sub sms_WxH_Perl_4loops( $m, $w, $h ) {
    my ( $W, $H ) = ( scalar $m-> [0]-> @*, scalar @$m );
    my @ret;
    for my $i ( 0 .. $H - $h ) {
        for my $j ( 0 .. $W - $w ) {
            for my $k ( 0 .. $w - 1 ) {
                for my $l ( 0 .. $h - 1 ) {
                    $ret[ $i ][ $j ] += $m-> [ $i + $l ][ $j + $k ];
                }
            }
        }
    }
    return \@ret
}

sub sms_WxH_Perl_3loops( $m, $w, $h ) {
    my ( $W, $H ) = ( scalar $m-> [0]-> @*, scalar @$m );
    my ( @tmp, @ret );
    for my $i ( 0 .. $H - 1 ) {
        for my $j ( 0 .. $W - $w ) {
            for my $k ( 0 .. $w - 1 ) {
                $tmp[ $i ][ $j ] += $m-> [ $i ][ $j + $k ];
            }
            next if $i < $h - 1;
            for my $k ( 0 .. $h - 1 ) {
                $ret[ $i - $h + 1 ][ $j ] += $tmp[ $i - $k ][ $j ];
            }
        }
    }
    return \@ret
}

$m = [ 
    map [ 
        map rand, 0 .. 149
    ], 0 .. 149
];

__END__

+
             Time (s) vs. N (NxN submatrix, 150x150 matrix)
       +----------------------------------------------------------+
       |+       +      +       +      +       +       +      +    |
    10 |-+                                                      +-|
       |+                                                        +|
       |+                                                        +|
       |+                      A   A  A   A                      +|
       |+                  A                  A                  +|
       |+              A                          A              +|
       |           A                                  A           |
       |                                                          |
     1 |-+                                               A      +-|
       |+       A                                                +|
       |+                                                        +|
       |+                                                        +|
       |+                                                    A   +|
       |    A                                                     |
       |+                                                        +|
       |               B   B   B   B  B   B                       |
   0.1 |-+      B  B                          B                 +-|
       |+                                         B              +|
       |+   B                                         B          +|
       |+                                                B       +|
       |+                                                        +|
       |                                                          |
       |+                                                    B   +|
       |                                                          |
  0.01 |-+                                                      +-|
       |+       +      +       +      +       +       +      +   +|
       +----------------------------------------------------------+
        0       20     40      60     80     100     120    140
                       sms_WxH_Perl_4loops    A
                       sms_WxH_Perl_3loops    B
  +-----+-------+-------+
  | N   | A     | B     |
  +-----+-------+-------+
  | 10  | 0.250 | 0.053 |
  | 20  | 0.809 | 0.091 |
  | 30  | 1.528 | 0.116 |
  | 40  | 2.288 | 0.134 |
  | 50  | 2.988 | 0.144 |
  | 60  | 3.512 | 0.147 |
  | 70  | 3.828 | 0.144 |
  | 80  | 3.862 | 0.134 |
  | 90  | 3.628 | 0.122 |
  | 100 | 3.122 | 0.103 |
  | 110 | 2.453 | 0.088 |
  | 120 | 1.675 | 0.066 |
  | 130 | 0.900 | 0.044 |
  | 140 | 0.284 | 0.022 |
  +-----+-------+-------+