Very interesting project, bravo !
Faust developer here, one remark: AFAICS you don't need to add the -ftz 2 option when compiling DSP to C++. The point is that when deploying native code, the FTZ will be done at hardware level (this is done automatically by JUCE in the audio thread AFAICS) and will be much better and faster.
The -ftz xx option in the Faust compiler has been developed to be used when compiling to WebAssemby, where hardware FTZ does not exits. And we know that this software-ftz xx is even not perfect.
Very interesting project, bravo ! Faust developer here, one remark: AFAICS you don't need to add the
-ftz 2
option when compiling DSP to C++. The point is that when deploying native code, the FTZ will be done at hardware level (this is done automatically by JUCE in the audio thread AFAICS) and will be much better and faster. The-ftz xx
option in the Faust compiler has been developed to be used when compiling to WebAssemby, where hardware FTZ does not exits. And we know that this software-ftz xx
is even not perfect.