view release on metacpan or  search on metacpan

TODO  view on Meta::CPAN

84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
  problem:
    we need to pass a filehandle back to inherit the lock.
    or we could save it globally somewheres?
 
pseudo code:
 
get_if_same:
 
if(!defined(is_same(...))) {
  if(my $FH = open_with_flock(LOCK_EX, block => 0)) {
    $cache::static::_curr_locked_FH = $FH;
  } else {
    #wait until it's been set, then return it
    open_with_flock(LOCK_SH, block => 1);
    return get(...);
  }
} else {
   return get(...);
}
 
set:
my $FH = $cache::static::_curr_locked_FH || open(...);
 
...
close($FH);
 
an optimization idea stolen from the lighthttpd folks:
  (http://www.lighttpd.net/documentation/performance.html)
                stat() cache
 
                A stat(2) can be expensive; caching it saves time and context switches.

TODO  view on Meta::CPAN

125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
        granted, they are talking about the context of httpd serving static
        pages, but it could still cut down on our number of stat calls by a
        large factor in the common case...
 
next up:
in the case where you allow old results:
  before writing, copy the cache file "$f" to "$f.bak"
  any get_if_same() during that time reads "$f.bak"
 
  when done, remove "$f.bak" (this results in race condition in reader:
    1 - find the file locked, try the .bak file
    2 - .bak file is removed, goto 1
  OR have a crontab that runs e.g. hourly to clean up .bak files...
 
 
  start the lock when generation starts, remove the lock when new value
  has been set
situation:
  refresh of key K takes 30 seconds.
  every 5 seconds, we get a hit.
  this means we regenerate 6 times, when we only needed to do it once.

doc/digests  view on Meta::CPAN

6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
available/installed.
 
MD5:     1.405/1.397/1.396
newhash: 0.923/0.920/0.925
 
#MD5 is much faster than SHA1 (at least in perl)
#MD4 is marginally faster than MD5 (but less available)
 
#Pearson, JHash, Elf, DJB, FNV all show promise but don't outperform newhash 
 
#testing done on Athlon-XP downclocked to 1000MHz
 
#general notes:
#OO is faster
#using addfile instead of add is MUCH faster!
 
time perl -e 'use Digest::MD5 qw(md5 md5_hex md5_base64); for my $i
(0..100) { open(FH, "foobar"); my $digest = md5(join ("", <FH>)); 
close(FH); } '
MD5:      4.806/4.731/4.752
MD4:      4.625/4.647/4.620