[16:37:23] <MasterDuke> i'm not subscribed to the mailing list but maybe someone who is could chime in to this thread https://www.nntp.perl.org/group/perl.perl6.users/2021/04/msg9878.html and mention that i believe the problem is in libtommath's mp_get_double and the gmp branch fixes it?
> On 14 Apr 2021, at 04:09, sisyphus <sisyphus...@gmail.com> wrote: > > Hi, > > C:\>raku -e "say 1.8446744073709552e+19 == 18446744073709551615" > True > > I think I understand why raku deems this to be true. > The LHS is 0x1p+64, which is identical to the double that the RHS rounds to. > (AFAIK, it's the same with perl5 and C.) > > Of course, for all $x in the Int range of 18446744073709551615 - 1023 to > 18446744073709551615, we find that "1.8446744073709552e+19 == $x" is True, > because each of those $x Int values round to 0x1p+64. > I understand all of that (I think). > > What I don't understand is the following: > > C:\>raku -e "say 1.8446744073709552e+19 == (18446744073709551615 - 1024)" > True > > C:\>raku -e "say 1.8446744073709552e+19 == (18446744073709551615 - 1087)" > True > > For those particular Int values (and all Int values in between) the RHS > rounds to 0x1.fffffffffffffp+63, > and the result should therefore be "False" ... but it is reported as "True". > > It's not until we subtract at least 1088 that the behaviour I'm expecting > appears: > > C:\>raku -e "say 1.8446744073709552e+19 == (18446744073709551615 - 1088)" > False > > Comparing the RHS Int values given in the last 2 one liners, I find they are > (respectively): > 0.1111111111111111111111111111111111111111111111111111101111000000E64 > 0.1111111111111111111111111111111111111111111111111111101110111111E64 > > It's not apparent to me how those 2 integer values can round to different > doubles. > Those 2 integer values can also be written as (respectively): > 18446744073709550528 > 18446744073709550527 > > On Windows, I have: > C:\>raku --version > This is Rakudo version 2020.05.1 built on MoarVM version 2020.05 > implementing Raku 6.d. > > On Ubuntu, I have: > $ raku --version > Welcome to Rakudo(tm) v2021.03. > Implementing the Raku(tm) programming language v6.d. > Built on MoarVM version 2021.03. > > It's the same behaviour on both systems. > > Is this a bug that I ought to report ? > > Cheers, > Rob > > > > > >
