Re: What is the runtime complexity of $(filter )?

[Peng Yu]

On Fri, Jan 8, 2010 at 12:34 PM, Paul Smith <address@hidden> wrote:
> On Sat, 2010-01-09 at 11:33 +1800, Peng Yu wrote:
>> I run $(filter ) with each argument of thousands of strings. It runs
>> slow. I'm wondering what the run time complexity of $(filter ) is.
>> Suppose I have n strings for each argument. Is the run time complexity
>> n*n or log(n)?
>
> If you run $(filter <N-strings>,<M-strings>) then the worst-case
> complexity is N*M of course.
>
> However, make does try to be smart.  If N and M are large enough (and
> actually, "large enough" is pretty small: if N>1 and N*M>9) then instead
> of doing the matching linearly, make will add the words (M, above) into
> a hash table and do hash lookups for matching rather than walking an
> array.
>
> This is true in make 3.81 and above; I didn't check 3.80 or below.

I want to measure much the runtimes for each of the three operations
that computes A, B and C. I'm wondering how to do it with gnu make.


$ cat Makefile
.PHONY: all

A=$(patsubst %.1, %, $(wildcard *.1))
B=$(patsubst %.2, %, $(wildcard *.2))
C=$(filter $(A),$(B))

all:
        echo $(C)

a:
        @touch a.1 b.1 c.1

b:
        @touch a.2 b.2 d.2