SlideShare a Scribd company logo
1 of 62
Download to read offline
RAVEN: Perception-aware Optimization
of Power Consumption for Mobile Games
Published in MobiCom’17
Chanyou Hwang et al.
1
Gabin An, Jinhan Kim
2018.10.31
Introduction
2
Background
● A mobile GPU constitutes a major portion of power consumption on the
devices.
● Visually appealing mobile games => High energy cost
GPU Battery
3
Motivation 1
● No matter the states, the game apps
maintain high frame rates.
4
In-Game Lobby
Crossy Road
(길건너친구들)
Motivation 2
● Large portions of adjacent frames show strong similarity.
5
Goal
● To reduce power consumption of mobile game playing without
degrading user experiences.
6
Goal
● To reduce power consumption of mobile game playing without
degrading user experiences.
7
Recap.
Challenges
1. The perceptual similarity should be measured efficiently.
2. The degradation of user experiences should be minimal.
3. The system should be transparent to the apps and avoid low-level
system optimization (since mobile gaming apps and GPU drivers are closed source).
8
RAVEN
RAVEN is a DC comics character
whose ability is ‘sense and alter.’
Methodology
9
RAVEN: System Overview
10
RAVEN: System Overview
11
RAVEN: System Overview
12
abstraction
RAVEN: System Overview
● Three major components for scaling game-frame rendering
○ F-Tracker (FrameDiff Tracker)
○ R-Regulator (Rate Regulator)
○ R-Injector (Rate Injector)
● Small virtual display
○ Low resolution to reduce overhead
13
RAVEN: Virtual Display (1)
Mobile gameplay (1920 x 1080)
Mobile gameplay (80 x 45)
Virtual Display
14
piggyback
RAVEN: F-Tracker (2)
Virtual Display
F-Tracker
Down-scaled pixels
Current Frame
(fN
)
Recently Rendered Frame
(fN-k
)
Calculate
Y-Diff 15
Determine how many
frames will be skipped
(No-Skipping | 1-Skipping | 3-Skipping)
RAVEN: R-Regulator (3)
R-Regulator
predict the similarity
scores with future frames
Y-Diff(fN-k
, fN
)
16
EPS: Estimated Perceptual Similarity
RAVEN: R-Injector (4)
R-Injector
Inject a delay into the
rendering loop of game app
The # of frames to be skipped
eglSwapBuffer()
Frames with delay
SurfaceFlinger
17
How can we calculate similarity between frames?
● SSIM [1]
○ Existing method to measure user's perception of visual changes
■ SSIM > 0.97
● A strong level of similarity between two images in mobile games [2]
■ Computationally expensive 😭
● YUV Color Space
○ Color encoding system like RGB
○ Leverage Y (luminance) values to approximate SSIM efficiently
[1] Wang, Zhou, et al. "Image quality assessment: from error visibility to structural similarity." IEEE transactions on image processing 13.4 (2004): 600-612.
[2] Cuervo, E., Wolman, A., Cox, L. P., Lebeck, K., Razeen, A., Saroiu, S., & Musuvathi, M. (2015, May). Kahawai: High-quality mobile gaming using gpu offload. In
Proceedings of the 13th Annual International Conference on Mobile Systems, Applications, and Services (pp. 121-135). ACM.
18
Y-Diff
● Y−Diff is the difference in Y values of two images in the YUV color space.
○ luminance (Y): brightness
○ chrominance (UV): color
● Computation cost: O(N)
○ N = the number of pixels
19
UV space
(Y′ value = 0.5)
Would it be reasonable to use Y-Diff?
the relation between Y−Diff and SSIM
20
Would it be reasonable to use Y-Diff?
the relation between Y−Diff and SSIM
21
Pearson’s correlation coefficient of -0.926
Would it be reasonable to use Y-Diff?
the relation between Y−Diff and SSIM
22
Pearson’s correlation coefficient of -0.926
SSIM and Y−Diff values are strongly correlated
and show a linear relationship.
Can we predict similarity of future frames?
● X-axis: Similarity between previous-current frames
● Y-axis: Similarity between current-future frames
23
Can we predict similarity of future frames?
● X-axis: Similarity between previous-current frames
● Y-axis: Similarity between current-future frames
24
Pearson’s correlation
coefficient of 0.93
Can we predict similarity of future frames?
● X-axis: Similarity between previous-current frames
● Y-axis: Similarity between current-future frames
25
Pearson’s correlation
coefficient of 0.93
The similarity with a future frame can be predicted
from the similarity with a previous one.
Perceptual Similarity Prediction
Y-Diff between current
frames and recently
rendered frame
26
● EPS estimates SSIM using Y-Diff (regression)
Perceptual Similarity Prediction
Constant
Set depending on app
and index k of previous frame (1, 2, 4)
27
● EPS estimates SSIM using Y-Diff (regression)
Constant
Perceptual Similarity Prediction
28
● EPS estimates SSIM using Y-Diff (regression)
Average of recent Y-Diff
scores over window size w
