By following the instructions to build here: https://intel.github.io/ScalableVectorSearch/start_cpp.html#building
I'm able to successfully build the library. However, the default compression does not use Leanvec or LVQ. This will lead to a poor user experience for first time C++ library users.
Ideally, if a customer (such as Faiss or Opensearch) tries to build/install SVS on an Intel machine (happy path), they should be able to leverage the best performance without any customization.
- Request to make LeanVec/LVQ the default compression option, and make default performance on Intel as most optimal.
- Make the shared library described in [0] easily accessible. Today its bundled under .gz files in a release. Please consider including the latest stable .so as part of the open-source repo, so it can be used to accomplish 1.
By following the instructions to build here: https://intel.github.io/ScalableVectorSearch/start_cpp.html#building
I'm able to successfully build the library. However, the default compression does not use Leanvec or LVQ. This will lead to a poor user experience for first time C++ library users.
Ideally, if a customer (such as Faiss or Opensearch) tries to build/install SVS on an Intel machine (happy path), they should be able to leverage the best performance without any customization.