docs: add parity benchmark instructions and table to CLAUDE.md

Author: prestwich

CLAUDE.md

@@ -145,3 +145,37 @@ This SHOULD be run alongside local tests and linting, especially for changes tha
 - Modify build configuration
 - Add new dependencies
 - Change platform-specific code
+
+## Parity Benchmarks
+
+Parity-tagged benchmarks (`// PARITY:`) have evmdb equivalents with
+identical parameters for cross-project comparison. Both repos use
+identical 32-byte binary keys and 128-byte binary values via
+`parity_key`/`parity_value` in `benches/utils.rs`.
+
+```bash
+# Run all parity benchmarks (native macOS or Linux)
+cargo bench --bench cursor_write -- "cursor_write::put::sync|cursor_write::append::sync"
+cargo bench --bench cursor -- "cursor::traverse::iter$"
+cargo bench --bench concurrent -- "readers_no_writer|readers_one_writer"
+cargo bench --bench scaling -- "sequential_get|random_get|full_iteration|append_ordered_put"
+
+# Cold-read parity benchmarks (Linux only — posix_fadvise is a no-op on macOS)
+cargo bench --bench scaling -- "cold_random_get|cold_sequential_scan"
+```
+
+### Parity bench list
+
+| Bench | File | evmdb counterpart |
+|-------|------|-------------------|
+| `cursor_write::put::sync` | cursor_write.rs | write_put_100 |
+| `cursor_write::append::sync` | cursor_write.rs | write_put_100_sorted |
+| `cursor::traverse::iter` | cursor.rs | cursor_seek_first_iterate |
+| `readers_no_writer` | concurrent.rs | readers_no_writer |
+| `readers_one_writer` | concurrent.rs | readers_with_writer |
+| `sequential_get` | scaling.rs | sequential_get |
+| `random_get` | scaling.rs | random_get |
+| `full_iteration` | scaling.rs | full_iteration |
+| `append_ordered_put` | scaling.rs | put_sorted |
+| `cold_random_get` | scaling.rs | cold_random_get |
+| `cold_sequential_scan` | scaling.rs | cold_sequential_scan |