On Thu, Jul 20, 2006 at 11:32:00AM +1000, Carlo Sogono wrote: > With regards to Benno's code: > After trying raw mallocs and performing quite well, I think it might > have to do with the code. I have different versions of my software,
ooh Have you checked your hash function/performance? How many land in the sam bucket? For really bad hash performance a basic lookup could be almost linear. > hashed linked lists, hashed binary trees, etc. With 87 million mallocs I > expect binary trees of a hash table with size 100,001 to have a 800 Forgive my ignorance, what's a hashed linked list or binary tree as compared to a normal linked list or binary tree? > entries each. Testing the same code with 1 million mallocs with a hash > table of 1,001 still gives me good results (that's an average of about > 1000 per tree) at less than 5 minutes. I tried to insert 87 million to a > hash of 100,001 yesterday at 4pm and it wasn't finished at 9am today. I > think that's just way too slow. By the way i just use a simple for loop > and % hash size so all trees should have the same depth. > > 87 million / 100,001 = 869 (4pm-9am still not complete) > 1 million / 1,001 = 999 (by 5 mins) > > I would attach my source but it contains proprietary code :/ > > Thanks in advance. > > Carlo _______________________________________________ coders mailing list [email protected] http://lists.slug.org.au/listinfo/coders
