Previously only atomic scheduler queues where supported. This was a bottleneck when the number of worker threads was increased. Added an option to choose scheduler queue sync mode (none, atomic, ordered).
Signed-off-by: Matias Elo <[email protected]> --- test/performance/odp_l2fwd.c | 94 ++++++++++++++++++++++++-------------------- 1 file changed, 51 insertions(+), 43 deletions(-) diff --git a/test/performance/odp_l2fwd.c b/test/performance/odp_l2fwd.c index ab3c143..34abddc 100644 --- a/test/performance/odp_l2fwd.c +++ b/test/performance/odp_l2fwd.c @@ -41,25 +41,19 @@ #define SHM_PKT_POOL_BUF_SIZE 1856 /** @def MAX_PKT_BURST - * @brief Maximum number of packet bursts + * @brief Maximum number of packet in a burst */ #define MAX_PKT_BURST 16 -/** @def APPL_MODE_PKT_BURST - * @brief The application will handle pakcets in bursts +/** + * Packet input mode */ -#define APPL_MODE_PKT_BURST 0 - -/** @def APPL_MODE_PKT_QUEUE - * @brief The application will handle packets in queues - */ -#define APPL_MODE_PKT_QUEUE 1 - -/** @def PRINT_APPL_MODE(x) - * @brief Macro to print the current status of how the application handles - * packets. - */ -#define PRINT_APPL_MODE(x) printf("%s(%i)\n", #x, (x)) +typedef enum pkt_in_mode_t { + DIRECT_RECV, + SCHED_NONE, + SCHED_ATOMIC, + SCHED_ORDERED, +} pkt_in_mode_t; /** Get rid of path in filename - only for unix-type paths using '/' */ #define NO_PATH(file_name) (strrchr((file_name), '/') ? \ @@ -71,7 +65,7 @@ typedef struct { int cpu_count; int if_count; /**< Number of interfaces to be used */ char **if_names; /**< Array of pointers to interface names */ - int mode; /**< Packet IO mode */ + pkt_in_mode_t mode; /**< Packet input mode */ int time; /**< Time in seconds to run. */ int accuracy; /**< Number of seconds to get and print statistics */ char *if_str; /**< Storage for interface names */ @@ -206,11 +200,11 @@ static inline int lookup_dest_port(odp_packet_t pkt) } /** - * Packet IO worker thread using bursts from/to IO resources + * Packet IO worker thread accessing IO resources directly * * @param arg thread arguments of type 'thread_args_t *' */ -static void *pktio_ifburst_thread(void *arg) +static void *pktio_direct_recv_thread(void *arg) { int thr; thread_args_t *thr_args; @@ -231,7 +225,7 @@ static void *pktio_ifburst_thread(void *arg) pktio_dst = gbl_args->pktios[dst_idx]; printf("[%02i] srcif:%s dstif:%s spktio:%02" PRIu64 - " dpktio:%02" PRIu64 " BURST mode\n", + " dpktio:%02" PRIu64 " DIRECT RECV mode\n", thr, gbl_args->appl.if_names[src_idx], gbl_args->appl.if_names[dst_idx], @@ -278,13 +272,11 @@ static void *pktio_ifburst_thread(void *arg) * * @param dev Name of device to open * @param pool Pool to associate with device for packet RX/TX - * @param mode Packet processing mode for this device (BURST or QUEUE) * * @return The handle of the created pktio object. * @retval ODP_PKTIO_INVALID if the create fails. */ -static odp_pktio_t create_pktio(const char *dev, odp_pool_t pool, - int mode) +static odp_pktio_t create_pktio(const char *dev, odp_pool_t pool) { char inq_name[ODP_QUEUE_NAME_LEN]; odp_queue_param_t qparam; @@ -292,10 +284,11 @@ static odp_pktio_t create_pktio(const char *dev, odp_pool_t pool, odp_pktio_t pktio; int ret; odp_pktio_param_t pktio_param; + odp_schedule_sync_t sync_mode; memset(&pktio_param, 0, sizeof(pktio_param)); - if (mode == APPL_MODE_PKT_BURST) + if (gbl_args->appl.mode == DIRECT_RECV) pktio_param.in_mode = ODP_PKTIN_MODE_RECV; else pktio_param.in_mode = ODP_PKTIN_MODE_SCHED; @@ -309,13 +302,20 @@ static odp_pktio_t create_pktio(const char *dev, odp_pool_t pool, printf("created pktio %" PRIu64 " (%s)\n", odp_pktio_to_u64(pktio), dev); - /* no further setup needed for burst mode */ - if (mode == APPL_MODE_PKT_BURST) + /* no further setup needed for direct receive mode */ + if (gbl_args->appl.mode == DIRECT_RECV) return pktio; + if (gbl_args->appl.mode == SCHED_ATOMIC) + sync_mode = ODP_SCHED_SYNC_ATOMIC; + else if (gbl_args->appl.mode == SCHED_ORDERED) + sync_mode = ODP_SCHED_SYNC_ORDERED; + else + sync_mode = ODP_SCHED_SYNC_NONE; + odp_queue_param_init(&qparam); qparam.sched.prio = ODP_SCHED_PRIO_DEFAULT; - qparam.sched.sync = ODP_SCHED_SYNC_ATOMIC; + qparam.sched.sync = sync_mode; qparam.sched.group = ODP_SCHED_GROUP_ALL; snprintf(inq_name, sizeof(inq_name), "%" PRIu64 "-pktio_inq_def", odp_pktio_to_u64(pktio)); @@ -470,8 +470,7 @@ int main(int argc, char *argv[]) odp_pool_print(pool); for (i = 0; i < gbl_args->appl.if_count; ++i) { - pktio = create_pktio(gbl_args->appl.if_names[i], - pool, gbl_args->appl.mode); + pktio = create_pktio(gbl_args->appl.if_names[i], pool); if (pktio == ODP_PKTIO_INVALID) exit(EXIT_FAILURE); gbl_args->pktios[i] = pktio; @@ -484,7 +483,7 @@ int main(int argc, char *argv[]) } /* Save interface default output queue */ - if (gbl_args->appl.mode == APPL_MODE_PKT_QUEUE) + if (gbl_args->appl.mode != DIRECT_RECV) gbl_args->outq_def[i] = odp_pktio_outq_getdef(pktio); /* Save destination eth address */ @@ -518,9 +517,9 @@ int main(int argc, char *argv[]) odp_cpumask_t thd_mask; void *(*thr_run_func) (void *); - if (gbl_args->appl.mode == APPL_MODE_PKT_BURST) - thr_run_func = pktio_ifburst_thread; - else /* APPL_MODE_PKT_QUEUE */ + if (gbl_args->appl.mode == DIRECT_RECV) + thr_run_func = pktio_direct_recv_thread; + else /* SCHED_NONE / SCHED_ATOMIC / SCHED_ORDERED */ thr_run_func = pktio_queue_thread; gbl_args->thread[i].src_idx = i % gbl_args->appl.if_count; @@ -632,7 +631,6 @@ static void parse_args(int argc, char *argv[], appl_args_t *appl_args) appl_args->time = 0; /* loop forever if time to run is 0 */ appl_args->accuracy = 1; /* get and print pps stats second */ - appl_args->mode = -1; /* Invalid, must be changed by parsing */ while (1) { opt = getopt_long(argc, argv, "+c:+t:+a:i:m:d:h", @@ -693,10 +691,14 @@ static void parse_args(int argc, char *argv[], appl_args_t *appl_args) break; case 'm': i = atoi(optarg); - if (i == 0) - appl_args->mode = APPL_MODE_PKT_BURST; + if (i == 1) + appl_args->mode = SCHED_NONE; + else if (i == 2) + appl_args->mode = SCHED_ATOMIC; + else if (i == 3) + appl_args->mode = SCHED_ORDERED; else - appl_args->mode = APPL_MODE_PKT_QUEUE; + appl_args->mode = DIRECT_RECV; break; case 'd': appl_args->dst_change = atoi(optarg); @@ -710,7 +712,7 @@ static void parse_args(int argc, char *argv[], appl_args_t *appl_args) } } - if (appl_args->if_count == 0 || appl_args->mode == -1) { + if (appl_args->if_count == 0) { usage(argv[0]); exit(EXIT_FAILURE); } @@ -746,10 +748,14 @@ static void print_info(char *progname, appl_args_t *appl_args) printf(" %s", appl_args->if_names[i]); printf("\n" "Mode: "); - if (appl_args->mode == APPL_MODE_PKT_BURST) - PRINT_APPL_MODE(APPL_MODE_PKT_BURST); - else - PRINT_APPL_MODE(APPL_MODE_PKT_QUEUE); + if (appl_args->mode == DIRECT_RECV) + printf("DIRECT_RECV"); + else if (appl_args->mode == SCHED_NONE) + printf("SCHED_NONE"); + else if (appl_args->mode == SCHED_ATOMIC) + printf("SCHED_ATOMIC"); + else if (appl_args->mode == SCHED_ORDERED) + printf("SCHED_ORDERED"); printf("\n\n"); fflush(NULL); } @@ -770,10 +776,12 @@ static void usage(char *progname) "\n" "Mandatory OPTIONS:\n" " -i, --interface Eth interfaces (comma-separated, no spaces)\n" - " -m, --mode 0: Burst send&receive packets (no queues)\n" - " 1: Send&receive packets through ODP queues.\n" "\n" "Optional OPTIONS\n" + " -m, --mode 0: Send&receive packets directly from NIC (default)\n" + " 1: Send&receive packets through scheduler sync none queues\n" + " 2: Send&receive packets through scheduler sync atomic queues\n" + " 3: Send&receive packets through scheduler sync ordered queues\n" " -c, --count <number> CPU count.\n" " -t, --time <number> Time in seconds to run.\n" " -a, --accuracy <number> Time in seconds get print statistics\n" -- 1.9.1 _______________________________________________ lng-odp mailing list [email protected] https://lists.linaro.org/mailman/listinfo/lng-odp
