06:26 <lsneff> whitequark: I've gotten the queryable db generated out now for vcd files, but I've written myself into a corner, where the vcd is converted into an intermediary format, and then the intermediary format has to be traversed with basically random access, backwards, to create the db. The traversing itself, simply reading from the mmapped intermediary format, takes almost half the total time.

14:08 <Sarayan> What is the currently correct way in C++ to lookup and drive pins and clocks in a cxxrtl-generated class?

16:40 <lkcl> lsneff: an off-the-wall suggestion / intuitive-side-step-if-you-know-what-i-mean: have you considered lumosql?

16:40 <lkcl> https://github.com/LumoSQL/lumosql

16:41 <lkcl> you can almost certainly get rust bindings to sqlite...

16:42 <lkcl> lumosql would give you "standard" sqlite query capability, with screamingly-fast backend read performance

16:45 <lkcl> mmm maybe not https://lumosql.org/benchmarking/

16:46 <lkcl> looks like there's still room for improvement there.

16:48 <lkcl> oh wait... err... 3.8.0-sqlite-lmdb completion time is a HUNDRED times quicker than 3.7.17 holy cow

22:29 <lsneff> lkcl: I'll look into that, but it needs to be query able for data I don't think sql is particularly good at.

22:29 <lsneff> Anyhow, I can load traces as they're looked at, making it much faster