Migrate performance enhancements from other branch.

CMakeLists.txt

@@ -5,7 +5,8 @@ set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/build)
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/build)
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
 
-set(CMAKE_CXX_FLAGS "-fstack-protector -fstack-protector-all -Wall -pedantic " )
+SET(CMAKE_C_FLAGS "-O2 -Wall -std=c11 -D_XOPEN_SOURCE=600" )
+set(CMAKE_CXX_FLAGS "-fstack-protector -fstack-protector-all -Wall -std=c++11 -D_XOPEN_SOURCE=600" )
 set(CMAKE_CXX_FLAGS_DEBUG "-O2 -Wall -ggdb" )
 set(CMAKE_CXX_FLAGS_RELEASE "-Os -Wall" )
 

prrt/proto/block.c

@@ -7,7 +7,7 @@
 #include "block.h"
 #include "codingParams.h"
 
-void gather_redundancy_packets(const PrrtBlock *block_ptr, gf *const *fec, int *idx_p)
+static void gather_redundancy_packets(const PrrtBlock *block_ptr, gf *const *fec, int *idx_p)
 {
     uint32_t i;
     uint32_t m = 0;
@@ -27,7 +27,7 @@ void gather_redundancy_packets(const PrrtBlock *block_ptr, gf *const *fec, int *
     }
 }
 
-void gather_data_packets(PrrtBlock *block_ptr, gf *const *fec, int *idx_p)
+static void gather_data_packets(PrrtBlock *block_ptr, gf *const *fec, int *idx_p)
 {
     LIST_FOREACH(block_ptr->dataPackets, first, next, current) {
         PrrtPacket *packet = current->value;
@@ -36,7 +36,7 @@ void gather_data_packets(PrrtBlock *block_ptr, gf *const *fec, int *idx_p)
     }
 }
 
-void clear_list(gf *const *src, uint8_t k)
+static void clear_list(gf *const *src, uint8_t k)
 {
     int j = 0;
     for(j = 0; j < k; j++) {

prrt/proto/packet.c

@@ -10,19 +10,19 @@
 #include "packet.h"
 #include "clock.h"
 
-void *encode_general_header(void *buf_ptr, const PrrtPacket *packet);
+static void *encode_general_header(void *buf_ptr, const PrrtPacket *packet);
 
-void *encode_data_header(void *buf_ptr, const void *payload);
+static void *encode_data_header(void *buf_ptr, const void *payload);
 
-void *encode_redundancy_header(void *buf_ptr, const void *payload);
+static void *encode_redundancy_header(void *buf_ptr, const void *payload);
 
-void *encode_feedback_header(void *buf_ptr, const void *payload);
+static void *encode_feedback_header(void *buf_ptr, const void *payload);
 
-void *decode_data_header(void *dstBuffer, const void *srcBuffer);
+static void *decode_data_header(void *dstBuffer, const void *srcBuffer);
 
-void *decode_redundancy_header(void *dstBuffer, const void *srcBuffer);
+static void *decode_redundancy_header(void *dstBuffer, const void *srcBuffer);
 
-void *decode_feedback_header(void *dstBuffer, const void *srcBuffer);
+static void *decode_feedback_header(void *dstBuffer, const void *srcBuffer);
 
 prrtPacketType_t PrrtPacket_type(PrrtPacket *packet_ptr)
 {
@@ -131,7 +131,7 @@ PrrtPacket *PrrtPacket_copy(PrrtPacket *original)
     return NULL;
 }
 
-PrrtPacket *create_header(uint8_t priority, prrtSequenceNumber_t seqno, prrtPacketLength_t size, uint8_t type, uint8_t index)
+static PrrtPacket *create_header(uint8_t priority, prrtSequenceNumber_t seqno, prrtPacketLength_t size, uint8_t type, uint8_t index)
 {
     PrrtPacket *packet = calloc(1, sizeof(PrrtPacket));
     check_mem(packet);
@@ -504,4 +504,4 @@ PrrtPacket *PrrtPacket_create_feedback_packet(uint8_t priority, uint8_t index, p
 
     error:
     return NULL;
-}
\ No newline at end of file
+}

prrt/proto/processes/dataReceiver.c

@@ -9,7 +9,7 @@
 #include "../socket.h"
 #include "dataReceiver.h"
 