(reflecting past history)
w = 1, 2, 5, and 10
Best
SSIM vs. EPS
29
Candy Crush Soda Saga Archery Master 3D
SSIM vs. EPS
30
Candy Crush Soda Saga Archery Master 3D
The predictions were more accurate within
the potential range of threshold values (> 0.97)
Determining the Scaling Factors (1)
31
1-Skipping
3-Skipping
Determining the Scaling Factors (2)
● False transition
○ Skipping of a frame which should have not been skipped
○ EPS > τ1
or τ3
, but SSIM < 0.97
● Increasing the τ1
and τ3
reduces the false transition ratio.
● For example,
○ when τ3
=0.97, the ratio of false transition is 27% for 3-Skipping
○ when τ3
=0.9945, the ratio of false transition is 2.7% for 3-Skipping
32
below the strong level of similaritya skip occur
Cloning the Primary Display: Virtual Display
● A virtual display could be allocated and maintained by SurfaceFlinger of
Android.
● The overhead of a virtual display is primarily proportional to the
resolution of the display.
33
Cloning the Primary Display: Why
a 0.9989 Pearson’s correlation
34
Cloning the Primary Display: Why
35
Cloning the Primary Display: Why
36
The Y−Diff results with 80×45 and 1920×1080 are nearly
identical with the selected game-play data, scoring a 0.9989
Pearson’s correlation coefficient
Recap.
37
RAVEN: Virtual Display (1)
Mobile gameplay (1920 x 1080)
Mobile gameplay (80 x 45)
Virtual Display
38
piggyback
RAVEN: F-Tracker (2)
Virtual Display
F-Tracker
Down-scaled pixels
Current Frame
(fN
)
Recently Rendered Frame
(fN-k
)
Calculate
Y-Diff 39
Determine how many
frames will be skipped
(No-Skipping | 1-Skipping | 3-Skipping)
RAVEN: R-Regulator (3)
R-Regulator
predict the similarity
scores with future frames
Y-Diff
40
EPS: Estimated Perceptual Similarity
RAVEN: R-Injector (4)
R-Injector
Inject a delay into the
rendering loop of game app
The # of frames to be skipped
eglSwapBuffer()
Frames with delay
SurfaceFlinger
41
Experiment
42
- Two different settings for the evaluation of RAVEN
- PAS : τ1
= 0.9975, τ3
= 0.9993 (more rigorous)
- PAS++: τ1
= 0.975, τ3
= 0.9983 (more generous)
- The lower the thresholds, the more the skips.
Evaluation
1-Skipping
3-Skipping
43
Four Different Settings to Compare
● PAS
● PAS++
● 30-FPS (without RAVEN)
● 60-FPS (without RAVEN)
44
Game Apps used for Experiments
13 commercial game apps with various
graphical characteristics
A. Static
○ Simple graphic effects
ex) puzzles and board games
B. Dynamic
○ Continuous graphical changes
C. Hybrid
○ Clear separation between the input and
the response phase
45
Solitaire (Static)
Cookie Run
(Dynamic)
Angry Birds
(Hybrid)
Baseline of Video Quality
● Perceptual similarity w.r.t. the original sequence of frames
○ SSIM
○ VMAF [3]: A state-of-the-art video quality metric
46
[3] Li, Z., Aaron, A., Katsavounidis, I., Moorthy, A., & Manohara, M. (2016). Toward a practical perceptual video quality metric. The Netflix Tech
Blog, 6.
Result
47
Result (1): Comparison of PAS and PAS++
48
Result (1): Comparison of PAS and PAS++
Summary
➔ PAS++ skipped more frames than PAS.
➔ In 7 out of the 12 cases, PAS++ skipped as many
frames as 30-FPS setting.
49
Result (2):
Video
Quality
Scores
50
Result (2):
Video
Quality
Scores
Summary
● PAS and PAS++ achieved high video-quality scores.
○ PAS > PAS++ for any cases
● PAS and PAS++ produced better results than 30-FPS,
especially for the games in the Dynamic group.
51
Result (3): Energy Saving
on Nexus 5X
52
Result (3): Energy Saving
Summary
➔ PAS does not save substantial amount of energy in the
dynamic games. Except those games, PAS saves at
least half the amount of energy saved by 30-FPS.
➔ PAS++ saves at least 10% energy consumption.
on Nexus 5X
53
User Study
● 12 participants (7 males and 5 females, in ages from 20 to 30)
● Assessment Protocol
○ Double Stimulus Impairment Scale (DSIS)
○ Double Stimulus Continuous Quality Scale (DSCQS)
54
User Study: Results
55
Conclusion
● RAVEN: a new, on-the-fly perception-aware rate scaling technique
○ A light-weight frame comparison technique using Y-Diff
○ A low resolution virtual display
● Up-to 35% of total device energy saving
○ w/o any substantial user experience degradation
56
Discussion
● Online-learning of the coefficients in the EPS regression equation
○ As RAVEN learns coefficients ( dfs and ) using recorded
gameplay video in prior, it would be better to learn them while
users are playing game.
● Application to other domains
○ Watching a video on Youtube or Netflix
○ It consumes GPU as much as gameplay.
57
58
Thanks!
Appendix
59
Background: Graphics Processing on Android
60
SurfaceFlinger and Framebuffer
61
SSIM
62

