SF: add VSyncPredictor implementation
Adds an implementation of VSyncPredictor, an object that will
predict vsync events for multiple fixed-rate vsync systems.
The prediction is based on both the HWVsync signal and the
presentation fence timings for the VSyncDispatch object.
Bug: 140201379
Fixes: 140302888
Test: 13 new unit tests
Change-Id: I195902cc70561d028741d822e4001ad21ab391cf
diff --git a/services/surfaceflinger/Android.bp b/services/surfaceflinger/Android.bp
index 3d94918..4b71bd8 100644
--- a/services/surfaceflinger/Android.bp
+++ b/services/surfaceflinger/Android.bp
@@ -165,8 +165,9 @@
"Scheduler/PhaseOffsets.cpp",
"Scheduler/Scheduler.cpp",
"Scheduler/SchedulerUtils.cpp",
- "Scheduler/VSyncDispatchTimerQueue.cpp",
"Scheduler/Timer.cpp",
+ "Scheduler/VSyncDispatchTimerQueue.cpp",
+ "Scheduler/VSyncPredictor.cpp",
"Scheduler/VSyncModulator.cpp",
"StartPropertySetThread.cpp",
"SurfaceFlinger.cpp",
diff --git a/services/surfaceflinger/Scheduler/VSyncPredictor.cpp b/services/surfaceflinger/Scheduler/VSyncPredictor.cpp
new file mode 100644
index 0000000..643c5d2
--- /dev/null
+++ b/services/surfaceflinger/Scheduler/VSyncPredictor.cpp
@@ -0,0 +1,208 @@
+/*
+ * Copyright 2019 The Android Open Source Project
+ *
+ * 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.
+ */
+
+#define ATRACE_TAG ATRACE_TAG_GRAPHICS
+//#define LOG_NDEBUG 0
+#include "VSyncPredictor.h"
+#include <android-base/logging.h>
+#include <cutils/compiler.h>
+#include <utils/Log.h>
+#include <utils/Trace.h>
+#include <algorithm>
+#include <chrono>
+#include "SchedulerUtils.h"
+
+namespace android::scheduler {
+static auto constexpr kNeedsSamplesTag = "SamplesRequested";
+static auto constexpr kMaxPercent = 100u;
+
+VSyncPredictor::~VSyncPredictor() = default;
+
+VSyncPredictor::VSyncPredictor(nsecs_t idealPeriod, size_t historySize,
+ size_t minimumSamplesForPrediction, uint32_t outlierTolerancePercent)
+ : kHistorySize(historySize),
+ kMinimumSamplesForPrediction(minimumSamplesForPrediction),
+ kOutlierTolerancePercent(std::min(outlierTolerancePercent, kMaxPercent)),
+ mIdealPeriod(idealPeriod) {
+ mRateMap[mIdealPeriod] = {idealPeriod, 0};
+}
+
+inline size_t VSyncPredictor::next(int i) const {
+ return (i + 1) % timestamps.size();
+}
+
+bool VSyncPredictor::validate(nsecs_t timestamp) const {
+ if (lastTimestampIndex < 0 || timestamps.empty()) {
+ return true;
+ }
+
+ auto const aValidTimestamp = timestamps[lastTimestampIndex];
+ auto const percent = (timestamp - aValidTimestamp) % mIdealPeriod * kMaxPercent / mIdealPeriod;
+ return percent < kOutlierTolerancePercent || percent > (kMaxPercent - kOutlierTolerancePercent);
+}
+
+void VSyncPredictor::addVsyncTimestamp(nsecs_t timestamp) {
+ std::lock_guard<std::mutex> lk(mMutex);
+
+ if (!validate(timestamp)) {
+ ALOGW("timestamp was too far off the last known timestamp");
+ return;
+ }
+
+ if (timestamps.size() != kHistorySize) {
+ timestamps.push_back(timestamp);
+ lastTimestampIndex = next(lastTimestampIndex);
+ } else {
+ lastTimestampIndex = next(lastTimestampIndex);
+ timestamps[lastTimestampIndex] = timestamp;
+ }
+
+ if (timestamps.size() < kMinimumSamplesForPrediction) {
+ mRateMap[mIdealPeriod] = {mIdealPeriod, 0};
+ return;
+ }
+
+ // This is a 'simple linear regression' calculation of Y over X, with Y being the
+ // vsync timestamps, and X being the ordinal of vsync count.