-void retrieve_data_blocks(PrrtSocket *sock_ptr, prrtSequenceNumber_t base_seqno, uint8_t k, const PrrtBlock *block)
+static void retrieve_data_blocks(PrrtSocket *sock_ptr, prrtSequenceNumber_t base_seqno, uint8_t k, const PrrtBlock *block)
 {
     List *res = List_create();
 
@@ -27,7 +27,7 @@ void retrieve_data_blocks(PrrtSocket *sock_ptr, prrtSequenceNumber_t base_seqno,
     List_destroy(res);
 }
 
-void decode_block(PrrtSocket *socket, PrrtBlock *block)
+static void decode_block(PrrtSocket *socket, PrrtBlock *block)
 {
     if(block != NULL && PrrtBlock_decode_ready(block)) {
         check(PrrtBlock_decode(block), "Decoding failed");
@@ -54,7 +54,7 @@ void decode_block(PrrtSocket *socket, PrrtBlock *block)
     PERROR("Decoding failed.%s", "")
 }
 
-bool send_feedback(const PrrtSocket *sock_ptr, struct sockaddr_in remote)
+static bool send_feedback(const PrrtSocket *sock_ptr, struct sockaddr_in remote)
 {
     uint16_t remote_port = ntohs(remote.sin_port);
     char *remote_host = inet_ntoa(remote.sin_addr);
@@ -94,7 +94,7 @@ bool send_feedback(const PrrtSocket *sock_ptr, struct sockaddr_in remote)
     return false;
 }
 
-void handle_data_packet(PrrtSocket *sock_ptr, PrrtPacket *packet, struct sockaddr_in remote)
+static void handle_data_packet(PrrtSocket *sock_ptr, PrrtPacket *packet, struct sockaddr_in remote)
 {
     PrrtPacketDataPayload *payload = packet->payload;
 
@@ -146,7 +146,7 @@ void handle_data_packet(PrrtSocket *sock_ptr, PrrtPacket *packet, struct sockadd
     return;
 }
 
-void handle_redundancy_packet(PrrtSocket *socket, PrrtPacket *packet)
+static void handle_redundancy_packet(PrrtSocket *socket, PrrtPacket *packet)
 {
     PrrtPacketRedundancyPayload *redundancyPayload = packet->payload;
 

prrt/proto/processes/dataTransmitter.c

@@ -11,7 +11,7 @@
 #include "dataTransmitter.h"
 
 
-bool send_packet(PrrtSocket *sock_ptr, PrrtPacket *packet)
+static bool send_packet(PrrtSocket *sock_ptr, PrrtPacket *packet)
 {
     uint8_t buf[MAX_PAYLOAD_LENGTH];
     memset(buf, 0, sizeof(buf));
@@ -24,19 +24,8 @@ bool send_packet(PrrtSocket *sock_ptr, PrrtPacket *packet)
     LIST_FOREACH(sock_ptr->receivers, first, next, cur) {
         PrrtReceiver *recv = cur->value;
 
