Six sections with concrete numbers: end-to-end latency vs kernel (40-52%
faster), CPU microbenchmarks (26.87 Gbps, 481M hash/s), 5-round
optimization effects (P0 19x improvement), architecture advantages
(zero-copy, zero-syscall), rdtsc latency traces (all <1us cross-core),
and code scale (6000 lines pure C).
Ready for forum/ blog post publishing.
Co-Authored-By: Claude <noreply@anthropic.com>
Wait up to 1s for PHY auto-negotiation after rte_eth_dev_start.
Reports link speed or 'link DOWN (no carrier)' to help diagnose
physical connectivity issues.
Co-Authored-By: Claude <noreply@anthropic.com>
- ws_client.c: When payload exceeds buffer, consume remaining bytes
before returning error to prevent parser desynchronization.
- market_feed.c: Increase receive buffer 4K→64K, support both
standard (symbol/timestamp) and compact (s/ts) JSON field names.
- ws_client.c: Disable per-frame debug print in release mode.
Co-Authored-By: Claude <noreply@anthropic.com>
Document all P0-P4 optimizations with code snippets, branch structure
(main vs feature/multi-core), performance comparison table, and future
optimization roadmap.
Co-Authored-By: Claude <noreply@anthropic.com>
wxj