+ // The calculated slope is the vsync period.
+ // Formula for reference:
+ // Sigma_i: means sum over all timestamps.
+ // mean(variable): statistical mean of variable.
+ // X: snapped ordinal of the timestamp
+ // Y: vsync timestamp
+ //
+ // Sigma_i( (X_i - mean(X)) * (Y_i - mean(Y) )
+ // slope = -------------------------------------------
+ // Sigma_i ( X_i - mean(X) ) ^ 2
+ //
+ // intercept = mean(Y) - slope * mean(X)
+ //
+ std::vector<nsecs_t> vsyncTS(timestamps.size());
+ std::vector<nsecs_t> ordinals(timestamps.size());
+
+ // normalizing to the oldest timestamp cuts down on error in calculating the intercept.
+ auto const oldest_ts = *std::min_element(timestamps.begin(), timestamps.end());
+ auto it = mRateMap.find(mIdealPeriod);
+ auto const currentPeriod = std::get<0>(it->second);
+ // TODO (b/144707443): its important that there's some precision in the mean of the ordinals
+ // for the intercept calculation, so scale the ordinals by 10 to continue
+ // fixed point calculation. Explore expanding
+ // scheduler::utils::calculate_mean to have a fixed point fractional part.
+ static constexpr int kScalingFactor = 10;
+
+ for (auto i = 0u; i < timestamps.size(); i++) {
+ vsyncTS[i] = timestamps[i] - oldest_ts;
+ ordinals[i] = ((vsyncTS[i] + (currentPeriod / 2)) / currentPeriod) * kScalingFactor;
+ }
+
+ auto meanTS = scheduler::calculate_mean(vsyncTS);
+ auto meanOrdinal = scheduler::calculate_mean(ordinals);
+ for (auto i = 0; i < vsyncTS.size(); i++) {
+ vsyncTS[i] -= meanTS;
+ ordinals[i] -= meanOrdinal;
+ }
+
+ auto top = 0ll;
+ auto bottom = 0ll;
+ for (auto i = 0; i < vsyncTS.size(); i++) {
+ top += vsyncTS[i] * ordinals[i];
+ bottom += ordinals[i] * ordinals[i];
+ }
+
+ if (CC_UNLIKELY(bottom == 0)) {
+ it->second = {mIdealPeriod, 0};
+ return;
+ }
+
+ nsecs_t const anticipatedPeriod = top / bottom * kScalingFactor;
+ nsecs_t const intercept = meanTS - (anticipatedPeriod * meanOrdinal / kScalingFactor);
+
+ it->second = {anticipatedPeriod, intercept};
+
+ ALOGV("model update ts: %" PRId64 " slope: %" PRId64 " intercept: %" PRId64, timestamp,
+ anticipatedPeriod, intercept);
+}
+
+nsecs_t VSyncPredictor::nextAnticipatedVSyncTimeFrom(nsecs_t timePoint) const {
+ std::lock_guard<std::mutex> lk(mMutex);
+
+ auto const [slope, intercept] = getVSyncPredictionModel(lk);
+
+ if (timestamps.empty()) {
+ auto const knownTimestamp = mKnownTimestamp ? *mKnownTimestamp : timePoint;
+ auto const numPeriodsOut = ((timePoint - knownTimestamp) / mIdealPeriod) + 1;
+ return knownTimestamp + numPeriodsOut * mIdealPeriod;
+ }
+
+ auto const oldest = *std::min_element(timestamps.begin(), timestamps.end());
+ auto const ordinalRequest = (timePoint - oldest + slope) / slope;
+ auto const prediction = (ordinalRequest * slope) + intercept + oldest;
+
+ ALOGV("prediction made from: %" PRId64 " prediction: %" PRId64 " (+%" PRId64 ") slope: %" PRId64
+ " intercept: %" PRId64,
+ timePoint, prediction, prediction - timePoint, slope, intercept);
+ return prediction;
+}
+
+std::tuple<nsecs_t, nsecs_t> VSyncPredictor::getVSyncPredictionModel() const {
