From: wn.jang <wn.jang@samsung.com>
Date: Thu, 13 Apr 2023 01:30:17 +0000 (+0900)
Subject: Add logs to measure xRT for synthesizing from tts engine
X-Git-Tag: accepted/tizen/unified/20230425.130131~5^2
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=refs%2Fchanges%2F84%2F291284%2F5;p=platform%2Fcore%2Fuifw%2Ftts.git

Add logs to measure xRT for synthesizing from tts engine

Change-Id: I9a43b12403df907db02d6309a5cdd4c5957a3be8
---

diff --git a/server/ttsd_server.c b/server/ttsd_server.c
index 3b7db8d2..3ddcfd90 100644
--- a/server/ttsd_server.c
+++ b/server/ttsd_server.c
@@ -32,11 +32,12 @@
 #define TIME_DIFF(start, end) \
 	((uint64_t)((end).tv_sec - (start).tv_sec) * 1000000 + (((end).tv_nsec - (start).tv_nsec) / 1000)) / 1000
 
-static struct timespec g_request_playing, g_start_playing;
+static struct timespec g_request_playing, g_start_playing, g_finish_playing;
 static double g_avg_latency;
 static double g_min_latency;
 static double g_max_latency;
 static int g_file_num;
+static int g_total_data_size;
 
 typedef struct {
 	unsigned int uid;
@@ -210,6 +211,23 @@ int ttsd_send_error(ttse_error_e error, const char* msg)
 	return 0;
 }
 
+static int ttsd_convert_audio_type_to_bytes(ttse_audio_type_e audio_type)
+{
+	int ret = 0;
+	switch (audio_type) {
+	case TTSE_AUDIO_TYPE_RAW_S16:
+		ret = 2; // 16bit
+		break;
+	case TTSE_AUDIO_TYPE_RAW_U8:
+		ret = 1; // 8bit
+		break;
+	default:
+		ret = 2; // Default 16bit
+		break;
+	}
+	return ret;
+}
+
 int ttsd_send_result(ttse_result_event_e event, const void* data, unsigned int data_size, ttse_audio_type_e audio_type, int rate, void* user_data)
 {
 	SLOG(LOG_DEBUG, tts_tag(), "@@@ SEND SYNTHESIS RESULT START");
@@ -266,13 +284,26 @@ int ttsd_send_result(ttse_result_event_e event, const void* data, unsigned int d
 		g_min_latency = (d_latency < g_min_latency) ? d_latency : g_min_latency;
 		g_max_latency = (d_latency > g_max_latency) ? d_latency : g_max_latency;
 		g_file_num++;
+		g_total_data_size = 0;
+		g_total_data_size += data_size;
 		SLOG(LOG_INFO, tts_tag(), "[Server] File num(%d), Avg Latency(%lf), Min(%lf), Max(%lf)", g_file_num, (g_avg_latency / (double)g_file_num), g_min_latency, g_max_latency);
 	} else if (TTSE_RESULT_EVENT_FINISH == event) {
 		SLOG(LOG_INFO, tts_tag(), "[SERVER] Event : TTSE_RESULT_EVENT_FINISH");
 		SECURE_SLOG(LOG_DEBUG, tts_tag(), "[SERVER] Result Info : uid(%u), utt(%d), data(%p), data size(%d) audiotype(%d) rate(%d)",
 			uid, uttid, data, data_size, audio_type, rate);
+
+		g_total_data_size += data_size;
+		int audio_type_bytes = ttsd_convert_audio_type_to_bytes(audio_type);
+		/* Calculate xRT */
+		clock_gettime(CLOCK_MONOTONIC_RAW, &g_finish_playing);
+		long long int time_processing = (uint64_t)TIME_DIFF(g_request_playing, g_finish_playing);
+		long long int time_speech = (long long int)(1000 * g_total_data_size * sizeof(short) / (rate * audio_type_bytes));
+		double xRT = (double)(time_processing) / (double)time_speech;
+		SLOG(LOG_INFO, tts_tag(), "[SERVER] time_processing : %lld, time_speech : %lld, total data size: %d", time_processing, time_speech, g_total_data_size);
+		SLOG(LOG_INFO, tts_tag(), "[SERVER] xRT : %lf", xRT);
 	} else {
 		/*if (TTSE_RESULT_EVENT_CONTINUE == event)  SLOG(LOG_DEBUG, tts_tag(), "[SERVER] Event : TTSE_RESULT_EVENT_CONTINUE");*/
+		g_total_data_size += data_size;
 	}
 
 	/* add wav data */