Re: [code-review] Benchmarking concatenation vs. interpolation

------=_20031204232817_21123
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 8bit

Adam Turoff said:
> I have this benchmark program, and I'd like a sanity check:
> Is this a valid benchmark?

Nope, because you're declaring the variables inside benchmark subs. In the
case of the concat sub this doesn't have a major effect on things
(although, it /does/ have an effect, as we'll see later). However, with
the interp sub perl realises that you assign to the variable and then
immediately discard it, so it can optimise it away. If you define the
variables outside of the benchmark subs, then perl will leave things alone
and you will get a proper result.

(Also add abigails reasons)

I've created a modified version of your benchmark script (please find
attached), that carries out the following runs:

interp and concat
  as you wrote them (_adam)
  assigning to already existing variables (_knew)
  as abigail wrote them (_abigail)
  as abigail but defining the variables inside the subs (_abi_adam)


And here are the results (first % column only to reduce width):

                    Rate interp_knew
interp_knew     267724/s          --
concat_adam     274652/s          3%
concat_knew     281098/s          5%
interp_abigail  288919/s          8%
interp_adam     438631/s         64%
concat_abi_adam 458605/s         71%
interp_abi_adam 461936/s         73%
concat_abigail  480052/s         79%


So, the fastest method is apparently to use concatenation,but within one
assignment statement.


-- 
Jody
knew (at) pimb (dot) org
------=_20031204232817_21123
Content-Type: application/octet-stream; name="benchmark_test"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="benchmark_test"
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------=_20031204232817_21123--