Description:
------------
Parsing file with 10000 lines of following format:

level + delim + [@xref_id@ + delim +] tag + [delim + line_value +] terminator

level		digit
delim		space
xref_id	alphanum
tag		alpha (english)
line_value	any (except terminator)
terminator	\r\n

With regexp:

$c=preg_match_all("/^\s*(\d+)\s+(@(\S+)@\s+)?(\w+)(\s+@(\S+)@\s*|.*)?$/Sm",$fp,$m,PREG_PATTERN_ORDER);

Setting PCRE_UTF8 modifier slows whole script down 30 times (from 300ms to 9000ms).

May be more accurate regexp here will be
$c=preg_match_all("/^ *(\d+) +(@([^@\\n]+)@ +)?([^ \\n]+)( +@([^@\\n]+)@ *| +[^\\n]*)?$/m",$fp,$m,PREG_PATTERN_ORDER);
But it changes nothing.