Performance optimizations

README.md

@@ -47,29 +47,35 @@
 
 ## Performance
 
+- **Test Overview**
+  One large packet with 100 Million bytes of data
+
 - **Hardware & Conditions**  
   Tested on a MacBook Air M3 (my daily driver machine, running multiple background applications)
 
 - **Interface**  
   Real Wi-Fi interface: en0
 
 - **Throughput**  
-  9.4 MB/s
+  20.9 MB/s
 
 - **Memory Usage**  
-  Peak memory footprint: 14 million bytes
+  **Peak memory footprint: 203 million bytes**
+  - 100M bytes are used for buffers on receiver and sender
+  - 3M bytes are used on internal data
+
 
 - **CPU Usage**  
-  Average utilization: ~7.7%
+  Average utilization: ~9%
 
 - **Detailed Timing (from /usr/bin/time -l)**  
-  - Real (wall-clock) time: 5.32 seconds  
-  - User CPU time: 0.24 seconds  
-  - System CPU time: 0.17 seconds  
-  - Total CPU time: 0.41 seconds
+  - Real (wall-clock) time: 4.78 seconds  
+  - User CPU time: 0.07 seconds  
+  - System CPU time: 0.31 seconds  
+  - Total CPU time: 0.38 seconds
 
 - **CPU Cycles**  
-  1 005 063 001 cycles
+  1 513 423 854 cycles
 
 ## Installation
 Follow these steps to install SwiftNet:

src/execute_packet_callback.c

@@ -59,52 +59,51 @@ void execute_packet_callback(
     pthread_mutex_t* const execute_callback_mtx,
     pthread_cond_t* const execute_callback_cond
 ) {
-    while (1) {
-        if (atomic_load_explicit(closing, memory_order_acquire) == true) {
-            break;
-        }
+check_packet:
+    if (unlikely(atomic_load_explicit(closing, memory_order_acquire) == true)) {
+        return;
+    }
 
-        pthread_mutex_lock(execute_callback_mtx);
+    pthread_mutex_lock(execute_callback_mtx);
 
-        const struct PacketCallbackQueueNode* const node = wait_for_next_packet_callback(queue);
-        if(node == NULL) {
-            pthread_cond_wait(execute_callback_cond, execute_callback_mtx);
+    const struct PacketCallbackQueueNode* const node = wait_for_next_packet_callback(queue);
+    if(node == NULL) {
+        pthread_cond_wait(execute_callback_cond, execute_callback_mtx);
 
-            pthread_mutex_unlock(execute_callback_mtx);
+        pthread_mutex_unlock(execute_callback_mtx);
 
-            continue;
-        }
+        goto check_packet;
+    }
 
-        pthread_mutex_unlock(execute_callback_mtx);
+    pthread_mutex_unlock(execute_callback_mtx);
 
-        atomic_thread_fence(memory_order_acquire);
+    if(node->packet_data == NULL) {
+        allocator_free(&packet_callback_queue_node_memory_allocator, (void*)node);
+        goto check_packet;
+    }
 
-        if(node->packet_data == NULL) {
-            allocator_free(&packet_callback_queue_node_memory_allocator, (void*)node);
-            continue;
-        }
+    if(node->pending_message != NULL) {
+        remove_pending_message_from_hashmap(pending_messages, node->pending_message);
+    }
 
-        if(node->pending_message != NULL) {
-            remove_pending_message_from_hashmap(pending_messages, node->pending_message);
+    void (*const packet_handler_loaded)(void* const, void* const) = atomic_load(packet_handler);
+    if (unlikely(packet_handler_loaded == NULL)) {
+        if (connection_type == CONNECTION_TYPE_CLIENT) {
+            swiftnet_client_destroy_packet_data(node->packet_data, connection);
+        } else {
+            swiftnet_client_destroy_packet_data(node->packet_data, connection);
         }
 
-        void (*const packet_handler_loaded)(void* const, void* const) = atomic_load(packet_handler);
-        if (unlikely(packet_handler_loaded == NULL)) {
-            if (connection_type == CONNECTION_TYPE_CLIENT) {
-                swiftnet_client_destroy_packet_data(node->packet_data, connection);
-            } else {
-                swiftnet_client_destroy_packet_data(node->packet_data, connection);
-            }
+        allocator_free(&packet_callback_queue_node_memory_allocator, (void*)node);
 