+ std::lock_guard<std::mutex> lk(mMutex);
+ return VSyncPredictor::getVSyncPredictionModel(lk);
+}
+
+std::tuple<nsecs_t, nsecs_t> VSyncPredictor::getVSyncPredictionModel(
+ std::lock_guard<std::mutex> const&) const {
+ return mRateMap.find(mIdealPeriod)->second;
+}
+
+void VSyncPredictor::setPeriod(nsecs_t period) {
+ ATRACE_CALL();
+
+ std::lock_guard<std::mutex> lk(mMutex);
+ static constexpr size_t kSizeLimit = 30;
+ if (CC_UNLIKELY(mRateMap.size() == kSizeLimit)) {
+ mRateMap.erase(mRateMap.begin());
+ }
+
+ mIdealPeriod = period;
+ if (mRateMap.find(period) == mRateMap.end()) {
+ mRateMap[mIdealPeriod] = {period, 0};
+ }
+
+ if (!timestamps.empty()) {
+ mKnownTimestamp = *std::max_element(timestamps.begin(), timestamps.end());
+ timestamps.clear();
+ lastTimestampIndex = 0;
+ }
+}
+
+bool VSyncPredictor::needsMoreSamples(nsecs_t now) const {
+ using namespace std::literals::chrono_literals;
+ std::lock_guard<std::mutex> lk(mMutex);
+ bool needsMoreSamples = true;
+ if (timestamps.size() >= kMinimumSamplesForPrediction) {
+ nsecs_t constexpr aLongTime =
+ std::chrono::duration_cast<std::chrono::nanoseconds>(500ms).count();
+ if (!(lastTimestampIndex < 0 || timestamps.empty())) {
+ auto const lastTimestamp = timestamps[lastTimestampIndex];
+ needsMoreSamples = !((lastTimestamp + aLongTime) > now);
+ }
+ }
+
+ ATRACE_INT(kNeedsSamplesTag, needsMoreSamples);
+ return needsMoreSamples;
+}
+
+} // namespace android::scheduler
diff --git a/services/surfaceflinger/Scheduler/VSyncPredictor.h b/services/surfaceflinger/Scheduler/VSyncPredictor.h
new file mode 100644
index 0000000..1590f49
--- /dev/null
+++ b/services/surfaceflinger/Scheduler/VSyncPredictor.h
@@ -0,0 +1,83 @@
+/*
+ * Copyright 2019 The Android Open Source Project
+ *
+ * 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.
+ */
+
+#pragma once
+
+#include <android-base/thread_annotations.h>
+#include <mutex>
+#include <unordered_map>
+#include <vector>
+#include "VSyncTracker.h"
+
+namespace android::scheduler {
+
+class VSyncPredictor : public VSyncTracker {
+public:
+ /*
+ * \param [in] idealPeriod The initial ideal period to use.
+ * \param [in] historySize The internal amount of entries to store in the model.
+ * \param [in] minimumSamplesForPrediction The minimum number of samples to collect before
+ * predicting. \param [in] outlierTolerancePercent a number 0 to 100 that will be used to filter
+ * samples that fall outlierTolerancePercent from an anticipated vsync event.
+ */
+ VSyncPredictor(nsecs_t idealPeriod, size_t historySize, size_t minimumSamplesForPrediction,
+ uint32_t outlierTolerancePercent);
+ ~VSyncPredictor();
+
+ void addVsyncTimestamp(nsecs_t timestamp) final;
+ nsecs_t nextAnticipatedVSyncTimeFrom(nsecs_t timePoint) const final;
+
+ /*
+ * Inform the model that the period is anticipated to change to a new value.
+ * model will use the period parameter to predict vsync events until enough
+ * timestamps with the new period have been collected.
+ *
+ * \param [in] period The new period that should be used.
