Re: Efficiency and ReplacePart?
- To: mathgroup at smc.vnet.net
- Subject: [mg86167] Re: Efficiency and ReplacePart?
- From: Szabolcs <szhorvat at gmail.com>
- Date: Tue, 4 Mar 2008 02:09:11 -0500 (EST)
- References: <fq3ab7$fmg$1@smc.vnet.net>
On Feb 27, 10:25=A0am, "W. Craig Carter" <ccar... at mit.edu> wrote:
> If someone can answer the following question, I will have
> learned something about efficiency in mathematica...
>
> I'd like each second part of a list to a zero-list:
>
> ftemp = Table[{RandomReal[1, {2}], RandomReal[1, {2}]}, {50000}];
>
> Compare:
> 1)
> Timing[ftemp /. {{a_, b_}, {c_, d_}} :> {{a, b}, {0, 0}}][[1]]
>
> 2)
> Timing[Map[(# = {#[[1]], {0, 0}} &), ftemp]][[1]]
Did you mean Timing[Map[{#[[1]], {0, 0}} &, ftemp]][[1]] here?
>
> 3) Timing[Map[ReplacePart[#, 2 -> {0, 0}] &, ftemp]][[1]]
>
> (*
> or if you like:
> times[n_] :=
> =A0 Module[{ftemp =
> =A0 =A0 =A0Table[{RandomReal[1, {2}], RandomReal[1, {2}]}, {n}]},
> =A0 =A0{Timing[ftemp /. {{a_, b_}, {c_, d_}} :> {{a, b}, {0,
> 0}}][[1]],
> =A0 =A0 Timing[Map[(# = {#[[1]], {0, 0}} &), ftemp]][[1]]}]
>
> ListPlot[Table[times[i], {i, 100, 10000, 100}]]
> *)
>
> 1 is faster than 2 is faster than 3. Why?
I don't think that there is anyone who could look at these 3 programs
and predict their performance without running them. Internal
implementations change with different versions: in v5.2, 1 is more
than twice as slow as 2 on my machine. But a different machine (CPU)
may behave in a different way. Look at the benchmark reports included
with Mathematica:
<< Benchmarking`
BenchmarkReport @@ $BenchmarkSystems
The timing ratios between systems A and B for different tests may be
very different.
Also, with a list of length 50000, these tests finish very quickly on
my computer (below 1 second). Based on my previous experience, I am
reluctant to attribute any significance to these timings ... at least
in the case when the tests were run on a Windows computer. Try
pushing up the timing to at least several seconds, and see if the
results change. You may be surprised.