Just tripled the nr of iterations to get a little better timing accuracy. Ptr.Single(...) : 15 seconds C 7 seconds
So it looks as if C is still around 2X faster. I can live with that difference ( although if RS can improve on that all the better) . But the debug version is unacceptably slow imo. I'm now going to go away and do as frank suggests and put both loops inside the C function. On 16/5/07 20:58, "Daniel Stenning" <[EMAIL PROTECTED]> wrote: > OK - very interesting. > > Just done the test in the RELEASE executable: > > Memoryblock.SingleValue(...) : 30 seconds > Ptr.Single(...) : 5 seconds !!!!! > > C dylib call version: around 2 seconds. ( almost too quick to time > accurately ) > > This is a great relief. Interestingly it seems to confirm two things: > > A) That there is a huge ( and I mean HUGE ) difference in loops when running > in debug mode > > > B) There is a huge performance hit in using the memoryblock methods to > access memory block variables versus using raw Ptr dereferencing. This > seems to me to confirm that the RB compiler is generating code to access the > single data directly rather than via a special function. > > > On 16/5/07 20:33, "Daniel Stenning" <[EMAIL PROTECTED]> wrote: > >> >> Good point. Indeed it was. I'll do the test again in a release build and >> report back with the results. > > Regards, > > Dan > > > > _______________________________________________ > Unsubscribe or switch delivery mode: > <http://www.realsoftware.com/support/listmanager/> > > Search the archives: > <http://support.realsoftware.com/listarchives/lists.html> > Regards, Dan _______________________________________________ Unsubscribe or switch delivery mode: <http://www.realsoftware.com/support/listmanager/> Search the archives: <http://support.realsoftware.com/listarchives/lists.html>
