Jeffrey, Absolutely! I think I am seeing my C++ QuantLib code reach a minimum that is much smaller (roughly 1e-6) vs the Julia implementation (1e-4) because of some precision issues.
Thanks! Chris On Thursday, December 17, 2015 at 7:34:54 PM UTC-5, Jeffrey Sarnoff wrote: > > Chris, > > That package is no longer supported and it is missing division, sqrt .. > > Would a double-double type supporting (+),(-),(*),(sqrt) suffice for your > work? > I have written that as part of something else, and am willing to separate > it out for others to use. > > Jeffrey > > > On Thursday, December 17, 2015 at 5:15:33 PM UTC-5, Christopher Alexander > wrote: >> >> I looked into using BigFloat, but the overhead majorly slowed everything >> down. I saw there was a package DoubleDouble that was supposed to do that >> level of precision but without the overhead, but it doesn't seem to have >> been updated and I couldn't get it to work with 0.4.2. Maybe I could try >> to get it up to speed. >> >> Chris >> >> On Thursday, December 17, 2015 at 5:05:23 PM UTC-5, Stephan Buchert wrote: >>> >>> In Julia there would be also BigFloat: >>> >>> julia> x=parse(BigFloat, "1.052500000000004") >>> >>> 1.052500000000003999999999999999999999999999999999999999999999999999999999999994 >>> >>> julia> x^6.0-x^6 >>> >>> 0.000000000000000000000000000000000000000000000000000000000000000000000000000000 >>> >>