+ */
+ void setPeriod(nsecs_t period);
+
+ /* Query if the model is in need of more samples to make a prediction at timePoint.
+ * \param [in] timePoint The timePoint to inquire of.
+ * \return True, if model would benefit from more samples, False if not.
+ */
+ bool needsMoreSamples(nsecs_t timePoint) const;
+
+ std::tuple<nsecs_t /* slope */, nsecs_t /* intercept */> getVSyncPredictionModel() const;
+
+private:
+ VSyncPredictor(VSyncPredictor const&) = delete;
+ VSyncPredictor& operator=(VSyncPredictor const&) = delete;
+
+ size_t const kHistorySize;
+ size_t const kMinimumSamplesForPrediction;
+ size_t const kOutlierTolerancePercent;
+
+ std::mutex mutable mMutex;
+ size_t next(int i) const REQUIRES(mMutex);
+ bool validate(nsecs_t timestamp) const REQUIRES(mMutex);
+ std::tuple<nsecs_t, nsecs_t> getVSyncPredictionModel(std::lock_guard<std::mutex> const&) const
+ REQUIRES(mMutex);
+
+ nsecs_t mIdealPeriod GUARDED_BY(mMutex);
+ std::optional<nsecs_t> mKnownTimestamp GUARDED_BY(mMutex);
+
+ std::unordered_map<nsecs_t, std::tuple<nsecs_t, nsecs_t>> mutable mRateMap GUARDED_BY(mMutex);
+
+ int lastTimestampIndex GUARDED_BY(mMutex) = 0;
+ std::vector<nsecs_t> timestamps GUARDED_BY(mMutex);
+};
+
+} // namespace android::scheduler
diff --git a/services/surfaceflinger/tests/unittests/Android.bp b/services/surfaceflinger/tests/unittests/Android.bp
index c6c3d29..0c4a752 100644
--- a/services/surfaceflinger/tests/unittests/Android.bp
+++ b/services/surfaceflinger/tests/unittests/Android.bp
@@ -56,6 +56,7 @@
"StrongTypingTest.cpp",
"VSyncDispatchTimerQueueTest.cpp",
"VSyncDispatchRealtimeTest.cpp",
+ "VSyncPredictorTest.cpp",
"mock/DisplayHardware/MockComposer.cpp",
"mock/DisplayHardware/MockDisplay.cpp",
"mock/DisplayHardware/MockPowerAdvisor.cpp",
diff --git a/services/surfaceflinger/tests/unittests/VSyncPredictorTest.cpp b/services/surfaceflinger/tests/unittests/VSyncPredictorTest.cpp
new file mode 100644
index 0000000..d0c8090
--- /dev/null
+++ b/services/surfaceflinger/tests/unittests/VSyncPredictorTest.cpp
@@ -0,0 +1,322 @@
+/*
+ * Copyright 2019 The Android Open Source Project
+ *
+ * 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.