-            allocator_free(&packet_callback_queue_node_memory_allocator, (void*)node);
+        goto check_packet;
+    }
 
-            continue;
-        }
+    (*packet_handler_loaded)(node->packet_data, atomic_load_explicit(user_data, memory_order_acquire));
 
-        (*packet_handler_loaded)(node->packet_data, atomic_load_explicit(user_data, memory_order_acquire));
+    allocator_free(&packet_callback_queue_node_memory_allocator, (void*)node);
 
-        allocator_free(&packet_callback_queue_node_memory_allocator, (void*)node);
-    }
+    goto check_packet;
 }
 
 void* execute_packet_callback_client(void* const void_client) {

src/handle_packets.c

@@ -66,8 +66,8 @@ static inline void swiftnet_handle_packets(
 	const struct pcap_pkthdr* hdr,
 	const uint8_t* packet,
     pthread_mutex_t* const process_packets_mtx,
-    pthread_cond_t* const process_packets_cond
-
+    pthread_cond_t* const process_packets_cond,
+    _Atomic bool *const processing_packets
 ) {
     uint8_t* const packet_buffer = allocator_allocate(&packet_buffer_memory_allocator);
     if (unlikely(packet_buffer == NULL)) {
@@ -100,15 +100,15 @@ static inline void swiftnet_handle_packets(
 
     struct PacketQueueNode* const node = construct_node(len, packet_buffer, sender_address);
 
-    atomic_thread_fence(memory_order_release);
-
-    pthread_mutex_lock(process_packets_mtx);
-
     insert_queue_node(node, packet_queue, connection_type);
 
-    pthread_cond_signal(process_packets_cond);
+    if (atomic_load_explicit(processing_packets, memory_order_acquire) != true) {
+        pthread_mutex_lock(process_packets_mtx);
 
-    pthread_mutex_unlock(process_packets_mtx);
+        pthread_cond_signal(process_packets_cond);
+
+        pthread_mutex_unlock(process_packets_mtx);
+    }
 }
 
 static void handle_client_init(struct SwiftNetClientConnection* user, const struct pcap_pkthdr* hdr, const uint8_t* buffer) {
@@ -177,7 +177,7 @@ static inline void handle_correct_receiver(const enum ConnectionType connection_
         if (client_connection->initialized == false) {
             handle_client_init(client_connection, hdr, packet);
         } else {
-            swiftnet_handle_packets(client_connection->port_info.source_port, &client_connection->process_packets_thread, client_connection, CONNECTION_TYPE_CLIENT, &client_connection->packet_queue, &client_connection->closing, client_connection->loopback, client_connection->addr_type, hdr, packet, &client_connection->process_packets_mtx, &client_connection->process_packets_cond);
+            swiftnet_handle_packets(client_connection->port_info.source_port, &client_connection->process_packets_thread, client_connection, CONNECTION_TYPE_CLIENT, &client_connection->packet_queue, &client_connection->closing, client_connection->loopback, client_connection->addr_type, hdr, packet, &client_connection->process_packets_mtx, &client_connection->process_packets_cond, &client_connection->processing_packets);
         }
     } else {
         LOCK_ATOMIC_DATA_TYPE(&listener->servers.atomic_lock);
@@ -189,7 +189,7 @@ static inline void handle_correct_receiver(const enum ConnectionType connection_
             return;
         }
 
-        swiftnet_handle_packets(server->server_port, &server->process_packets_thread, server, CONNECTION_TYPE_SERVER, &server->packet_queue, &server->closing, server->loopback, server->addr_type, hdr, packet, &server->process_packets_mtx, &server->process_packets_cond);
+        swiftnet_handle_packets(server->server_port, &server->process_packets_thread, server, CONNECTION_TYPE_SERVER, &server->packet_queue, &server->closing, server->loopback, server->addr_type, hdr, packet, &server->process_packets_mtx, &server->process_packets_cond, &server->processing_packets);
     }
 }
 

src/initialize_client_connection.c

@@ -56,7 +56,7 @@ void* request_server_information(void* const request_server_information_args_voi
             }
         #endif
 
