Giovanni Rovatti wrote:> Is there any documentation about it?
Not that I'm aware of.
> So far I'm building with the TI_C5X_ASM;CONFIG_TI_C55X #define.
I was able to get it to build with the following in my config.h:
#define FIXED_POINT
#define DOUBLE_PRECISION
#define MIXED_PRECISION
#define DISABLE_FLOAT_API
#define ENABLE_TI_DSPLIB
#define TI_C5X_ASM
#define CONFIG_TI_C55X
You may also want to use
#define STATIC_MODES
The last one will eliminate all floating point routines so that you
don't have to link against the FPU emulation library, giving a
significant code size reduction. In order to use that, you have to do a
build (on your development machine, not the DSP, should be fine) without
--enable-static-modes and run dump_modes to generate the .c and .h files.
Note that if you are using the free Linux tools from TI (though I think
they were recently taken off the TI website), and not
CodeComposerStudio, a large number of their compiler intrinsics are
simply missing (_lmpy, _lmpyu, and _lmpysu, at a minimum). You will have
to comment out the macros that use them in fixed_c5x.h.