Re: LuaJIT string concat

[Francesco Abbate <francesco.bbt@...>]

2011/4/13 Geoff Leyland <geoff_leyland@fastmail.fm>:
> Why not profile it? (you'll need to look at the output in a monospaced font)
>
> $ lua -luatrace.profile concat-bench.lua
> ...
>
> ==================================================================================
> Visits  Total   Self  Child Line | concat-bench.lua - Times in milliseconds
> ----------------------------------------------------------------------------------
>     1   0.00   0.00   0.00    1 | local t = {}
>   101   0.01   0.01   0.00    2 | for i = 1, 100 do
>   100   0.25   0.25   0.00    3 |   t[i] = "abc"..tostring(i)
>     .      .      .      .    4 | end
>     .      .      .      .    5 |
>     .      .      .      .    6 | local sformat, rnd, tostring, a =
>     1   0.00   0.00   0.00    7 |       string.format, math.random, tostring, nil
>     .      .      .      .    8 |
> 100001   9.28   9.28   0.00    9 | for i= 1,100000 do
> 100000 648.02 648.02   0.00   10 |   local s = table.concat(t, ",")
> 100000  96.93  96.93   0.00   11 |   a = sformat("[%s]", s)
>     .      .      .      .   12 | end
>     .      .      .      .   13 |
> 100001   9.40   9.40   0.00   14 | for i= 1,100000 do
> 100000 644.82 644.82   0.00   15 |   local s = table.concat(t, ",")
> 100000  60.55  60.55   0.00   16 |   a = "["..s.."]"
>     1   0.00   0.00   0.00   17 | end

Hi all,

I see that the benchmark tell that the concatenation method is a
little bit faster. I'm a little bit surprised but it is true that for
the case with very small strings many small elements came to play to
determine the final timing and the benchmarks can contradict some
common beliefs.

My only doubt now is what happens is the string is big... I was
wondering if in this case the results can be different but may be that
I still have some wrong belief!! :-)


-- 
Francesco