Avoid unnecessary memory allocations, copies, and inefficient memory usage patterns that can impact performance. This includes several key practices: **Avoid unnecessary allocations:** - Use static allocation instead of heap allocation when possible: `static NullIoStream globalNullStream;` instead of `static auto globalNullStream = kj::heap();` - Avoid intermediate string copies: prefer direct construction over `kj::str(...)` when the result is immediately wrapped **Optimize parameter passing:** - Use const references for non-trivial types to prevent copies: `const tracing::TraceId& traceId` instead of `tracing::TraceId traceId` - Use move semantics appropriately, but be aware that explicit `kj::mv()` can interfere with NRVO (Named Return Value Optimization) **Pre-allocate when size is known:** ```cpp // Instead of letting vector grow dynamically kj::Vector> promises; // Reserve space when size is known promises.reserve(filenames.size()); ``` **Choose efficient memory usage patterns:** - Pump data to null streams instead of buffering in memory when data will be discarded: avoid `readAllBytes()` that buffers unnecessarily - Use more efficient construction methods: `kj::Path({"tmp"})` instead of `kj::Path::parse("tmp")` These optimizations are particularly important in hot paths or when processing large amounts of data, as small inefficiencies can compound significantly.