jtamate updated this revision to Diff 33505. jtamate marked 3 inline comments as done. jtamate edited the summary of this revision. jtamate edited the test plan for this revision. jtamate added a comment.
Fixed the ordered insertion. Using std::vector and std::find_if. Initialize everything to try to detect a change of type in a "insert" over a different type. Using templates to reduce somehow the code size. The last 3 structures are tested 3 times: - Fill the structure - Compare two structures - Read 3 values > When you say "scales better", we're talking about the number of fields in the udsentry, not the number of items. But kioslaves don't fill in 1000 fields, so I have the feeling that scaling with the number of fields isn't a requirement. Yes, I was talking about the number of fields in the udsentry. I had to test it, just in case. > Are those benchmarks run in Release (or RelWithDebInfo) mode, rather than Debug (which is a big no no for benchmarks)? Qt should be compiled with optimizations enabled too. Yes, since the last comment of D11487 <https://phabricator.kde.org/D11487> everything is compiled with -O2 -mtune=native Qt is the one provided by OpenSuse. REPOSITORY R241 KIO CHANGES SINCE LAST UPDATE https://phabricator.kde.org/D12659?vs=33480&id=33505 REVISION DETAIL https://phabricator.kde.org/D12659 AFFECTED FILES autotests/udsentry_benchmark.cpp To: jtamate, dfaure, #frameworks Cc: bruns, michaelh
