Re: Re: Eek! goto?

[BrowserUk]

Updated: now I got pfaut's code to go.

Neat! And faster (just) than the goto.

However, with some adaption, pfaut's insight at Re: Eek! goto? allow's me to double the speed for this bit and save cycles earlier by avoiding spliting or unpacking the string to emulate the char[].

           Rate demerphq   gotoit    pfaut
demerphq  916/s       --      -1%     -49%
gotoit    923/s       1%       --     -49%
pfaut    1809/s      98%      96%       --

Benchmark and full results

c:\test>
#! perl -slw
use strict;
use Benchmark qw[cmpthese];

sub demerphq {
    my ($len, @k) = @_;
    my ($a, $b, $c) = (0)x3;
    {
        ($len       or last) and $a+= $k[0];
        ($len >=  2 or last) and $a+= $k[1]  <<8;
        ($len >=  3 or last) and $a+= $k[2]  <<16;
        ($len >=  4 or last) and $a+= $k[3]  <<24;
        ($len >=  5 or last) and $b+= $k[4];
        ($len >=  6 or last) and $b+= $k[5]  <<8;
        ($len >=  7 or last) and $b+= $k[6]  <<16;
        ($len >=  8 or last) and $b+= $k[7]  <<24;
        ($len >=  9 or last) and $c+= $k[8]  <<8;
        ($len >= 10 or last) and $c+= $k[9]  <<16;
        ($len >= 11 or last) and $c+= $k[10] <<24;
    }
    return ($a+$b+$c);
}

sub gotoit {
    my ($len, @k) = @_;
    my ($a, $b, $c) = (0)x3;

    goto "LAST".$len;
    LAST11: $c+= $k[10] <<24;
    LAST10: $c+= $k[9]  <<16;
    LAST9:  $c+= $k[8]  <<8;
    LAST8:  $b+= $k[7]  <<24;
    LAST7:  $b+= $k[6]  <<16;
    LAST6:  $b+= $k[5]  <<8;
    LAST5:  $b+= $k[4];
    LAST4:  $a+= $k[3]  <<24;
    LAST3:  $a+= $k[2]  <<16;
    LAST2:  $a+= $k[1]  <<8;
    LAST1:  $a+= $k[0];
    LAST0:     return ($a+$b+$c);
}

sub pfaut {
    my ($k) = @_;
    my ($a, $b, $c) = unpack 'L3', $k.(chr(0)x11) ;
    return ($a||0) + ($b||0) + ($c||0)<<8;
}

print 'demerphq', demerphq( $_, (65) x $_) for 0 .. 11;
print 'gotoit'  ,   gotoit( $_, (65) x $_) for 0 .. 11;
print 'pfaut'  ,     pfaut(     'A' x $_) for 0 .. 11;

cmpthese( -10, {
    demerphq    => q[demerphq( $_, (65) x $_) for 0 .. 11;],
    gotoit        => q[  gotoit( $_, (65) x $_) for 0 .. 11;],
     pfaut        => q[   pfaut(      'A' x $_) for 0 .. 11;],
});
__END__
c:\test>234554.pl
demerphq0
demerphq65
demerphq16705
demerphq4276545
demerphq1094795585
demerphq1094795650
demerphq1094812290
demerphq1099072130
demerphq2189591170
demerphq2189607810
demerphq2193867650
demerphq3284386690
gotoit0
gotoit65
gotoit16705
gotoit4276545
gotoit1094795585
gotoit1094795650
gotoit1094812290
gotoit1099072130
gotoit2189591170
gotoit2189607810
gotoit2193867650
gotoit3284386690
pfaut0
pfaut16640
pfaut4276480
pfaut1094795520
pfaut1094795520
pfaut1094812160
pfaut1099072000
pfaut2189591040
pfaut2189591040
pfaut2189607680
pfaut2193867520
pfaut3284386560
Benchmark:
running
 demerphq, gotoit, pfaut
, each for at least 10 CPU seconds
...

  demerphq: 11 wallclock secs (10.56 usr +  0.00 sys = 10.56 CPU) @ 91
+5.87/s (n=9667)

    gotoit: 11 wallclock secs (10.80 usr +  0.00 sys = 10.80 CPU) @ 92
+2.72/s (n=9970)

     pfaut: 10 wallclock secs (10.14 usr +  0.00 sys = 10.14 CPU) @ 18
+09.27/s (n=18337)

           Rate demerphq   gotoit    pfaut
demerphq  916/s       --      -1%     -49%
gotoit    923/s       1%       --     -49%
pfaut    1809/s      98%      96%       --

c:\test>

</code>

---

Examine what is said, not who speaks.

The 7th Rule of perl club is -- pearl clubs are easily damaged. Use a diamond club instead.

[demerphq]

I thought up another approach last night after I had logged out. It still isn't faster than pfauts, but it is an interesting example of using code templates.
Read more... (3 kB)

Benchmark: running demerphq, gotoit, pfaut, template, 
each for at least 10 CPU seconds...
  demerphq: 11 wallclock secs (10.52 usr) @ 4809.81/s (n=50580)
    gotoit: 10 wallclock secs (10.31 usr) @ 5002.33/s (n=51584)
     pfaut: 10 wallclock secs (10.52 usr) @ 9265.21/s (n=97433)
  template: 11 wallclock secs (10.56 usr) @ 5934.77/s (n=62689)
           Rate demerphq   gotoit template    pfaut
demerphq 4810/s       --      -4%     -19%     -48%
gotoit   5002/s       4%       --     -16%     -46%
template 5935/s      23%      19%       --     -36%
pfaut    9265/s      93%      85%      56%       --

TMTOWTDI

;-)

--- demerphq
my friends call me, usually because I'm late....

[pfaut]

Mine may be fast but it makes some assumptions about the data. It assumes each value in @k is a byte value (between 0 and 255) and that $a, $b, and $c start out at 0. If the first is true but the second isn't, it might still work if the results of the unpack are stored in different variables and added to $a, $b, and $c. But I think BrowserUK said that this is part of some sort of encryption routine so it's quite possible the first assumption is wrong and thus the whole thing falls apart.

---
print map { my ($m)=1<<hex($_)&11?' ':'';
$m.=substr('AHJPacehklnorstu',hex($_),1) }
split //,'2fde0abe76c36c914586c';