+ */
+
+#undef LOG_TAG
+#define LOG_TAG "LibSurfaceFlingerUnittests"
+#define LOG_NDEBUG 0
+
+#include "Scheduler/VSyncPredictor.h"
+
+#include <gmock/gmock.h>
+#include <gtest/gtest.h>
+#include <algorithm>
+#include <chrono>
+#include <utility>
+
+using namespace testing;
+using namespace std::literals;
+
+namespace android::scheduler {
+
+MATCHER_P2(IsCloseTo, value, tolerance, "is within tolerance") {
+ return arg <= value + tolerance && value >= value - tolerance;
+}
+
+std::vector<nsecs_t> generateVsyncTimestamps(size_t count, nsecs_t period, nsecs_t bias) {
+ std::vector<nsecs_t> vsyncs(count);
+ std::generate(vsyncs.begin(), vsyncs.end(),
+ [&, n = 0]() mutable { return n++ * period + bias; });
+ return vsyncs;
+}
+
+struct VSyncPredictorTest : testing::Test {
+ nsecs_t mNow = 0;
+ nsecs_t mPeriod = 1000;
+ static constexpr size_t kHistorySize = 10;
+ static constexpr size_t kMinimumSamplesForPrediction = 6;
+ static constexpr size_t kOutlierTolerancePercent = 25;
+ static constexpr nsecs_t mMaxRoundingError = 100;
+
+ VSyncPredictor tracker{mPeriod, kHistorySize, kMinimumSamplesForPrediction,
+ kOutlierTolerancePercent};
+};
+
+TEST_F(VSyncPredictorTest, reportsAnticipatedPeriod) {
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+
+ EXPECT_THAT(slope, Eq(mPeriod));
+ EXPECT_THAT(intercept, Eq(0));
+
+ auto const changedPeriod = 2000;
+ tracker.setPeriod(changedPeriod);
+ std::tie(slope, intercept) = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, Eq(changedPeriod));
+ EXPECT_THAT(intercept, Eq(0));
+}
+
+TEST_F(VSyncPredictorTest, reportsSamplesNeededWhenHasNoDataPoints) {
+ for (auto i = 0u; i < kMinimumSamplesForPrediction; i++) {
+ EXPECT_TRUE(tracker.needsMoreSamples(mNow += mPeriod));
+ tracker.addVsyncTimestamp(mNow);
+ }
+ EXPECT_FALSE(tracker.needsMoreSamples(mNow));
+}
+
+TEST_F(VSyncPredictorTest, reportsSamplesNeededAfterExplicitRateChange) {
+ for (auto i = 0u; i < kMinimumSamplesForPrediction; i++) {
+ tracker.addVsyncTimestamp(mNow += mPeriod);
+ }
+ EXPECT_FALSE(tracker.needsMoreSamples(mNow));
+
+ auto const changedPeriod = mPeriod * 2;
+ tracker.setPeriod(changedPeriod);
+ EXPECT_TRUE(tracker.needsMoreSamples(mNow));
+
+ for (auto i = 0u; i < kMinimumSamplesForPrediction; i++) {
+ EXPECT_TRUE(tracker.needsMoreSamples(mNow += changedPeriod));
+ tracker.addVsyncTimestamp(mNow);
+ }
+ EXPECT_FALSE(tracker.needsMoreSamples(mNow));
+}
+
+TEST_F(VSyncPredictorTest, transitionsToModelledPointsAfterSynthetic) {
+ auto last = mNow;
+ auto const bias = 10;
+ for (auto i = 0u; i < kMinimumSamplesForPrediction; i++) {
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow), Eq(last + mPeriod));
+ mNow += mPeriod - bias;
+ last = mNow;
+ tracker.addVsyncTimestamp(mNow);
+ mNow += bias;
+ }
+
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow), Eq(mNow + mPeriod - bias));
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow + 100), Eq(mNow + mPeriod - bias));
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow + 990), Eq(mNow + 2 * mPeriod - bias));
+}
+
+TEST_F(VSyncPredictorTest, uponNotifiedOfInaccuracyUsesSynthetic) {
+ auto const slightlyLessPeriod = mPeriod - 10;
+ auto const changedPeriod = mPeriod - 1;
+ for (auto i = 0u; i < kMinimumSamplesForPrediction; i++) {
+ tracker.addVsyncTimestamp(mNow += slightlyLessPeriod);
+ }
+
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow), Eq(mNow + slightlyLessPeriod));
+ tracker.setPeriod(changedPeriod);
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow), Eq(mNow + changedPeriod));
