Or are you upset that https://github.com/SwuduSusuwu/SubStack/blob/trunk/.clang-tidy excludes some of the altera-* tests?
altera-id-dependent-backward-branch gives lots of false positives, some of which the new for(auto &X : S) loops address.
altera-struct-pack-align has lots of false positives about class ResultList (that the std::vectors are not aligned),
altera-unroll-loops has false positives for almost all loops.
First of search results: https://x.com/Lenkafomenka (posts about hardware accelerators) Do you want Neon/AVX2 paths? https://github.com/SwuduSusuwu/SubStack/blob/trunk/cxx/ClassCns.hxx implementations can benefit from this, but would not have as much general-purposeness. https://github.com/SwuduSusuwu/SubStack/blob/trunk/cxx/ClassResultList.hxx is all templates, thus
./build.sh --release(which passes CXXFLAGS-O2) should optimize for you. Forgot to pass-march=native, which perhaps can optimize for GPUs.Or are you upset that https://github.com/SwuduSusuwu/SubStack/blob/trunk/.clang-tidy excludes some of the
altera-*tests?altera-id-dependent-backward-branchgives lots of false positives, some of which the newfor(auto &X : S)loops address.altera-struct-pack-alignhas lots of false positives aboutclass ResultList(that thestd::vectorsare not aligned),altera-unroll-loopshas false positives for almost all loops.