fixed point arithmetic

[umlaeute]

while packaging faust-2.74.6 for Debian, I tried to test the new faust2comparator with faust2comparator foo.dsp -fx-size 48 as described in

faust/architecture/comparator/README.md

Line 9 in b498ef5

For example, to impose a fixed width of 48 to fixed-point types in `foo.dsp`, run `faust2comparator foo.dsp -fx-size 48`

however, this gives me build errors:

1. it cannot find float.cpp (which is generated in the cwd, so adding -I. to the compiler invocation fixes the problem)
2. however, float.cpp then goes on to include ap_fixed.h, which is supposed to live in /usr/local/include/ap_fixed/; at least if I read the compiler invocation correctly:

   faust/tools/faust2appls/faust2comparator

   Line 60 in b498ef5

   $CXX $CXXFLAGS -g "$FAUSTARCH/comparator/compclass.cpp" "-I/usr/local/include/ap_fixed" `pkg-config --cflags --static --libs sndfile jack gtk+-2.0` -o "${f%.dsp}"

now, I do not have a /usr/local/include/ap_fixed/ap_fixed.h file (nor the entire directory).

where does it come from?
probably something like https://github.com/Xilinx/HLS_arbitrary_Precision_Types.git, but using the Xilinx headers gives me template-errors (unfortunately i didn't write them down before deleting the container).

It would be nice if there were some hints on how to use the fixed-point backends (e.g. what to install where)

[sletz]

The -fx option is not considered usable for now.

[umlaeute]

thanks for the quick response. what would be a usable invocation of faust2comparator then?

EDIT: or is faust2comparator not usable right now? (in which case I would just exclude it from the packaging)

[sletz]

Probably nothing for now.