tag:blogger.com,1999:blog-2715968472735546962.post8851345405614600090..comments2023-12-14T02:21:18.222+01:00Comments on Bannalia: trivial notes on themes diverse: Measuring lookup times for C++ unordered associative containersJoaquín M López Muñozhttp://www.blogger.com/profile/08579853272674211100noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-2715968472735546962.post-43310941657662384532013-11-26T22:55:28.651+01:002013-11-26T22:55:28.651+01:00Hi Candy,
I'm not sure I understand your prop...Hi Candy,<br /><br />I'm not sure I understand your proposal. The random sequence rnd() generates unsigned ints (the type of the elements in the container), not addresses. How am I supposed to look for addresses? Maybe you could elaborate. Thank you,Joaquín M López Muñozhttps://www.blogger.com/profile/08579853272674211100noreply@blogger.comtag:blogger.com,1999:blog-2715968472735546962.post-81861630473151786142013-11-26T20:09:53.472+01:002013-11-26T20:09:53.472+01:00What if you didn't use a truly random sequence...What if you didn't use a truly random sequence but a random real-world program allocation addresses (from operator new() or malloc()) to keep? That would be a more real-world scenario where I think Boost's reasoning will come to light in having non-Power-of-2 hash functions. In any case it'll be a lot less synthetic than these tests.<br /><br />For example, take GCC compiling GCC as test data.Peter Bindelshttps://www.blogger.com/profile/02642770336702427006noreply@blogger.com