JobServer: print completion percentage and ETA

If the JobServer is provided a total number of experiments by the
campaign, it now prints a completion percentage and an estimated
remaining runtime along the usual progress reports.

Change-Id: Ibd781ba8bff9af3a85683bbd29728216e316da57

src/core/cpn/CampaignManager.cc

@@ -42,5 +42,15 @@ void CampaignManager::noMoreParameters()
 	m_jobserver->setNoMoreExperiments();
 }
 
+void CampaignManager::setTotalCount(uint64_t count)
+{
+	m_jobserver->setTotalCount(count);
+}
+
+void CampaignManager::skipJobs(uint64_t count)
+{
+	m_jobserver->skipJobs(count);
+}
+
 void CampaignManager::done() { m_jobserver->done(); }
 } // end-of-namespace: fail

src/core/cpn/CampaignManager.hpp

@@ -64,6 +64,15 @@ public:
 	 * Signal, that there will not come any further parameter sets.
 	 */
 	void noMoreParameters();
+	/**
+	 * Can optionally be used to tell the JobServer how many jobs to expect in
+	 * total.  This count is used for progress reporting.  Make sure you also
+	 * call skipJobs() if some of these early-on announced jobs will not be
+	 * sent after all (e.g. because the campaign already found results for them
+	 * in the database).
+	 */
+	void setTotalCount(uint64_t count);
+	void skipJobs(uint64_t count);
 	 /**
 	 * User campaign has finished.
 	 */

src/core/cpn/DatabaseCampaign.cc

@@ -215,6 +215,7 @@ bool DatabaseCampaign::run_variant(Database::Variant variant) {
 
 	log_send << "Found " << experiment_count << " jobs in database. ("
 			 << variant.variant << "/" << variant.benchmark << ")" << std::endl;
+	campaignmanager.setTotalCount(experiment_count);
 
 	// abstraction of injection point:
 	// must not be initialized in loop, because hop chain calculator would lose
@@ -229,6 +230,7 @@ bool DatabaseCampaign::run_variant(Database::Variant variant) {
 		unsigned pilot_id        = strtoul(row[0], NULL, 10);
 		if (existing_results_for_pilot(pilot_id) == expected_results) {
 			skipped_pilots++;
+			campaignmanager.skipJobs(1);
 			continue;
 		}
 

src/core/cpn/JobServer.cc

@@ -276,9 +276,9 @@ void CommThread::print_progress(const enum ProgressType type,
 				const uint32_t w_id, const uint32_t count)
 {
 	using namespace std::chrono;
-	static system_clock::time_point last;
 	const auto now = system_clock::now();
 	const auto delay = milliseconds{500};
+	static system_clock::time_point last = system_clock::now() - delay;
 
 	if (last + delay > now) {
 		return;
@@ -297,6 +297,19 @@ void CommThread::print_progress(const enum ProgressType type,
 		<< std::setw(6) << m_js.m_runningJobs.Size() << " run/"
 		<< std::setw(6) << donecount_cur << " tot/ ("
 		<< std::setw(6) << std::setprecision(1) << std::fixed << rate << "/s)  ";
+	if (m_js.m_TotalCount) {
+		float percentage = (float) donecount_cur / *m_js.m_TotalCount * 100.0;
+		std::cout << std::setw(4) << std::setprecision(1) << std::fixed << percentage << "%";
+		if (rate > 0) {
+			float ETA_s = std::max(.0f, (*m_js.m_TotalCount - donecount_cur) / rate);
+			std::cout << " (ETA " << std::dec
+				<< std::setw(2) << std::setfill('0') <<  ((int64_t)ETA_s / (60*60)) << ':'
+				<< std::setw(2) << std::setfill('0') << (((int64_t)ETA_s / 60) % 60) << ':'
+				<< std::setw(2) << std::setfill('0') <<  ((int64_t)ETA_s % 60)
+				<< std::setfill(' ') << ')';
+		}
+		std::cout << "   ";
+	}
 	const char *sep;
 	if (type == ProgressType::Send) {
 		sep = " >";

src/core/cpn/JobServer.hpp

@@ -16,6 +16,7 @@
 #ifndef __puma
 #include <boost/thread.hpp>
 #endif
+#include <boost/optional.hpp>
 
 namespace fail {
 
@@ -52,6 +53,7 @@ private:
 	uint64_t m_runid;
 
 	volatile uint64_t m_DoneCount = 0; //! the number of finished jobs
+	boost::optional<uint64_t> m_TotalCount; //! the total number of jobs to be expected
 #ifdef SERVER_PERFORMANCE_MEASURE
 #ifndef __puma
 	boost::thread* m_measureThread; //! the performance measurement thread
@@ -114,6 +116,15 @@ public:
 	 * distribution.
 	 */
 	void setNoMoreExperiments();
+	/**
+	 * Can optionally be used to tell the JobServer how many jobs to expect in
+	 * total.  This count is used for progress reporting.  Make sure you also
+	 * call skipJobs() if some of these early-on announced jobs will not be
+	 * sent after all (e.g. because the campaign already found results for them
+	 * in the database).
+	 */
+	void setTotalCount(uint64_t count) { m_TotalCount = count; }
+	void skipJobs(uint64_t count) { ++m_DoneCount; /* FIXME assume atomic */ }
 	/**
 	 * Checks whether there are no more experiment parameter sets.
 	 * @return \c true if no more parameter sets available, \c false otherwise