+}
+
+TEST_F(VSyncPredictorTest, adaptsToFenceTimelines_60hzHighVariance) {
+ // these are precomputed simulated 16.6s vsyncs with uniform distribution +/- 1.6ms error
+ std::vector<nsecs_t> const simulatedVsyncs{
+ 15492949, 32325658, 49534984, 67496129, 84652891,
+ 100332564, 117737004, 132125931, 149291099, 165199602,
+ };
+ auto constexpr idealPeriod = 16600000;
+ auto constexpr expectedPeriod = 16639242;
+ auto constexpr expectedIntercept = 1049341;
+
+ tracker.setPeriod(idealPeriod);
+ for (auto const& timestamp : simulatedVsyncs) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, IsCloseTo(expectedPeriod, mMaxRoundingError));
+ EXPECT_THAT(intercept, IsCloseTo(expectedIntercept, mMaxRoundingError));
+}
+
+TEST_F(VSyncPredictorTest, adaptsToFenceTimelines_90hzLowVariance) {
+ // these are precomputed simulated 11.1 vsyncs with uniform distribution +/- 1ms error
+ std::vector<nsecs_t> const simulatedVsyncs{
+ 11167047, 22603464, 32538479, 44938134, 56321268,
+ 66730346, 78062637, 88171429, 99707843, 111397621,
+ };
+ auto idealPeriod = 11110000;
+ auto expectedPeriod = 11089413;
+ auto expectedIntercept = 94421;
+
+ tracker.setPeriod(idealPeriod);
+ for (auto const& timestamp : simulatedVsyncs) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, IsCloseTo(expectedPeriod, mMaxRoundingError));
+ EXPECT_THAT(intercept, IsCloseTo(expectedIntercept, mMaxRoundingError));
+}
+
+TEST_F(VSyncPredictorTest, adaptsToFenceTimelinesDiscontinuous_22hzLowVariance) {
+ // these are 11.1s vsyncs with low variance, randomly computed, between -1 and 1ms
+ std::vector<nsecs_t> const simulatedVsyncs{
+ 45259463, // 0
+ 91511026, // 1
+ 136307650, // 2
+ 1864501714, // 40
+ 1908641034, // 41
+ 1955278544, // 42
+ 4590180096, // 100
+ 4681594994, // 102
+ 5499224734, // 120
+ 5591378272, // 122
+ };
+ auto idealPeriod = 45454545;
+ auto expectedPeriod = 45450152;
+ auto expectedIntercept = 469647;
+
+ tracker.setPeriod(idealPeriod);
+ for (auto const& timestamp : simulatedVsyncs) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, IsCloseTo(expectedPeriod, mMaxRoundingError));
+ EXPECT_THAT(intercept, IsCloseTo(expectedIntercept, mMaxRoundingError));
+}
+
+TEST_F(VSyncPredictorTest, againstOutliersDiscontinuous_500hzLowVariance) {
+ std::vector<nsecs_t> const simulatedVsyncs{
+ 1992548, // 0
+ 4078038, // 1
+ 6165794, // 2
+ 7958171, // 3
+ 10193537, // 4
+ 2401840200, // 1200
+ 2403000000, // an outlier that should be excluded (1201 and a half)
+ 2405803629, // 1202
+ 2408028599, // 1203
+ 2410121051, // 1204
+ };
+ auto idealPeriod = 2000000;
+ auto expectedPeriod = 1999892;
+ auto expectedIntercept = 175409;
+
+ tracker.setPeriod(idealPeriod);
+ for (auto const& timestamp : simulatedVsyncs) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, IsCloseTo(expectedPeriod, mMaxRoundingError));
+ EXPECT_THAT(intercept, IsCloseTo(expectedIntercept, mMaxRoundingError));
+}
+
+TEST_F(VSyncPredictorTest, handlesVsyncChange) {
+ auto const fastPeriod = 100;
+ auto const fastTimeBase = 100;
+ auto const slowPeriod = 400;
+ auto const slowTimeBase = 800;
+ auto const simulatedVsyncsFast =
+ generateVsyncTimestamps(kMinimumSamplesForPrediction, fastPeriod, fastTimeBase);
+ auto const simulatedVsyncsSlow =
+ generateVsyncTimestamps(kMinimumSamplesForPrediction, slowPeriod, slowTimeBase);
+
+ tracker.setPeriod(fastPeriod);
+ for (auto const& timestamp : simulatedVsyncsFast) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+
+ auto const mMaxRoundingError = 100;