-        struct hostent *hp;
-
-        struct sockaddr_in targetaddr;
-        memset((char *) &targetaddr, 0, sizeof(targetaddr));
-        targetaddr.sin_family = AF_INET;
-        targetaddr.sin_port = htons(recv->port);
-
-        hp = gethostbyname(recv->host_name);
-        check(hp != NULL, "Could not resolve host '%s'.", recv->host_name)
-        memcpy((void *) &targetaddr.sin_addr, hp->h_addr_list[0], (size_t) hp->h_length);
-
         // TODO: [LATENCY] By knowing the time encoding etc. that happens upfront, one could make an adjustment here.
-        check(sendto(sock_ptr->dataSocketFd, buf, length, 0, (struct sockaddr *) &targetaddr, sizeof(targetaddr)) ==
+        check(sendto(sock_ptr->dataSocketFd, buf, length, 0, recv->ai->ai_addr, recv->ai->ai_addrlen) ==
               length, "Sendto failed.");
         usleep(1);
     }
@@ -61,17 +50,14 @@ void *send_data_loop(void *ptr)
     while(1) {
         check(pthread_mutex_lock(&sock_ptr->outQueueFilledMutex) == 0, "Lock failed.");
         while(List_count(sock_ptr->outQueue) == 0) {
-            check(pthread_mutex_lock(&sock_ptr->closingMutex) == 0, "Lock failed.");
-            if(sock_ptr->closing) {
+            if (atomic_load_explicit(&sock_ptr->closing, memory_order_acquire)) {
                 PrrtCodingParams_destroy(cpar);
                 if(block != NULL) {
                     PrrtBlock_destroy(block);
                 }
-                check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
                 check(pthread_mutex_unlock(&sock_ptr->outQueueFilledMutex) == 0, "Unlock failed.");
                 return NULL;
             }
-            check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
             check(pthread_cond_wait(&sock_ptr->outQueueFilledCv, &sock_ptr->outQueueFilledMutex) == 0,
                   "Cond wait failed.");
         }

prrt/proto/processes/feedbackReceiver.c

@@ -10,7 +10,7 @@
 #include "../socket.h"
 #include "feedbackReceiver.h"
 
-void handle_feedback(PrrtSocket *prrtSocket, const size_t length)
+static void handle_feedback(PrrtSocket *prrtSocket, const size_t length)
 {
     char bufin[MAX_PAYLOAD_LENGTH];
     PrrtPacket *prrtPacket = NULL;
@@ -50,18 +50,14 @@ void *receive_feedback_loop(void *ptr)
 {
     PrrtSocket *sock_ptr = ptr;
 
-    check(pthread_mutex_lock(&sock_ptr->closingMutex) == 0, "Lock failed.");
-    while(sock_ptr->closing == false) {
-        check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
+    while (!atomic_load_explicit(&sock_ptr->closing, memory_order_acquire)) {
         handle_feedback(sock_ptr, MAX_PAYLOAD_LENGTH);
         usleep(1);
-        check(pthread_mutex_lock(&sock_ptr->closingMutex) == 0, "Lock failed.");
     }
-    check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
 
     return NULL;
 
-    error:
-    PERROR("Feedback reception failed.%s","");
-    return NULL;
-}
\ No newline at end of file
+//    error:
+//    PERROR("Feedback reception failed.%s","");
+//    return NULL;
+}

prrt/proto/receiver.c

@@ -8,18 +8,34 @@ PrrtReceiver *PrrtReceiver_create(const char *host, uint16_t port)
     PrrtReceiver *recv = calloc(1, sizeof(PrrtReceiver));
     check_mem(recv);
     recv->host_name = strdup(host);
+	check_mem(recv->host_name);
     recv->port = port;
 
+	struct addrinfo *info;
+	struct addrinfo hints;
+	char portstr[sizeof(port)*8+1];
+	snprintf(portstr, sizeof(portstr), "%u", (unsigned int) port);
+	memset(&hints, 0x0, sizeof(hints));
+	hints.ai_family = AF_INET;
+	hints.ai_socktype = SOCK_DGRAM;
+	check(0 == getaddrinfo(host, portstr, &hints, &info), "getaddrinfo");
+
+	recv->ai = info;
+
     return recv;
 
     error:
-    if(recv != NULL) { free(recv); }
-    PERROR("Memory issue.%s","");
+    if (recv != NULL) {
+		free((void *) recv->host_name);
+		free(recv);
+	}
+    //PERROR("Memory issue.%s","");
     return NULL;
 }
 
 bool PrrtReceiver_destroy(PrrtReceiver *receiver)
 {
+	freeaddrinfo(receiver->ai);
     free((void *) receiver->host_name);
     free(receiver);
     return true;

prrt/proto/receiver.h

@@ -3,10 +3,14 @@
 
 #include <stdbool.h>
 #include <stdint.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <netdb.h>
 
 typedef struct prrtReceiver {
     const char* host_name;
     uint16_t port;
+	struct addrinfo *ai;
 } PrrtReceiver;
 