-        swiftnet_pcap_send(request_server_information_args->pcap, request_server_information_args->data, request_server_information_args->size);
+        SWIFTNET_PCAP_SEND_SAFE(request_server_information_args->pcap, request_server_information_args->data, request_server_information_args->size);
 
         usleep(250000);
     }
@@ -98,6 +98,7 @@ static inline struct SwiftNetClientConnection* const construct_client_connection
     UNLOCK_ATOMIC_DATA_TYPE(&new_connection->packet_queue.locked);
     UNLOCK_ATOMIC_DATA_TYPE(&new_connection->packet_callback_queue.locked);
 
+    atomic_store_explicit(&new_connection->processing_packets, true, memory_order_release);
     atomic_store_explicit(&new_connection->closing, false, memory_order_release);
     atomic_store_explicit(&new_connection->initialized, false, memory_order_release);
     atomic_store_explicit(&new_connection->packet_handler_user_arg, NULL, memory_order_release);

src/initialize_server.c

@@ -39,6 +39,7 @@ static inline struct SwiftNetServer* const construct_server(const bool loopback,
     UNLOCK_ATOMIC_DATA_TYPE(&new_server->packet_queue.locked);
     UNLOCK_ATOMIC_DATA_TYPE(&new_server->packet_callback_queue.locked);
 
+    atomic_store_explicit(&new_server->processing_packets, true, memory_order_release);
     atomic_store_explicit(&new_server->packet_handler, NULL, memory_order_release);
     atomic_store_explicit(&new_server->packet_handler_user_arg, NULL, memory_order_release);
     atomic_store_explicit(&new_server->closing, false, memory_order_release);

src/internal/datatype_allocator.c

@@ -62,8 +62,6 @@ static inline void set_memory_status(struct SwiftNetMemoryAllocator* const memor
             return;
         }
     }
-
-    printf("doesnt fall\n");
 }
 
 static inline bool is_already_free(struct SwiftNetMemoryAllocator* const memory_allocator, void* const memory_location) {
@@ -129,7 +127,7 @@ struct SwiftNetMemoryAllocatorStack* const find_free_pointer_stack(struct SwiftN
     if((bitmap | invalid_bitmap) == UINT64_MAX) {
         return NULL;
     } else {
-        first_free = __builtin_ctz(~(bitmap | invalid_bitmap));
+        first_free = __builtin_ctzll(~(bitmap | invalid_bitmap));
     }
 
     if (first_free >= stacks_allocated) {

src/internal/internal.h

@@ -3,6 +3,7 @@
 #include <arpa/inet.h>
 #include <pcap/pcap.h>
 #include <stdint.h>
+#include <arm_acle.h>
 #include <string.h>
 #include <netinet/in.h>
 #include <stdatomic.h>
@@ -27,6 +28,11 @@
     #define DISABLE_INTERNAL_CHECK
 #endif
 
+#define SWIFTNET_PCAP_SEND_SAFE(pcap, buffer, len) \
+    while(swiftnet_pcap_send(pcap, buffer, len) == -2) { \
+        usleep(2000); \
+    } \
+
 enum RequestLostPacketsReturnType {
     REQUEST_LOST_PACKETS_RETURN_UPDATED_BIT_ARRAY = 0x00,
     REQUEST_LOST_PACKETS_RETURN_COMPLETED_PACKET  = 0x01
@@ -82,8 +88,8 @@ enum RequestLostPacketsReturnType {
 
 // Simple crc16 call with proper memory order
 #define HANDLE_CHECKSUM(buffer, size, prepend_size) \
-    uint16_t checksum = htons(crc16(buffer, size)); \
-    memcpy(buffer + prepend_size + offsetof(struct ip, ip_sum), &checksum, sizeof(checksum));
+    const uint32_t checksum = crc32(buffer, size); \
+    memcpy(buffer + prepend_size + sizeof(struct ip) + offsetof(struct SwiftNetPacketInfo, checksum), &checksum, sizeof(checksum));
 