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, IsCloseTo(fastPeriod, mMaxRoundingError));
+ EXPECT_THAT(intercept, IsCloseTo(0, mMaxRoundingError));
+
+ tracker.setPeriod(slowPeriod);
+ for (auto const& timestamp : simulatedVsyncsSlow) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ std::tie(slope, intercept) = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, IsCloseTo(slowPeriod, mMaxRoundingError));
+ EXPECT_THAT(intercept, IsCloseTo(0, mMaxRoundingError));
+}
+
+TEST_F(VSyncPredictorTest, willBeAccurateUsingPriorResultsForRate) {
+ auto const fastPeriod = 101000;
+ auto const fastTimeBase = fastPeriod - 500;
+ auto const fastPeriod2 = 99000;
+
+ auto const slowPeriod = 400000;
+ auto const slowTimeBase = 800000 - 201;
+ auto const simulatedVsyncsFast =
+ generateVsyncTimestamps(kMinimumSamplesForPrediction, fastPeriod, fastTimeBase);
+ auto const simulatedVsyncsSlow =
+ generateVsyncTimestamps(kMinimumSamplesForPrediction, slowPeriod, slowTimeBase);
+ auto const simulatedVsyncsFast2 =
+ generateVsyncTimestamps(kMinimumSamplesForPrediction, fastPeriod2, fastTimeBase);
+
+ auto idealPeriod = 100000;
+ tracker.setPeriod(idealPeriod);
+ for (auto const& timestamp : simulatedVsyncsFast) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ auto [slope, intercept] = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, Eq(fastPeriod));
+ EXPECT_THAT(intercept, Eq(0));
+
+ tracker.setPeriod(slowPeriod);
+ for (auto const& timestamp : simulatedVsyncsSlow) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+
+ // we had a model for 100ns mPeriod before, use that until the new samples are
+ // sufficiently built up
+ tracker.setPeriod(idealPeriod);
+ std::tie(slope, intercept) = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, Eq(fastPeriod));
+ EXPECT_THAT(intercept, Eq(0));
+
+ for (auto const& timestamp : simulatedVsyncsFast2) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ std::tie(slope, intercept) = tracker.getVSyncPredictionModel();
+ EXPECT_THAT(slope, Eq(fastPeriod2));
+ EXPECT_THAT(intercept, Eq(0));
+}
+
+TEST_F(VSyncPredictorTest, willBecomeInaccurateAfterA_longTimeWithNoSamples) {
+ auto const simulatedVsyncs = generateVsyncTimestamps(kMinimumSamplesForPrediction, mPeriod, 0);
+
+ for (auto const& timestamp : simulatedVsyncs) {
+ tracker.addVsyncTimestamp(timestamp);
+ }
+ auto const mNow = *simulatedVsyncs.rbegin();
+ EXPECT_FALSE(tracker.needsMoreSamples(mNow));
+
+ // TODO: would be better to decay this as a result of the variance of the samples
+ static auto constexpr aLongTimeOut = 1000000000;
+ EXPECT_TRUE(tracker.needsMoreSamples(mNow + aLongTimeOut));
+}
+
+TEST_F(VSyncPredictorTest, idealModelPredictionsBeforeRegressionModelIsBuilt) {
+ auto const simulatedVsyncs =
+ generateVsyncTimestamps(kMinimumSamplesForPrediction + 1, mPeriod, 0);
+ nsecs_t const mNow = 0;
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(mNow), Eq(mPeriod));
+
+ nsecs_t const aBitOfTime = 422;
+
+ for (auto i = 0; i < kMinimumSamplesForPrediction; i++) {
+ tracker.addVsyncTimestamp(simulatedVsyncs[i]);
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(simulatedVsyncs[i] + aBitOfTime),
+ Eq(mPeriod + simulatedVsyncs[i]));
+ }
+
+ for (auto i = kMinimumSamplesForPrediction; i < simulatedVsyncs.size(); i++) {
+ tracker.addVsyncTimestamp(simulatedVsyncs[i]);
+ EXPECT_THAT(tracker.nextAnticipatedVSyncTimeFrom(simulatedVsyncs[i] + aBitOfTime),
+ Eq(mPeriod + simulatedVsyncs[i]));
+ }
+}
+
+} // namespace android::scheduler