summaryrefslogtreecommitdiffstats
path: root/VNFs/DPPD-PROX/display_latency.c
diff options
context:
space:
mode:
Diffstat (limited to 'VNFs/DPPD-PROX/display_latency.c')
-rw-r--r--VNFs/DPPD-PROX/display_latency.c154
1 files changed, 154 insertions, 0 deletions
diff --git a/VNFs/DPPD-PROX/display_latency.c b/VNFs/DPPD-PROX/display_latency.c
new file mode 100644
index 00000000..04382e46
--- /dev/null
+++ b/VNFs/DPPD-PROX/display_latency.c
@@ -0,0 +1,154 @@
+/*
+// Copyright (c) 2010-2017 Intel Corporation
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+*/
+
+#include "display.h"
+#include "display_latency.h"
+#include "stats_latency.h"
+#include "lconf.h"
+
+static struct display_column *min_col;
+static struct display_column *max_col;
+static struct display_column *avg_col;
+static struct display_column *stddev_col;
+static struct display_column *accuracy_limit_col;
+static struct display_column *used_col;
+static struct display_column *lost_col;
+static struct display_page display_page_latency;
+
+static void display_latency_draw_frame(struct screen_state *screen_state)
+{
+ const uint32_t n_latency = stats_get_n_latency();
+ struct display_column *core_col;
+ struct display_column *port_col;
+
+ display_page_init(&display_page_latency);
+
+ struct display_table *core = display_page_add_table(&display_page_latency);
+ struct display_table *port = display_page_add_table(&display_page_latency);
+ struct display_table *lat = display_page_add_table(&display_page_latency);
+ struct display_table *acc = display_page_add_table(&display_page_latency);
+ struct display_table *other = display_page_add_table(&display_page_latency);
+
+ display_table_init(core, "Core");
+ core_col = display_table_add_col(core);
+ display_column_init(core_col, "Nb", 4);
+
+ display_table_init(port, "Port Nb");
+ port_col = display_table_add_col(port);
+ display_column_init(port_col, "RX", 8);
+
+ if (screen_state->toggle == 0)
+ display_table_init(lat, "Measured Latency per interval");
+ else
+ display_table_init(lat, "Measured Latency since reset");
+
+ min_col = display_table_add_col(lat);
+ display_column_init(min_col, "Min (us)", 20);
+ max_col = display_table_add_col(lat);
+ display_column_init(max_col, "Max (us)", 20);
+ avg_col = display_table_add_col(lat);
+ display_column_init(avg_col, "Avg (us)", 20);
+ stddev_col = display_table_add_col(lat);
+ display_column_init(stddev_col, "Stddev (us)", 20);
+
+ display_table_init(acc, "Accuracy ");
+ used_col = display_table_add_col(acc);
+ display_column_init(used_col, "Used Packets (%)", 16);
+ accuracy_limit_col = display_table_add_col(acc);
+ display_column_init(accuracy_limit_col, "limit (us)", 16);
+
+ display_table_init(other, "Other");
+
+ lost_col = display_table_add_col(other);
+ display_column_init(lost_col, "Lost Packets", 16);
+
+ display_page_draw_frame(&display_page_latency, n_latency);
+
+ for (uint16_t i = 0; i < n_latency; ++i) {
+ uint32_t lcore_id = stats_latency_get_core_id(i);
+ uint32_t task_id = stats_latency_get_task_id(i);
+ struct task_args *targ = &lcore_cfg[lcore_id].targs[task_id];
+
+ display_column_print(core_col, i, "%2u/%1u", lcore_id, task_id);
+ display_column_port_ring(port_col, i, targ->rx_port_queue, targ->nb_rxports, targ->rx_rings, targ->nb_rxrings);
+ }
+}
+
+#define AFTER_POINT 1000000
+
+static void display_stats_latency_entry(int row, struct stats_latency *stats_latency)
+{
+ struct time_unit_err avg = stats_latency->avg;
+ struct time_unit_err min = stats_latency->min;
+ struct time_unit_err max = stats_latency->max;
+ struct time_unit_err stddev = stats_latency->stddev;
+ struct time_unit accuracy_limit = stats_latency->accuracy_limit;
+
+ uint32_t used = 0;
+
+ if (stats_latency->tot_all_packets)
+ used = stats_latency->tot_packets * (100 * AFTER_POINT) / stats_latency->tot_all_packets;
+
+ char dst[32];
+
+ if (stats_latency->tot_packets) {
+ display_column_print(min_col, row, "%s", print_time_unit_err_usec(dst, &min));
+ display_column_print(max_col, row, "%s", print_time_unit_err_usec(dst, &max));
+ display_column_print(avg_col, row, "%s", print_time_unit_err_usec(dst, &avg));
+ display_column_print(stddev_col, row, "%s", print_time_unit_err_usec(dst, &stddev));
+ } else {
+ display_column_print(min_col, row, "%s", "N/A");
+ display_column_print(max_col, row, "%s", "N/A");
+ display_column_print(avg_col, row, "%s", "N/A");
+ display_column_print(stddev_col, row, "%s", "N/A");
+ }
+
+ display_column_print(accuracy_limit_col, row, "%s", print_time_unit_usec(dst, &accuracy_limit));
+ display_column_print(lost_col, row, "%16"PRIu64"", stats_latency->lost_packets);
+ display_column_print(used_col, row, "%3u.%06u", used / AFTER_POINT, used % AFTER_POINT);
+}
+
+static void display_latency_draw_stats(struct screen_state *screen_state)
+{
+ const uint32_t n_latency = stats_get_n_latency();
+ struct stats_latency *stats_latency;
+
+ for (uint16_t i = 0; i < n_latency; ++i) {
+ if (screen_state->toggle == 0)
+ stats_latency = stats_latency_get(i);
+ else
+ stats_latency = stats_latency_tot_get(i);
+
+ display_stats_latency_entry(i, stats_latency);
+ }
+}
+
+static int display_latency_get_height(void)
+{
+ return stats_get_n_latency();
+}
+
+static struct display_screen display_screen_latency = {
+ .draw_frame = display_latency_draw_frame,
+ .draw_stats = display_latency_draw_stats,
+ .get_height = display_latency_get_height,
+ .title = "latency",
+};
+
+struct display_screen *display_latency(void)
+{
+ return &display_screen_latency;
+}