 PrrtReceiver* PrrtReceiver_create(const char *host, uint16_t port);

prrt/proto/socket.c

@@ -47,11 +47,6 @@ PrrtSocket *PrrtSocket_create(const bool is_sender)
     check(sock_ptr->feedbackSocketFd = socket(AF_INET, SOCK_DGRAM, 0), "Cannot create feedback socket.");
     check(setsockopt(sock_ptr->feedbackSocketFd, SOL_SOCKET, SO_BROADCAST, &enabled, sizeof(enabled)) == EXIT_SUCCESS, "Socket option set failed.");
 
-    pthread_mutexattr_t attr;
-    check(pthread_mutexattr_init(&attr) == 0, "Mutex attr init failed.");
-    check(pthread_mutexattr_settype(&attr, PTHREAD_MUTEX_RECURSIVE) == 0, "Setting type failed.");
-    check(pthread_mutex_init(&sock_ptr->closingMutex, &attr) == 0, "Mutex init failed.");
-
     if (is_sender) {
         check(pthread_mutex_init(&sock_ptr->outQueueFilledMutex, NULL) == 0, "Mutex init failed.");
         check(pthread_cond_init(&sock_ptr->outQueueFilledCv, NULL) == 0, "Cond init failed.");
@@ -143,13 +138,10 @@ int32_t PrrtSocket_recv(PrrtSocket *sock_ptr, void *buf_ptr) {
     while (1) {
         check(pthread_mutex_lock(&sock_ptr->inQueueFilledMutex) == 0, "Lock failed.");
         while (List_count(sock_ptr->inQueue) == 0) {
-            check(pthread_mutex_lock(&sock_ptr->closingMutex) == 0, "Lock failed.");
-            if (sock_ptr->closing) {
-                check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
+            if (atomic_load_explicit(&sock_ptr->closing, memory_order_acquire)) {
                 check(pthread_mutex_unlock(&sock_ptr->inQueueFilledMutex) == 0, "Unlock failed.");
                 return -1;
             }
-            check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
             check(pthread_cond_wait(&sock_ptr->inQueueFilledCv, &sock_ptr->inQueueFilledMutex) == 0, "Wait failed.");
         }
         PrrtPacket *packet = List_shift(sock_ptr->inQueue);
@@ -169,9 +161,7 @@ int32_t PrrtSocket_recv(PrrtSocket *sock_ptr, void *buf_ptr) {
 }
 
 int PrrtSocket_interrupt(PrrtSocket *sock_ptr) {
-    check(pthread_mutex_lock(&sock_ptr->closingMutex) == 0, "Lock failed.");
-    sock_ptr->closing = true;
-    check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
+    atomic_store_explicit(&sock_ptr->closing, true, memory_order_release);
 
     void **res = NULL;
     if (sock_ptr->sendDataThread != 0) {
@@ -215,12 +205,8 @@ int PrrtSocket_interrupt(PrrtSocket *sock_ptr) {
 
 int PrrtSocket_close(PrrtSocket *sock_ptr) {
     debug("Closing socket.");
-    check(pthread_mutex_lock(&sock_ptr->closingMutex) == 0, "Lock failed.");
-    if (!sock_ptr->closing) {
-        check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
+    if (!atomic_load_explicit(&sock_ptr->closing, memory_order_acquire)) {
         check(PrrtSocket_interrupt(sock_ptr) == EXIT_SUCCESS, "Interrupt failed.");
-    } else {
-        check(pthread_mutex_unlock(&sock_ptr->closingMutex) == 0, "Unlock failed.");
     }
 
     if(sock_ptr->dataPacketStore != NULL) {
@@ -266,8 +252,6 @@ int PrrtSocket_close(PrrtSocket *sock_ptr) {
         sock_ptr->packetTimeoutTable = NULL;
     }
 
