Re: [FWP] map BLOCK LIST vs. map EXPR, LIST

[Ronald J Kimball <rjk@linguist.dartmouth.edu>]

On Thu, Apr 13, 2000 at 04:59:07PM -0400, Jeff Pinyan wrote:
> ---------- Forwarded message ----------
> From: Jeff Pinyan <jeffp@crusoe.net>
> To: Bennett Todd <bet@rahul.net>
> 
> >I only ever used map BLOCK LIST, never paid any attention to the
> >alternative. Makes sense that EXPR, LIST would run faster.
> 
> I wince slightly when I see constructs like
> 
>   @silly_way_to_unpack = map { ord } split //, $name;
> 
> (Yes, poor example, since unpack("C*", $name) is much better.)
> 
> I've done benchmarks that show that
> 
>   map ord, split //, $name;
> 
> is faster.

I don't see it.

#!perl

use Benchmark;

$name = join '', 'a' .. 'z';

timethese(1 << shift,
          {
           expr  => sub { @tmp1 = map   ord,  split //, $name; },
           block => sub { @tmp2 = map { ord } split //, $name; },
          }
         );

__END__

% perl bm.pl 11
Benchmark: timing 2048 iterations of block, expr...
     block:  1 wallclock secs ( 0.42 usr +  0.00 sys =  0.42 CPU)
      expr:  1 wallclock secs ( 0.55 usr +  0.00 sys =  0.55 CPU)
% perl bm.pl 12
Benchmark: timing 4096 iterations of block, expr...
     block:  1 wallclock secs ( 0.96 usr +  0.01 sys =  0.97 CPU)
      expr:  2 wallclock secs ( 0.96 usr +  0.00 sys =  0.96 CPU)
% perl bm.pl 13
Benchmark: timing 8192 iterations of block, expr...
     block:  3 wallclock secs ( 1.93 usr +  0.01 sys =  1.94 CPU)
      expr:  3 wallclock secs ( 1.79 usr +  0.01 sys =  1.80 CPU)
% perl bm.pl 14
Benchmark: timing 16384 iterations of block, expr...
     block:  5 wallclock secs ( 3.72 usr +  0.01 sys =  3.73 CPU)
      expr:  5 wallclock secs ( 3.59 usr +  0.01 sys =  3.60 CPU)
% perl bm.pl 15
Benchmark: timing 32768 iterations of block, expr...
     block: 11 wallclock secs ( 7.59 usr +  0.03 sys =  7.62 CPU)
      expr: 10 wallclock secs ( 7.66 usr +  0.02 sys =  7.68 CPU)
% perl bm.pl 16
Benchmark: timing 65536 iterations of block, expr...
     block: 20 wallclock secs (15.30 usr +  0.05 sys = 15.35 CPU)
      expr: 20 wallclock secs (15.34 usr +  0.05 sys = 15.39 CPU)
% 


And I don't think it would make sense for map EXPR, LIST to run faster than
map { EXPR } LIST.  Either way Perl just applies EXPR to each element of
the list.


Ronald

==== Want to unsubscribe from Fun With Perl?  Well, if you insist...
==== Send email to <fwp-request@technofile.org> with message _body_
====   unsubscribe