On Mon, Jul 24, 2017 at 2:54 PM, Gunnar Morling <gun...@hibernate.org> wrote:
> Awesome, thanks a lot for getting the numbers and applying the > improvements. > > So to be sure, at runtime (non-metadata retrieval), CR3 already is faster > than 5.4, also without your latest optimisations from the pending PR? You > mentioned this morning that having the first 6 commits from your PR would > be beneficial for the Final, but I think that's about reducing memory > usage, correct? > Nope. CR3 is faster than 5.4 when not dealing with cascaded validation. CR3 is significantly slower for cascading than 5.4 due to one call to TypeVariableBindings executed at runtime. When I say significantly, it's 2 time slower in the CascadedValidation test of our performance tests, which is the worst case (e.g. the validation is really fast as you only have a very simple constraint to apply). The first 6 commits of the PR (the ones referencing HV-1438 <https://github.com/hibernate/hibernate-validator/pull/814/commits/b60220c91d119003d5ca38d7e208bf42274a2de5>) are about this issue (e.g. have a list of compatible value extractors and avoid playing with the type arguments at runtime). -- Guillaume _______________________________________________ hibernate-dev mailing list hibernate-dev@lists.jboss.org https://lists.jboss.org/mailman/listinfo/hibernate-dev