-    check(pthread_mutex_destroy(&sock_ptr->closingMutex) == 0, "Mutex destroy failed.");
-
     if(sock_ptr->address != NULL) {
         free(sock_ptr->address);
     }

prrt/proto/socket.h

 #ifndef PRRT_SOCKET_H
 #define PRRT_SOCKET_H
 
+#include <stdatomic.h>
+
 #include "../defines.h"
 #include "packet.h"
 #include "stores/forwardPacketTable.h"
@@ -45,8 +47,7 @@ typedef struct prrtSocket {
 
     List* receivers;
 
-    pthread_mutex_t closingMutex;
-    bool closing;
+    atomic_bool closing;
 
     prrtSequenceNumber_t packetsCount;
     prrtSequenceNumber_t sequenceNumberSource;

prrt/proto/vdmcode/block_code.c

@@ -314,7 +314,7 @@ static int invert_mat(PrrtCoder *cod, gf *src, int k) {
         goto fail;
     }
 
-    bzero(id_row, k * sizeof(gf));
+    memset(id_row, 0x0, k * sizeof(gf));
     /*
     * ipiv marks elements already used as pivots.
     */
@@ -391,7 +391,7 @@ static int invert_mat(PrrtCoder *cod, gf *src, int k) {
         * we can optimize the addmul).
         */
         id_row[icol] = 1;
-        if (bcmp(pivot_row, id_row, k * sizeof(gf)) != 0) {
+        if (memcmp(pivot_row, id_row, k * sizeof(gf)) != 0) {
             for (p = src, ix = 0; ix < k; ix++, p += k) {
                 if (ix != icol) {
                     c = p[icol];
@@ -435,7 +435,7 @@ static int invert_mat(PrrtCoder *cod, gf *src, int k) {
  * q = values of the polynomial (known)
  */
 
-int invert_vdm(PrrtCoder *cod, gf *src, int k) {
+static int invert_vdm(PrrtCoder *cod, gf *src, int k) {
     int i, j, row, col;
     gf *b, *c, *p;
     gf t, xx;
@@ -617,7 +617,7 @@ int prrt_coder_create(PrrtCoder **cod, uint8_t k, uint8_t n) {
     /*
     * the upper matrix is I so do not bother with a slow multiply
     */
-    bzero((*cod)->params.enc_matrix, k * k * sizeof(gf));
+    memset((*cod)->params.enc_matrix, 0x0, k * k * sizeof(gf));
     for (p = (*cod)->params.enc_matrix, col = 0; col < k; col++, p += k + 1) {
         *p = 1;
     }
@@ -645,11 +645,11 @@ void PrrtCoder_encode(PrrtCoder *cod, gf **src, gf *fec, int index, int sz) {
     }
 
     if (index < k) {
-        bcopy(src[index], fec, sz * sizeof(gf));
+        memcpy(fec, src[index], sz * sizeof(gf));
     }
     else if (index < cod->params.n) {
         p = &(cod->params.enc_matrix[index * k]);
-        bzero(fec, sz * sizeof(gf));
+        memset(fec, 0x0, sz * sizeof(gf));
         for (i = 0; i < k; i++) {
             addmul(cod, fec, src[i], p[i], sz);
         }
@@ -672,12 +672,12 @@ static gf *prrt_coder_build_matrix(PrrtCoder *cod, int *index) {
     for (i = 0, p = matrix; i < k; i++, p += k) {
 #if 1 /* this is simply an optimization, not very useful indeed */
         if (index[i] < k) {
-            bzero(p, k * sizeof(gf));
+            memset(p, 0x0, k * sizeof(gf));
             p[i] = 1;
         } else
 #endif
         if (index[i] < cod->params.n) {
-            bcopy(&(cod->params.enc_matrix[index[i] * k]), p, k * sizeof(gf));
+            memcpy(p, &(cod->params.enc_matrix[index[i] * k]), k * sizeof(gf));
         }
         else {
             PERROR("decode: invalid index %d (max %d)\n",
@@ -729,7 +729,7 @@ int PrrtCoder_decode(PrrtCoder *cod, gf **pkt, int *index, int sz) {
         err = -1;
         goto error;
     }
-    bzero(new_pkt, k * sizeof(gf *));
+    memset(new_pkt, 0x0, k * sizeof(gf *));
 
