On Thursday, June 5, 2014 3:17:35 PM UTC+2, Jameson wrote:
> n should be a global const or a function parameter, so that you are
> testing array performance rather than global variable performance, before
> drawing any conclusions
>
Ah!
My revised results, then (with the updated Gist):
| Row # | Function | Average | Relative | Replications |
| 1 | "push_int_any" | 0.00378676 | 2.88618 | 100 |
| 2 | "asgn_int_any" | 0.00350896 | 2.67445 | 100 |
| 3 | "asgn_check_int_any" | 0.00333424 | 2.54128 | 100 |
| 4 | "asgn_check_id_uint" | 0.00462921 | 3.52829 | 100 |
| 5 | "push_id_uint" | 0.00516361 | 3.93559 | 100 |
| 6 | "push_int_uint" | 0.00558508 | 4.25682 | 100 |
| 7 | "push_entry_any" | 0.00131203 | 1.0 | 100 |
| 8 | "push_enryid_uint" | 0.00133314 | 1.01609 | 100 |
| 9 | "push_int_int" | 0.00141462 | 1.0782 | 100 |
| 10 | "push_uint_uint" | 0.00142525 | 1.08629 | 100 |
Actually, I ran these with @inbounds – running without it was slightly
faster, for some reason. Possibly noise. (I would have thought there should
have ben a noticable improvement when using @inbounds?)
Hm. I'm still a bit confused. But the easiest solution (wrt. the rest of my
code) of simply pushing objects/pointers onto an Array{Any} doesn't seem
like an obviously bad choice, at least.