More Related Content

What's hot

COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...
COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...
COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...Hemantha Kulathilake
 
4.intensity transformations
4.intensity transformations4.intensity transformations
4.intensity transformationsYahya Alkhaldi
 
Image Interpolation Techniques with Optical and Digital Zoom Concepts
Image Interpolation Techniques with Optical and Digital Zoom ConceptsImage Interpolation Techniques with Optical and Digital Zoom Concepts
Image Interpolation Techniques with Optical and Digital Zoom Conceptsmmjalbiaty
 
1536 graphics &amp; graphical programming
1536 graphics &amp; graphical  programming1536 graphics &amp; graphical  programming
1536 graphics &amp; graphical programmingDr Fereidoun Dejahang
 
Image Enhancement in Spatial Domain
Image Enhancement in Spatial DomainImage Enhancement in Spatial Domain
Image Enhancement in Spatial DomainA B Shinde
 
05 histogram processing DIP
05 histogram processing DIP05 histogram processing DIP
05 histogram processing DIPbabak danyal
 
Digital Image Processing_ ch2 enhancement spatial-domain
Digital Image Processing_ ch2 enhancement spatial-domainDigital Image Processing_ ch2 enhancement spatial-domain
Digital Image Processing_ ch2 enhancement spatial-domainMalik obeisat
 
//STEIM Workshop: A Vernacular of File Formats
//STEIM Workshop: A Vernacular of File Formats//STEIM Workshop: A Vernacular of File Formats
//STEIM Workshop: A Vernacular of File FormatsRosa ɯǝukɯɐn
 
6.frequency domain image_processing
6.frequency domain image_processing6.frequency domain image_processing
6.frequency domain image_processingNashid Alam
 
Naveen 9911103606 major ppt
Naveen 9911103606 major pptNaveen 9911103606 major ppt
Naveen 9911103606 major pptNaveen Rajgariya
 
New approach for generalised unsharp masking alogorithm
New approach for generalised unsharp masking alogorithmNew approach for generalised unsharp masking alogorithm
New approach for generalised unsharp masking alogorithmeSAT Publishing House
 
Point processing
Point processingPoint processing
Point processingpanupriyaa7
 

What's hot (20)

COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...
COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...
COM2304: Intensity Transformation and Spatial Filtering – III Spatial Filters...
 