     for (row = 0; row < k; row++) {
         if (index[row] >= k) {
@@ -738,7 +738,7 @@ int PrrtCoder_decode(PrrtCoder *cod, gf **pkt, int *index, int sz) {
                 PERROR("No memory for packet row\n");
                 goto free_rows;
             }
-            bzero(new_pkt[row], sz * sizeof(gf));
+            memset(new_pkt[row], 0x0, sz * sizeof(gf));
             for (col = 0; col < k; col++) {
                 addmul(cod, new_pkt[row], pkt[col], m_dec[row * k + col], sz);
             }
@@ -749,7 +749,7 @@ int PrrtCoder_decode(PrrtCoder *cod, gf **pkt, int *index, int sz) {
     */
     for (row = 0; row < k; row++) {
         if (index[row] >= k) {
-            bcopy(new_pkt[row], pkt[row], sz * sizeof(gf));
+            memcpy(pkt[row], new_pkt[row], sz * sizeof(gf));
             free(new_pkt[row]);
         }
     }
@@ -771,4 +771,4 @@ int PrrtCoder_decode(PrrtCoder *cod, gf **pkt, int *index, int sz) {
     free(m_dec);
 
     return -1;
-}
\ No newline at end of file
+}

prrt/util/bitmap.c

@@ -12,9 +12,10 @@ Bitmap *Bitmap_create(bool initialValue, uint32_t elementCount)
     check_mem(bitmap);
     bitmap->count = elementCount;
 
-    uint32_t blockCount = (uint32_t) ceil(elementCount / 32.0);
+    uint32_t blockCount = (elementCount + 31) / 32;
 
     bitmap->data = calloc(blockCount, sizeof(uint32_t));
+	check_mem(bitmap->data);
     if(initialValue == true) {
         memset(bitmap->data, -1, blockCount * sizeof(uint32_t));
     }
@@ -25,32 +26,22 @@ Bitmap *Bitmap_create(bool initialValue, uint32_t elementCount)
         return NULL;
 }
 
-bool Bitmap_get(Bitmap *bitmap, uint32_t position)
+void Bitmap_set_range_0(Bitmap *bitmap, uint32_t start, uint32_t length)
 {
-    uint32_t byte_offset = position / 32;
-    uint32_t bit_offset = position % 32;
-    return (bool) ((bitmap->data[byte_offset] & (1 << bit_offset)) != 0);
-}
-
-bool Bitmap_set(Bitmap *bitmap, uint32_t position, bool value)
-{
-    uint32_t byte_offset = position / 32;
-    uint32_t bit_offset = position % 32;
-    if(value) {
-        bitmap->data[byte_offset] |= (1 << bit_offset);
-    } else {
-        bitmap->data[byte_offset] &= ~(1 << bit_offset);
-    }
-    return true;
+    uint32_t i = 0;
+    // TODO: highly inefficient - optimize
+	for(i = 0; i < length; i++) {
+		Bitmap_set_0(bitmap, start + i);
+	}
 }
 
-void Bitmap_set_range(Bitmap *bitmap, uint32_t start, uint32_t length, bool value)
+void Bitmap_set_range_1(Bitmap *bitmap, uint32_t start, uint32_t length)
 {
     uint32_t i = 0;
     // TODO: highly inefficient - optimize
-    for(i = 0; i < length; i++) {
-        Bitmap_set(bitmap, start + i, value);
-    }
+	for(i = 0; i < length; i++) {
+		Bitmap_set_1(bitmap, start + i);
+	}
 }
 
 uint32_t Bitmap_sum_ones(Bitmap *bitmap, uint32_t start, uint32_t length)
@@ -66,13 +57,7 @@ uint32_t Bitmap_sum_ones(Bitmap *bitmap, uint32_t start, uint32_t length)
 
 uint32_t Bitmap_sum_zeros(Bitmap *bitmap, uint32_t start, uint32_t length)
 {
-    uint32_t sum = 0;
-    uint32_t i = 0;
-    // TODO: highly inefficient - optimize
-    for(i = 0; i < length; i++) {
-        sum += (Bitmap_get(bitmap, start + i) == false);
-    }
-    return sum;
+    return length - Bitmap_sum_ones(bitmap, start, length);
 }
 
 bool Bitmap_destroy(Bitmap *bitmap)

prrt/util/bitmap.h

@@ -10,11 +10,56 @@ typedef struct bitmap {
 } Bitmap;
 
