Hi,
Przemysław Czerpak wrote:
As you can see in the last test DLMT has even better scalability what allows to reduce the speed difference and DLMT was finally a little bit faster. Anyhow this test uses only 3 threads which very nicely play with DLMT and I expect that when much bigger number of threads is used on multi CPU hardware then Hoard will be more efficient i.e. in test (3) we have 56 threads and here Hoard is still faster. But of course it does not change the fact that DLMT results are really good.
Thank You for test results. I like the simplicity of DLMT, and I think this implementation is better than trying to adapt more complex PT*, NED* or other ALLOCs. 16 separate mspaces is enough for most applications. Only on some heavy loaded servers we may need more active threads (minority of us have such number of CPU). This number 16 can be increased for a specific applications. Perhaps the only drawback of current implementation is that unused mspaces are destroyed only on application exit, but it this can be improved using some logic dependent on GC activation or thread destruction.
Regards, Mindaugas _______________________________________________ Harbour mailing list (attachment size limit: 40KB) Harbour@harbour-project.org http://lists.harbour-project.org/mailman/listinfo/harbour
