Elvis Zhang

Another issue which may be caused by newly introduced code: ``` src/terark/util/sortable_strvec.cpp: In function ‘void terark::byte_swap_in(terark::uint128_t&, mpl_::true_)’: src/terark/util/sortable_strvec.cpp:1183:6: error: ‘__builtin_bswap128’ was not declared in this scope; did you mean ‘__builtin_bswap32’?...

- The boost library has been installed in my host. But when build db_bench: ``` ln -fs librocksdb.so.7.6.0 librocksdb.so CCLD db_bench sideplugin/topling-zip/build/Linux-x86_64-g++-9.4-bmi2-0/lib_shared/libterark-core-g++-9.4-r.so：undefined reference to ‘boost::fibers::context::terminate()’ sideplugin/topling-zip/build/Linux-x86_64-g++-9.4-bmi2-0/lib_shared/libterark-zbs-g++-9.4-r.so：undefined reference to ‘boost::fibers::context::active()’ sideplugin/topling-zip/build/Linux-x86_64-g++-9.4-bmi2-0/lib_shared/libterark-core-g++-9.4-r.so：undefined...

- Build successed. And segment fault happenes when run dbbench ``` ~/toplingdb# ./db_bench -json lcompact_community.yaml -num 10000000 -disable_wal=true -value_size 2000 -benchmarks=fillrandom,readrandom -batch_size=10 WARN: MADV_POPULATE_READ requires kernel 5.14+, fallback to mlock...

- The path `/root/toplingdb_data` exists. - And I tried to use default path and re-run. The error changed. ``` ~/toplingdb# ./db_bench -json lcompact_community.yaml -num 10000000 -disable_wal=true -value_size 2000 -benchmarks=fillrandom,readrandom -batch_size=10...

Build and run successfully. Thanks~

Thanks for your reply! Are the parameters `numFlusherThreads` and `numCompactorThreads` corresponding to the max flush and compaction jobs in RocksDB? Are there any recommended default parameters to compare the performance...

Yes, you are right! The restart_index is not be updated in time for the first entry, but I found it had little effect. I'm not sure if introduce any new...