Re: roff.c question

[Ingo Schwarze <schwarze@usta.de>]

Hi Joerg,

Joerg Sonnenberger wrote on Fri, Dec 10, 2010 at 09:52:03PM +0100:
> On Fri, Dec 10, 2010 at 09:45:13PM +0100, Ingo Schwarze wrote:

>> If we want to go into micro-optimization, we could do this:
>> 
>>         for (cp = *bufp + pos; *cp; cp++) {
>>                 if ('\\' != *cp)
>>                         continue;

> ... or use a single strchr() call to start the search.

Sure, that would be correct, and not even obfuscation.

> For the typical case of medium long lines and no \, strchr
> can operate mostly on dwords...

Here is our implementation of strchr():

char *
strchr(const char *p, int ch)
{
	for (;; ++p) {
		if (*p == ch)
			return((char *)p);
		if (!*p)
			return((char *)NULL);
	}
	/* NOTREACHED */
}

You mean, the compiler will optimize that into dword
comparisons, but will not optimize this code:

	for (cp = *bufp + pos; *cp; cp++) {
		if ('\\' != *cp)
			continue;

And the difference will buy us enough to warrant the extra
function call?

I doubt any of this is relevant, but i'm tempted to do some
measurements...

Yours,
  Ingo
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv