Hi Martin, On 2019-01-02 20:00, Martin Buchholz wrote:
Looks good to me.
thanks!
Can you explain why there are two identical test methods get and get2 ?
It was to provoke the expected level of mono-, bi- and megamorphicness at each callsite depending on the method under test, while ensuring the different benchmarks doesn't pollute eachother. I should add a comment about that...
I was surprised by the use of SALT in SetN. I'm guessing you could improve SetN by adapting the tricky circular array traversal code from ArrayDeque. I'm not convinced that the extra non-determinism from negative SALT pulls its weight.
This is one of Stuart's designs - the goal being to avoid situations where tests and production code create an unintended dependency on the iteration order - giving us more freedom to change hash algorithm etc. Since the SALT is calculated only once the untaken paths should be DCE'd, so we're not paying any real cost for this. (I'm guilty of moving the use of SALT from creation time to the iterators, though, to at least ensure consistent run-to-run memory layout of SetN/MapN so that we can archive these collections) /Claes
On Wed, Jan 2, 2019 at 10:12 AM Claes Redestad
