On Wed, Oct 04, 2006 at 03:48:20PM +0200, Abdelrazak Younes wrote: > Note that I didn't say that it is slow, Andre did.
It accounts for 7% of the total run time, a bit more than rowBreakPoint() which used to be a well-known bottleneck. > I was just investigating why it could be slow. Some of the solutions > offered in this thread (the resrve() amd/or the inlining) may very > well bring a speed advantage on Linux; I am just saying that there's > no improvement in Windows. Yes, 'no cache' brings a 30% speedup on Linux. Andre'
