Iain Buclaw
Iain Buclaw
> it breaks some parts of std.math which currently work in CTFE. I don't particularly buy that, as many functions that do pointer casting are not CTFE-able anyway. Typical getter/setters...
I'm in the middle of testing ieeeQuadruple and ibmExtended, and finding bugs in both. I'd want to make sure that is complete before introducing yours. To make sure we start...
Rebased, but I'll have to double-check that ibmExtended is still fine (did some educated guesswork on reimplementing the floorImpl I added in the new format).
Is this part of a zlib release?
> > Is this part of a zlib release? > > Not yet, no. The latest release is 1.2.12, which we already have in Phobos. These commits are from the...
Looks like some testsuites in the wild assume x87 precision in pow() too, despite using doubles from the outside.
Testsuite failure will be fixed in dlang/dmd#14270.
> Thanks for pushing this forward! This was one of the things on my personal backlog to tackle grin > > It's now time for LLVM/GCC intrinsics to shine. >...
> `std/internal/math/gammafunction.d(569): [unittest] unittest failure` > `assert(logGamma(-50.2L) == log(fabs(gamma(-50.2L))));` > e: relaxed to > `assert(feqrel(logGamma(-50.2L),log(fabs(gamma(-50.2L)))) > real.mant_dig-2);` I'll have to check, but I think that similar relaxations also exist in...