18:10 <korvo> Increasingly finding a single particular mutable-bitset abstraction to be useful: https://bpa.st/SH5A

18:14 <nimaje> I think you have a bug with s = Bitset(); s.set(2); t = Bitset(); t.set(3); (s.intersect(t)).any()

18:18 <korvo> Nice.

18:18 <korvo> If only `or any(self.ds)` were legal RPython.

18:19 <nimaje> and no reset or toggle operation?

18:19 <korvo> Everything I've been doing is append-only, now that I think about it.

18:23 <nimaje> you could ensure that the highest part has to be non-zero in intersect, then bool(len(self.ds)) works for any, you just have to ensure that for every operation that could loose bits (and disallow Bitset([0]) or similar)

18:25 <korvo> Yep. Which would save memory, too.

18:25 <nimaje> ah, well l = [1]; s = Bitset(l); l[0] = 0; s.any() would still be buggy

18:31 <nimaje> and are you sure that you want to use SHIFT from rbigint instead of SHIFT=64 (or maybe even 128)? and intmask seems unused

18:42 <korvo> SHIFT is indeed 128 on one of my machines; it uses the xmm0, xmm1, etc. registers.

18:47 <nimaje> hm, did rbigint change since I last pulled? I remember SHIFT being 63 or 31 depending if you have support for int128 or not