SVectors + @reset from Accessors: Strange Benchmarks

General Usage
, , ,
5

I agree, something more is going on. I found this from a couple years ago which probably explains it:

In other words, I think @set a[i].s = 1 is interpreted as : “give me a copy of a in which the x field of the i-th element is set to 1”. As opposed to “give me a copy of a[i] in which the x field is set to 1” (which would be the desired semantics here).
Container of mutable structs without 100x slowdown? - #5 by ffevotte

It looks like a proposed fix was rejected?