Free42 possible accuracy flaw

03222022, 05:39 PM
(This post was last modified: 03222022 05:43 PM by Thomas Okken.)
Post: #4




RE: Free42 possible accuracy flaw
I'm reluctant to simply remove the repeatedsquaring logic. That logic is there for a reason, bid128_pow() is not very good when raising integers to integer powers, and the 2.0u2 revision of the library even includes a patch that does the same thing as the workaround I'm applying to 2.0u1. But it does make sense that bid128_pow() without the workaround works well for (1+1e8)^1e8, since it does seem to specifically handle bases close to 1. So the question to me is when to use the repeatedsquaring workaround and when not to, and that question is relevant to the binary version as well, since the standard library's pow() has the same problem, so there's the question when to use pow() and when to use log1p() and expm1().


« Next Oldest  Next Newest »

User(s) browsing this thread: 1 Guest(s)