 Bitmap *Bitmap_create(bool initialValue, uint32_t elementCount);
-bool Bitmap_get(Bitmap* bitmap, uint32_t position);
-bool Bitmap_set(Bitmap* bitmap, uint32_t position, bool value);
-void Bitmap_set_range(Bitmap *bitmap, uint32_t start, uint32_t length, bool value);
+void Bitmap_set_range_0(Bitmap *bitmap, uint32_t start, uint32_t length);
+void Bitmap_set_range_1(Bitmap *bitmap, uint32_t start, uint32_t length);
 uint32_t Bitmap_sum_ones(Bitmap *bitmap, uint32_t start, uint32_t length);
 uint32_t Bitmap_sum_zeros(Bitmap *bitmap, uint32_t start, uint32_t length);
 bool Bitmap_destroy(Bitmap* bitmap);
 
+#ifdef __GNUC__
+#  define BITMAP_EXPOSED_FUNCTION static inline __attribute__((__always_inline__,__unused__))
+#else
+#  define BITMAP_EXPOSED_FUNCTION static inline
+#endif
+
+BITMAP_EXPOSED_FUNCTION bool Bitmap_get(Bitmap *bitmap, uint32_t position)
+{
+    uint32_t byte_offset = position / 32;
+    uint32_t bit_offset = position % 32;
+    return (bool) ((bitmap->data[byte_offset] & (1 << bit_offset)) != 0);
+}
+
+BITMAP_EXPOSED_FUNCTION bool Bitmap_set_0(Bitmap *bitmap, uint32_t position)
+{
+    uint32_t byte_offset = position / 32;
+    uint32_t bit_offset = position % 32;
+    bitmap->data[byte_offset] &= ~(1 << bit_offset);
+	return true;
+}
+
+BITMAP_EXPOSED_FUNCTION bool Bitmap_set_1(Bitmap *bitmap, uint32_t position)
+{
+    uint32_t byte_offset = position / 32;
+    uint32_t bit_offset = position % 32;
+    bitmap->data[byte_offset] |= (1 << bit_offset);
+	return true;
+}
+
+BITMAP_EXPOSED_FUNCTION bool Bitmap_set(Bitmap *bitmap, uint32_t position, bool value)
+{
+	if (value)
+		Bitmap_set_1(bitmap, position);
+	else
+		Bitmap_set_0(bitmap, position);
+    return true;
+}
+
+BITMAP_EXPOSED_FUNCTION void Bitmap_set_range(Bitmap *bitmap, uint32_t start, uint32_t length, bool value)
+{
+	if (value)
+		Bitmap_set_range_1(bitmap, start, length);
+	else
+		Bitmap_set_range_0(bitmap, start, length);
+}
+
 #endif //PRRT_BITMAP_H

prrt/util/common.c

@@ -11,6 +11,7 @@ int print_buffer(const char *buf, const int length) {
         printf("%2x", buf[i]);
     }
     printf("\n");
+	return 0;
 }
 
 void print_gf(const gf *start, const int len) {
@@ -24,4 +25,4 @@ void print_gf(const gf *start, const int len) {
         }
     }
     printf("\n");
-}
\ No newline at end of file
+}

