Unsafe.{get,put}-X-Unaligned performance

Andrew Haley aph at redhat.com
Fri Mar 13 08:54:12 UTC 2015


On 12/03/15 22:02, Vitaly Davidovich wrote:
> Is vectorization coming soon? AFAIK, only memory copies are vectorized
> currently but not any arithmetic or the like.

That's true, but the idea is that this is future-proof and will work
well when we get vectorized scatter/gather memory accesses.

The workload is important when measuring performance, and which
approach works best depends on exactly what your benchmark looks like.
It's important not to optimize for random memory access patterns if
(as I suspect) random patterns are unlikely.  The current code is a
good compromise between complexity, efficiency, and generality, IMO.
I have tried a variety of scenarios and we get decent code quality for
all of them.

Andrew.



More information about the core-libs-dev mailing list