-bit contain sign of float then exponent and fraction, I am not sure but DSP float has diferent order - exponent - sign - fraction. Several years ago I used TIC compiler for C40x processors containing library with functions vieee2flt() that convert PC float to DSP float. I need equivalent of vieee2flt() for CCS OMAP-L137. I am looking for eguivalent of TIC compiler library consist vcoordtrans3Dsf(), vsub3f(), mtransposef(). My next question is: what size is float defined in CCS for OMAP-L137, 32 or 40 bit?

The C40 did not use IEEE standard floats and hence you needed the function.  The DSP core (C674x) inside the OMAP-L137 uses IEEE standard floats.  Here's a screenshot from the "DSP Features and Options" section of the C674x CPU and Instruction Set Reference Guide.

Hi Brad,

You are right, but what about 40 bit precision operation on C++ CCS level.

I can't speak to C++ as I do everything in C.  From a C perspective we have type "long" which is a 40-bit integer.  In terms of floating point we have only IEEE standard floats (32-bit single and 64-bit double).  There is not a 40-bit floating point data type.

The C3x/C4x floating point forman (non-IEEE) is documented on p. 114 of http://focus.ti.com/lit/ug/spru031f/spru031f.pdf

Later in Chapter 5 are example of how to convert to/from IEEE.