Markus Wallerberger

I implemented something here which seems to work: https://github.com/tuwien-cms/xprec/blob/v1.2.13/csrc/dd_arith.c#L162 Might be worth taking a look at ....

@oscardssmith I took the `exp` from the QD library, so that problem is "imported" from there. For me, 10 ulps are acceptable, but of course you guys may feel differently....

@aptthorpe: > attempt to recover from the IndexError exception in RegexLexer::line_tokens(), try to detect a Hollerith field with a character-by-character scan or multiple-regex-pass, and continue on to raising the LexerError...

No, this is not implemented yet. Thanks for the pointer!

Hmm, the problem is that the standard specifies that `include` is not a statement at all -- it is called "include line" and is much closer in spirit to a...

Here's a C++ implementation of `expm1` and `exp` that is accurate to 2 ulps across the range and takes around 400 flops: https://github.com/tuwien-cms/libxprec/blob/v0.4.2/src/exp.cxx The idea is similar to what you...

Hi Hugo, I think the `G4iw(X)` have been renamed, as you also suspect, to `P(X+1)iwPH`, that would mean `P3iwPH` is what you want. In other words, its a two-frequency object...

~Doesn't this result in a mandatory Ninja dependency?~ Ah, no, this should still work with cmake ... could we check both as part of the CI?