src/prrt/timestamp.c

+#include "timestamp.h"
+
+#include <stdatomic.h>
+
+void PrrtTimestampTableDumpHeader(FILE *out)
+{
+	fprintf(out, "SeqNo" );
+#	define OUT(id) fprintf(out, ",%s_T,%s_C", #id, #id);
+	PP_foreach(PP_join_space, OUT, TIMESTAMP_ID_LIST)
+#	undef OUT
+	fprintf(out, "\n");
+}
+
+static inline unsigned long long timestampByTime(struct timespec *ts)
+{
+	// convert timespec to microseconds
+	unsigned long long x = ts->tv_sec;
+	x *= 1000000ULL;
+	x += ts->tv_nsec / 1000;
+	return x;
+}
+
+void PrrtTimestampTableDump(FILE *out, unsigned int seqno, PrrtTimestampTable *table)
+{
+	fprintf(out, "%u", seqno);
+#	define OUT(id) fprintf(out, ",%llu,%llu", timestampByTime(&table->time[ts_##id].actual.t), (unsigned long long) table->time[ts_##id].actual.c);
+	PP_foreach(PP_join_space, OUT, TIMESTAMP_ID_LIST)
+#	undef OUT
+	fprintf(out, "\n");
+}
+

src/prrt/timestamp.h

+#ifndef PRRT_TIMESTAMP_H
+#define PRRT_TIMESTAMP_H
+
+#include <stdio.h>
+#include <stdint.h>
+#include <time.h>
+
+/*
+ * Each timestamp contains both a cycle value and a time value, but some values
+ * might remain zero
+ */
+typedef union PrrtTimestamp {
+	struct PrrtActualTimestamp {
+		uint64_t c;
+		struct timespec t;
+	} actual;
+	char _cacheline[64];
+} PrrtTimestamp;
+
+#include "../util/pp.h"
+
+/*
+ * Comma-separated list of timestamp IDs
+ */
+#define TIMESTAMP_ID_LIST \
+	PrrtSendStart, \
+	PrrtSendEnd, \
+	PrrtEncodeStart, \
+	PrrtEncodeEnd, \
+	LinkTransmitStart, \
+	LinkTransmitEnd, \
+	LinkReceive, \
+	PrrtDeliver \
+
+#define TIMESSTAMP_ID_TO_NAME(id) ts_##id
+
+/*
+ * enum containing all timestamp IDs
+ */
+typedef enum PrrtTimestampId {
+	PP_foreach(PP_join_comma, TIMESSTAMP_ID_TO_NAME, TIMESTAMP_ID_LIST),
+	ts_count
+} PrrtTimestampId;
+
+
+/*
+ * Table that stores timestamps for each timestamp ID
+ */
+typedef struct PrrtTimestampTable {
+	PrrtTimestamp time[ts_count];
+} PrrtTimestampTable;
+
+/*
+ * update the clock value of a timestamp
+ *
+ * This macro will cause a SIGSEGV if the application does not install a
+ * timestamp table to the prrt socket.
+ */
+#define PrrtTimeStampClock(sck, id) do { \
+		clock_gettime(CLOCK_MONOTONIC, &atomic_load_explicit(&(sck)->tstable, memory_order_acquire)->time[ts_##id].actual.t); \
+	} while (0);
+
+/*
+ * update the cycle value of a timestamp
+ *
+ * This macro will cause a SIGSEGV if the application does not install a
+ * timestamp table to the prrt socket.
+ */
+#define PrrtTimeStampCycle(sck, id) do { \
+		atomic_load_explicit(&(sck)->tstable, memory_order_acquire)->time[ts_##id].actual.c = __builtin_ia32_rdtsc(); \
+	} while (0);
+
+/*
+ * install a time stamp table to a socket
+ */
+#define PrrtTimestampTableInstall(sck, tstp) do { \
+		PrrtTimestampTable *__tstp = (tstp); \
+		memset(__tstp, 0, sizeof(PrrtTimestampTable)); \
+		atomic_store_explicit(&(sck)->tstable, __tstp, memory_order_release); \
+	} while (0)
+
+/*
+ * print a timestamp dump header
+ */
+extern void PrrtTimestampTableDumpHeader(FILE *);
+
+/*
+ * dump a timestamp table
+ */
+extern void PrrtTimestampTableDump(FILE *, unsigned int, PrrtTimestampTable *);
+
+/*
+ * use a specific timestamp table
+ */
+extern void PrrtTimestampTableUse(PrrtTimestampTable *);
+
+
+#endif // PRRT_TIMESTAMP_H