Clangd Indexing Memory Limits for Large C++ Projects

Answer: clangd memory‑usage controls

The clangd language server does not expose a direct option to cap its RAM consumption. The only built‑in lever is the number of background‑indexing threads.

Confirmed facts

• Clangd provides the flag --background-index-threads=N to limit how many threads perform indexing.
• Setting --background-index=false disables background indexing entirely, which stops the symbol database from growing but also disables features such as “Find Usages” and live completion.
• The symbol index is built from the compilation database (compile_commands.json) that clangd receives; there is no supported way to paginate the index or to restrict it to a subset of directories without changing that file.

Likely explanation / practical steps

1. Check the clangd version with clangd --version. (If you are using a release prior to 18.x, the flag names may differ.)
2. Start clangd with a reduced thread count, for example:

   clangd --background-index-threads=1

3. Observe the process memory with top or htop while the IDE indexes a large project.
4. If memory is still too high, consider disabling background indexing:

   clangd --background-index=false

5. To keep indexing limited to specific parts of the codebase, prune compile_commands.json so it only lists the source directories you need indexed, or run a separate clangd instance per subproject.

What to verify

• Run clangd --version and confirm the flag names match the output of clangd --help.
• After applying the flag, monitor RAM usage; a noticeable drop indicates the setting is effective.
• Optionally, temporarily remove a directory from compile_commands.json and verify that indexing time and memory usage decrease.

Please let me know the exact clangd version you are using, as older releases may use different flag names or lack the background‑index option.