Regarding performance, we have tested this with Csound and in that system, doubles are actually faster in 64bit systems (but slower, not by much, in 32-bit). Audio quality has reportedly been much higher.
Victor ----- Original Message ----- From: Pedro Lopez-Cabanillas <address@hidden> Date: Tuesday, September 28, 2010 10:30 pm Subject: [fluid-dev] floats or doubles? To: address@hidden
> Hi, > > The old autotools build system defaults to using single > precision floating > point numbers in DSP calculations and buffers (symbol WITH_FLOAT > defined in > config.h) but there is an alternative, if the user provides to > the "configure" > script the argument "--with-doubles=no", that in despite of > logic has the > effect of undefining WITH_FLOAT in config.h, enabling double > precision > floating point numbers. > > The new CMake build system uses by default double precision > floating point > numbers, but there is an option ("-Denable-floats=yes" as a > command line > argument) that defines WITH_FLOAT in config.h forcing the usage > of single > precision floating point numbers in DSP buffers and calculations. > > Double precision real numbers are bigger, and probably have > worse performance > than single precision floats, but maybe they can produce better > sound in some > circumstances. Floats are probably safer because they have been > tested more > deeply, as the ticket #87 has shown. > > So, what do you think? leave doubles as default in CMake, or > change the > default to floats, like in the old autotools? > > Regards, > Pedro > > _______________________________________________ > fluid-dev mailing list > address@hidden > http://lists.nongnu.org/mailman/listinfo/fluid-dev
Dr Victor Lazzarini, Senior Lecturer, Dept. of Music, National University of Ireland, Maynooth