 // Number used in ip.proto
 #define PROT_NUMBER 253
@@ -99,50 +105,22 @@ enum RequestLostPacketsReturnType {
 
 #define MIN(one, two) (one > two ? two : one)
 
-static const uint16_t crc16_table[256] = {
-    0x0000, 0xC0C1, 0xC181, 0x0140, 0xC301, 0x03C0, 0x0280, 0xC241,
-    0xC601, 0x06C0, 0x0780, 0xC741, 0x0500, 0xC5C1, 0xC481, 0x0440,
-    0xCC01, 0x0CC0, 0x0D80, 0xCD41, 0x0F00, 0xCFC1, 0xCE81, 0x0E40,
-    0x0A00, 0xCAC1, 0xCB81, 0x0B40, 0xC901, 0x09C0, 0x0880, 0xC841,
-    0xD801, 0x18C0, 0x1980, 0xD941, 0x1B00, 0xDBC1, 0xDA81, 0x1A40,
-    0x1E00, 0xDEC1, 0xDF81, 0x1F40, 0xDD01, 0x1DC0, 0x1C80, 0xDC41,
-    0x1400, 0xD4C1, 0xD581, 0x1540, 0xD701, 0x17C0, 0x1680, 0xD641,
-    0xD201, 0x12C0, 0x1380, 0xD341, 0x1100, 0xD1C1, 0xD081, 0x1040,
-    0xF001, 0x30C0, 0x3180, 0xF141, 0x3300, 0xF3C1, 0xF281, 0x3240,
-    0x3600, 0xF6C1, 0xF781, 0x3740, 0xF501, 0x35C0, 0x3480, 0xF441,
-    0x3C00, 0xFCC1, 0xFD81, 0x3D40, 0xFF01, 0x3FC0, 0x3E80, 0xFE41,
-    0xFA01, 0x3AC0, 0x3B80, 0xFB41, 0x3900, 0xF9C1, 0xF881, 0x3840,
-    0x2800, 0xE8C1, 0xE981, 0x2940, 0xEB01, 0x2BC0, 0x2A80, 0xEA41,
-    0xEE01, 0x2EC0, 0x2F80, 0xEF41, 0x2D00, 0xEDC1, 0xEC81, 0x2C40,
-    0xE401, 0x24C0, 0x2580, 0xE541, 0x2700, 0xE7C1, 0xE681, 0x2640,
-    0x2200, 0xE2C1, 0xE381, 0x2340, 0xE101, 0x21C0, 0x2080, 0xE041,
-    0xA001, 0x60C0, 0x6180, 0xA141, 0x6300, 0xA3C1, 0xA281, 0x6240,
-    0x6600, 0xA6C1, 0xA781, 0x6740, 0xA501, 0x65C0, 0x6480, 0xA441,
-    0x6C00, 0xACC1, 0xAD81, 0x6D40, 0xAF01, 0x6FC0, 0x6E80, 0xAE41,
-    0xAA01, 0x6AC0, 0x6B80, 0xAB41, 0x6900, 0xA9C1, 0xA881, 0x6840,
-    0x7800, 0xB8C1, 0xB981, 0x7940, 0xBB01, 0x7BC0, 0x7A80, 0xBA41,
-    0xBE01, 0x7EC0, 0x7F80, 0xBF41, 0x7D00, 0xBDC1, 0xBC81, 0x7C40,
-    0xB401, 0x74C0, 0x7580, 0xB541, 0x7700, 0xB7C1, 0xB681, 0x7640,
-    0x7200, 0xB2C1, 0xB381, 0x7340, 0xB101, 0x71C0, 0x7080, 0xB041,
-    0x5000, 0x90C1, 0x9181, 0x5140, 0x9301, 0x53C0, 0x5280, 0x9241,
-    0x9601, 0x56C0, 0x5780, 0x9741, 0x5500, 0x95C1, 0x9481, 0x5440,
-    0x9C01, 0x5CC0, 0x5D80, 0x9D41, 0x5F00, 0x9FC1, 0x9E81, 0x5E40,
-    0x5A00, 0x9AC1, 0x9B81, 0x5B40, 0x9901, 0x59C0, 0x5880, 0x9841,
-    0x8801, 0x48C0, 0x4980, 0x8941, 0x4B00, 0x8BC1, 0x8A81, 0x4A40,
-    0x4E00, 0x8EC1, 0x8F81, 0x4F40, 0x8D01, 0x4DC0, 0x4C80, 0x8C41,
-    0x4400, 0x84C1, 0x8581, 0x4540, 0x8701, 0x47C0, 0x4680, 0x8641,
-    0x8201, 0x42C0, 0x4380, 0x8341, 0x4100, 0x81C1, 0x8081, 0x4040
-};
-
-static inline uint16_t crc16(const uint8_t *data, size_t length) {
-    uint16_t crc = 0xFFFF;
+static inline uint32_t crc32(uint8_t* data, uint32_t length) {
+    uint32_t crc = 0xFFFFFFFF;
 
