About ebtree, it is partly the compiler, partly not ... the container_of macro came from warnings whereas ebtree functions was a bit my own ... my_memmem function is also my own ... all the rest is compiler warnings.
Cheers. On 6 January 2016 at 12:54, Willy TARREAU <wtarr...@haproxy.com> wrote: > Hi David, > > On Tue, Dec 29, 2015 at 01:28:13PM +0000, David CARLIER wrote: >> Hi all, >> >> these are a set of independent patches which basically remove some >> void pointers arithmetics operations in various places. At least they >> can be applied individually or not ... >> >> That should be my last submission for this year :-) > > Thank you. Just out of curiosity, did you do this just as a cleanup or > did you get some warnings from a compiler ? I'm asking because in the > case of the ebtree lookup functions, I remember that we can get a > performance impact of 10-20% on 32-bit machines just due to register > allocation and pointer aliasing, so this part was carefully tuned for > this and will need to be rechecked. > > Best regards, > Willy >