[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 7 Jan 2013 23:19:08 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: mask-mode speed
On Mon, Jan 07, 2013 at 09:48:27AM -0500, jfoug@....net wrote:
> Here is an extern run. This extern (listed at the end of the post), is probably about as fast as it can be done externally.
Actually, we can do a bit faster.
> void generate()
> {
> pos = 9;
> ++word[pos];
> while (word[pos] == ':') {
> word[pos] = '0';
> --pos;
> if (pos == -1) { word = 0; return; }
> ++word[pos];
> }
> }
Here's a possible replacement for the above:
void generate()
{
if (++word[9] <= '9')
return;
word[9] = '0';
pos = 8;
while (++word[pos] > '9') {
if (pos) {
word[pos--] = '0';
} else {
word = 0;
return;
}
}
}
The hard-coded 9 and 8 can be avoided with no performance impact by
putting them in variables in init().
The external mode compiler is (mostly) not optimizing - e.g., it does
not detect common subexpressions (and even if it would, quickly reusing
them in the stack-based VM would not always be possible without some
additional rearrangement to keep them on top of stack at the right
times). So e.g. "++word[pos = 9]" is quicker than "pos = 9;
++word[pos];" Similarly, directly using this expression as a
conditional in an "if" or "while" statement is quicker than repeating
"word[pos]" in there separately.
Alexander
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
