Vladimir Panteleev wrote:
On Sunday, 25 December 2011 at 01:08:04 UTC, Piotr Szturmaj wrote:
Anyway, I believe you can do without memcpy by using array copy? Array
copy might even be faster, since memcpy is not a DMD compiler intrinsic
like in many C/C++ compilers.
I converted memcpy calls to array copy but it become about 1 Mbps slower.
That should be MBps.
That's strange. I've tried optimizing some of my code today, and
changing slice copies to memcpy had the opposite effect. You are
benchmarking with -O -release -inline, right?
Yes. Here are the results: http://pastebin.com/rD8kiaQy. This is
observed only with Windows DMD.
