feat: introduce cpu-arch, dashmap, parking_lot

* Change mutex usage from std to parking_lot (faster impl)
* Calculate checksums/ip-addr with cpu-native ASM
* Move from locked hashmap to dashmap (lockfree)

.cargo/config.toml

@@ -1,3 +1,6 @@
 [target.aarch64-unknown-linux-gnu]
 linker = "aarch64-linux-gnu-gcc"
 runner = ["qemu-aarch64-static"] # use qemu user emulation for cargo run and test
+
+[build]
+rustflags = ["-C", "target-cpu=native"]

lightway-core/Cargo.toml

@@ -29,6 +29,7 @@ metrics.workspace = true
 more-asserts.workspace = true
 num_enum = "0.7.0"
 once_cell = "1.19.0"
+parking_lot = "0.12"
 pnet.workspace = true
 rand.workspace = true
 rand_core = "0.6.4"

lightway-core/src/connection.rs

@@ -6,13 +6,14 @@ mod io_adapter;
 mod key_update;
 
 use bytes::{Bytes, BytesMut};
+use parking_lot::Mutex;
 use rand::Rng;
 use std::borrow::Cow;
 use std::net::AddrParseError;
 use std::num::{NonZeroU16, Wrapping};
 use std::{
     net::SocketAddr,
-    sync::{Arc, Mutex},
+    sync::Arc,
     time::{Duration, Instant},
 };
 use thiserror::Error;
@@ -936,7 +937,7 @@ impl<AppState: Send> Connection<AppState> {
                 ref mut pending_session_id,
                 ..
             } => {
-                let new_session_id = rng.lock().unwrap().gen();
+                let new_session_id = rng.lock().gen();
 
                 self.session.io_cb_mut().set_session_id(new_session_id);
 

lightway-core/src/connection/builders.rs

@@ -278,7 +278,7 @@ impl<'a, AppState: Send + 'static> ServerConnectionBuilder<'a, AppState> {
         let auth = ctx.auth.clone();
         let ip_pool = ctx.ip_pool.clone();
 
-        let session_id = ctx.rng.lock().unwrap().gen();
+        let session_id = ctx.rng.lock().gen();
 
         let outside_mtu = MAX_OUTSIDE_MTU;
 

lightway-core/src/context.rs

@@ -1,8 +1,9 @@
 pub mod ip_pool;
 mod server_auth;
 
+use parking_lot::Mutex;
 use rand::SeedableRng;
-use std::sync::{Arc, Mutex};
+use std::sync::Arc;
 use thiserror::Error;
 
 use crate::{