Hi Josh, Thank you for the pointers. Although I'm also interested in implementing quad-floats using doubles, such a correct implementation would take much more time than the FFI approach as well as what I can afford given my time constraints.
Shaobo Josh Rubin <[email protected]> 于2019年8月23日周五 下午7:08写道： > > On 8/23/2019 8:39 PM, Shaobo He wrote: > > Hello everyone, > > > > I'm in need of quad-precision floating-point type/operations in Racket > > (the bigfloat module doesn't work for me). It appears there's none, am > > I right? > > > > If so, I think I will try to write one myself, which is going to be a > > wrapper to libquadmath. Any suggestions about how I should approach > > it? The first issue seems to be marshalling the C `__float128` type > > into a Racket type. > > > > Thanks, > > Shaobo > > I'm not sure exactly what you have in mind, and I am a total ignoramus > about Racket numerics, but it is possible to implement quad precision > floating point arithmetic using only double precision floating point, > while maintaining rigorous error bounds. > (... and so on, doubling precision precision at each step.) > > These sources might be useful. > > Design and Implementation of a High Precision Arithmetic with Rigorous > Error Bounds > Alexander WittigMSUHEP-081126December 2008 > > https://pdfs.semanticscholar.org/8450/3383403eca34dd7e92a12261a3ed7ea2340a.pdf > > The Great Internet Mersenne Prime Search (GIMPS) project carries this to > extremes. > They use Fourier multiplication on integers with millions of bits - and > must have exact results. > https://www.mersenne.org/ > > Good luck with your project. I am interested in seeing the result! > > -- > > Josh Rubin > [email protected] > > > -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/CAMePyd-SvEtjqsNaaVa7MsF0MTLX2wmmhpCO-neiUdiNTAQk_A%40mail.gmail.com.
