|
|
||||
|
Re: First compiled XMark results
by
Michael Kay
For Q11, I discovered that the static type analysis was making an incorrect inference that ($p/profile/@income > (5000 * $i)) was necessarily a numeric comparison. (5000+$i is always a number, but 5000*$i might be a duration.) Correcting this bug increases the time taken by the interpretive code from 4018ms to 7456ms, but decreases the compiled time from 5347ms to 5295. Don't ask me why.
|
Search
Recent Comments
Recent Articles
Month Archive
|
|||