-    for (size_t i = 0; i < length; i++) {
-        uint8_t byte = data[i];
-        crc = (crc >> 8) ^ crc16_table[(crc ^ byte) & 0xFF];
+    while (length >= 8) {
+        crc = __crc32d(crc, *(const uint64_t*)data);
+        data += 8;
+        length -= 8;
+    }
+
+    while (length > 0) {
+        crc = __crc32b(crc, *data);
+        data++;
+        length--;
     }
 
-    return crc ^ 0xFFFF;
+    return ~crc;
 }
 
 enum StackCreatingState {
@@ -325,7 +303,8 @@ static inline struct SwiftNetPacketInfo construct_packet_info(const uint32_t pac
         .chunk_amount = chunk_amount,
         .chunk_index = chunk_index,
         .maximum_transmission_unit = maximum_transmission_unit,
-        .port_info = port_info
+        .port_info = port_info,
+        .checksum = 0x00
     };
 }
 
@@ -339,7 +318,7 @@ static struct ip construct_ip_header(struct in_addr destination_addr, const uint
         .ip_id = htons(packet_id), // Packet id
         .ip_off = htons(0), // Not used
         .ip_ttl = 64,// Time to live
-        .ip_sum = htons(0), // Checksum
+        .ip_sum = htons(0), // not used!!
         .ip_src = private_ip_address, // Source ip
         .ip_dst = destination_addr // Destination ip
     };

src/internal/pcap_open.c

@@ -1,4 +1,5 @@
 #include "internal.h"
+#include <pcap/pcap.h>
 
 pcap_t* swiftnet_pcap_open(const char* interface) {
     char errbuf[PCAP_ERRBUF_SIZE];

src/internal/pcap_send.c

@@ -1,7 +1,15 @@
 #include "internal.h"
+#include <errno.h>
+#include <pcap/pcap.h>
 
 int swiftnet_pcap_send(pcap_t *pcap, const u_char *data, int len) {
-    if (pcap_inject(pcap, data, len) == -1) {
+    int ret = pcap_inject(pcap, data, len);
+
+    if (ret == -1) {
+        if (errno == ENOBUFS) {
+            return -2;
+        }
+
         PRINT_ERROR("inject error: %s", pcap_geterr(pcap));
         return -1;
     }

src/memory_cleanup_background_service.c

@@ -59,13 +59,15 @@ void* memory_cleanup_background_service() {
 
         gettimeofday(&end, NULL);
 
-        uint32_t diff_us = (uint32_t)((end.tv_sec - start.tv_sec) * 1000000 + (end.tv_usec - start.tv_usec));
+        int64_t elapsed_us = (int64_t)(end.tv_sec - start.tv_sec) * 1000000LL + (end.tv_usec - start.tv_usec);
 
-        if (diff_us > PACKET_HISTORY_STORE_TIME * 1000000) {
-            continue;
-        }
+        if (elapsed_us < 0) elapsed_us = 0;
+
+        uint64_t target_us = (uint64_t)PACKET_HISTORY_STORE_TIME * 1000000ULL;
 
-        usleep((PACKET_HISTORY_STORE_TIME * 1000000) - diff_us);
+        if ((uint64_t)elapsed_us < target_us) {
+            usleep(target_us - (uint64_t)elapsed_us);
+        }
     }
 
     return NULL;