Two numbers

2.8318757E10
2.734893e10

The first is FLOPS for LW.  The second for SBCL.  Same source code, M1 using four cores, speed 3 safety 0 for both and float 0 for LW.

This is for a version of my naive-gravitational-many-body benchmark (naive means it very much doesn't use any fancy approach, intentionally) using my SOA code (coming fairly soon to a repo near you) instead of its ancestor.  I don't know what's changed, as SBCL used to win by a noticable amount: the only thing I can think of is that LW does slightly better 2d-array access (the old code was all 1d arrays, SOA uses rank n+1 where the 'user rank' of the array is n.

I always used to think that LW was significantly slower than SBCL, but worth it for the much nicer experience.  Turns out not always.

--tim


_______________________________________________
Lisp Hug - the mailing list for LispWorks users
[email protected]
http://www.lispworks.com/support/lisp-hug.html