aero · December 14, 2015 07:59 · yongbin · Feb 28, 2013 · yongbin · Feb 28, 2013
diff --git a/Inline_bench.pl b/Inline_bench.pl
 #!/usr/bin/env perl
 use 5.010;
 use strict;
 use warnings;
 use Benchmark;
 use Data::Dumper;

 timethese(1000000, {
        perl       => sub { sub_perl(100) },
        c          => sub { sub_c(100) },
 });

 my @a = (1..500);
 my @b = (501..1000);

 #say Dumper( sub_perl_array(scalar @a, @a, @b) );
 #say Dumper( sub_c_array(scalar @a, @a, @b) );

 timethese(1000, {
        perl_array => sub { sub_perl_array(scalar @a, @a, @b) },
        c_array    => sub { sub_c_array(scalar @a, @a, @b) },
 });


 sub sub_perl {
    return $_[0]+1;
 }
 sub sub_perl_array {
    my ($n) = @_;
    my @ret;
    foreach (1..$n) {
        push @ret, ( $_[$_] * $_[$_+$n] );
    }
    return \@ret;
 }

 use Inline C => <<'END_C';

 int sub_c (int i) {
    return i+1;
 }

 AV * sub_c_array(SV * x, ...) {
     Inline_Stack_Vars;
     long array_size, i;
     AV * ret = newAV();

     array_size = SvUV(Inline_Stack_Item(0));

 /* check that array_size matches our expectations,
   based on the total number of arguments supplied */
   if(Inline_Stack_Items != (2 * array_size) + 1)
     croak("Mismatch in number of arguments supplied to foo()");

 /* for efficiency, make the array big enough in one fell swoop */
     av_extend(ret, array_size - 1);

 /* multiply corresponding array elements and push the result into ret +*/
     for(i = 1; i <= array_size; i++)
         av_push(ret, newSVnv(SvNV(Inline_Stack_Item(i)) *
                              SvNV(Inline_Stack_Item(i + array_size))));

     return ret;
 }

 END_C


 __END__
 <result>
 Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz:
 Microsoft Windows 7 Enterprise K

 Benchmark: timing 1000000 iterations of c, perl...
         c:  1 wallclock secs ( 0.23 usr +  0.00 sys =  0.23 CPU) @ 4273504.27/s (n=1000000)
            (warning: too few iterations for a reliable count)
      perl:  0 wallclock secs ( 0.25 usr +  0.00 sys =  0.25 CPU) @ 4000000.00/s (n=1000000)
            (warning: too few iterations for a reliable count)
 Benchmark: timing 1000 iterations of c_array, perl_array...
   c_array:  0 wallclock secs ( 0.06 usr +  0.00 sys =  0.06 CPU) @ 16129.03/s (n=1000)
            (warning: too few iterations for a reliable count)
 perl_array:  1 wallclock secs ( 0.17 usr +  0.00 sys =  0.17 CPU) @ 5847.95/s (n=1000)
            (warning: too few iterations for a reliable count)
	#!/usr/bin/env perl
	use 5.010;
	use strict;
	use warnings;
	use Benchmark;
	use Data::Dumper;

	timethese(1000000, {
	perl => sub { sub_perl(100) },
	c => sub { sub_c(100) },
	});

	my @a = (1..500);
	my @b = (501..1000);

	#say Dumper( sub_perl_array(scalar @a, @a, @b) );
	#say Dumper( sub_c_array(scalar @a, @a, @b) );

	timethese(1000, {
	perl_array => sub { sub_perl_array(scalar @a, @a, @b) },
	c_array => sub { sub_c_array(scalar @a, @a, @b) },
	});


	sub sub_perl {
	return $_[0]+1;
	}
	sub sub_perl_array {
	my ($n) = @_;
	my @ret;
	foreach (1..$n) {
	push @ret, ( $_[$_] * $_[$_+$n] );
	}
	return \@ret;
	}

	use Inline C => <<'END_C';

	int sub_c (int i) {
	return i+1;
	}

	AV * sub_c_array(SV * x, ...) {
	Inline_Stack_Vars;
	long array_size, i;
	AV * ret = newAV();

	array_size = SvUV(Inline_Stack_Item(0));

	/* check that array_size matches our expectations,
	based on the total number of arguments supplied */
	if(Inline_Stack_Items != (2 * array_size) + 1)
	croak("Mismatch in number of arguments supplied to foo()");

	/* for efficiency, make the array big enough in one fell swoop */
	av_extend(ret, array_size - 1);

	/* multiply corresponding array elements and push the result into ret +*/
	for(i = 1; i <= array_size; i++)
	av_push(ret, newSVnv(SvNV(Inline_Stack_Item(i)) *
	SvNV(Inline_Stack_Item(i + array_size))));

	return ret;
	}

	END_C


	__END__
	<result>
	Intel(R) Core(TM)2 Quad CPU Q9400 @ 2.66GHz:
	Microsoft Windows 7 Enterprise K

	Benchmark: timing 1000000 iterations of c, perl...
	c: 1 wallclock secs ( 0.23 usr + 0.00 sys = 0.23 CPU) @ 4273504.27/s (n=1000000)
	(warning: too few iterations for a reliable count)
	perl: 0 wallclock secs ( 0.25 usr + 0.00 sys = 0.25 CPU) @ 4000000.00/s (n=1000000)
	(warning: too few iterations for a reliable count)
	Benchmark: timing 1000 iterations of c_array, perl_array...
	c_array: 0 wallclock secs ( 0.06 usr + 0.00 sys = 0.06 CPU) @ 16129.03/s (n=1000)
	(warning: too few iterations for a reliable count)
	perl_array: 1 wallclock secs ( 0.17 usr + 0.00 sys = 0.17 CPU) @ 5847.95/s (n=1000)
	(warning: too few iterations for a reliable count)