On Sunday 15 April 2007, Johannes Schindelin wrote:
quoted text
> On Sun, 15 Apr 2007, Shawn O. Pearce wrote:
> >  static int simple_length(const char *match)
> >  {
> > -	const char special[256] = {
> > -		[0] = 1, ['?'] = 1,
> > -		['\'] = 1, ['*'] = 1,
> > -		['['] = 1
> > -	};
> >  	int len = -1;
> >  
> >  	for (;;) {
> >  		unsigned char c = *match++;
> >  		len++;
> > -		if (special[c])
> > +		switch (c) {
> > +		case 0: case '?':
> > +		case '\': case '*':
> > +		case '[':
> >  			return len;
> > +		}
> >  	}
> >  }
> 
> You are replacing a table-based check with a switch based, which might be 
> substantially slower (depends on how often cmp_name() is called).

Or faster. When the table gives a cache miss and has to be
loaded from main memory, I am quite sure that 5 compares in a row are
faster than the cache miss.

Actually, with the switch, the compiler is free to implement it with a
table (and gcc usually does this, probably even using a substantially
smaller table). The table-based check in contrast looks
like some kind of micro-optimization which makes the code IMHO more
difficult to read, and which only would be justified with meassured
improvements.

Josef
-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html