4.intensity transformations
4.intensity transformations4.intensity transformations
4.intensity transformations
 
Image Interpolation Techniques with Optical and Digital Zoom Concepts
Image Interpolation Techniques with Optical and Digital Zoom ConceptsImage Interpolation Techniques with Optical and Digital Zoom Concepts
Image Interpolation Techniques with Optical and Digital Zoom Concepts
 
Bit plane coding
Bit plane codingBit plane coding
Bit plane coding
 
Spatial filtering
Spatial filteringSpatial filtering
Spatial filtering
 
1536 graphics &amp; graphical programming
1536 graphics &amp; graphical  programming1536 graphics &amp; graphical  programming
1536 graphics &amp; graphical programming
 
Lecture 7
Lecture 7Lecture 7
Lecture 7
 
Project004
Project004Project004
Project004
 
SPATIAL FILTER
SPATIAL FILTERSPATIAL FILTER
SPATIAL FILTER
 
Image Enhancement in Spatial Domain
Image Enhancement in Spatial DomainImage Enhancement in Spatial Domain
Image Enhancement in Spatial Domain
 
05 histogram processing DIP
05 histogram processing DIP05 histogram processing DIP
05 histogram processing DIP
 
Digital Image Processing_ ch2 enhancement spatial-domain
Digital Image Processing_ ch2 enhancement spatial-domainDigital Image Processing_ ch2 enhancement spatial-domain
Digital Image Processing_ ch2 enhancement spatial-domain
 
Lecture 3
Lecture 3Lecture 3
Lecture 3
 
Sharpening spatial filters
Sharpening spatial filtersSharpening spatial filters
Sharpening spatial filters
 
//STEIM Workshop: A Vernacular of File Formats
//STEIM Workshop: A Vernacular of File Formats//STEIM Workshop: A Vernacular of File Formats
//STEIM Workshop: A Vernacular of File Formats
 
6.frequency domain image_processing
6.frequency domain image_processing6.frequency domain image_processing
6.frequency domain image_processing
 
Lecture 4
Lecture 4Lecture 4
Lecture 4
 
Naveen 9911103606 major ppt
Naveen 9911103606 major pptNaveen 9911103606 major ppt
Naveen 9911103606 major ppt
 
New approach for generalised unsharp masking alogorithm
New approach for generalised unsharp masking alogorithmNew approach for generalised unsharp masking alogorithm
New approach for generalised unsharp masking alogorithm
 
Point processing
Point processingPoint processing
Point processing
 

Similar to RAVEN: Perception-aware Optimization of Power Consumption for Mobile Games

A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images
A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images
A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images IOSR Journals
 
Comparison of image segmentation
Comparison of image segmentationComparison of image segmentation
Comparison of image segmentationHaitham Ahmed
 
JPEG XR objective and subjective evaluations
JPEG XR objective and subjective evaluationsJPEG XR objective and subjective evaluations
JPEG XR objective and subjective evaluationsTouradj Ebrahimi
 
Survey paper on image compression techniques
Survey paper on image compression techniquesSurvey paper on image compression techniques
Survey paper on image compression techniquesIRJET Journal
 
Depth estimation do we need to throw old things away
Depth estimation do we need to throw old things awayDepth estimation do we need to throw old things away
Depth estimation do we need to throw old things awayNAVER Engineering
 
Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...
Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...
Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...NAVER Engineering
 
Gpu based-image-quality-assessment-using-structural-similarity-(ssim)-index
Gpu based-image-quality-assessment-using-structural-similarity-(ssim)-indexGpu based-image-quality-assessment-using-structural-similarity-(ssim)-index
Gpu based-image-quality-assessment-using-structural-similarity-(ssim)-indexMahesh Khadatare
 
Domainspecificsubgraph extraction ieee-bigdata2016
Domainspecificsubgraph extraction ieee-bigdata2016Domainspecificsubgraph extraction ieee-bigdata2016
Domainspecificsubgraph extraction ieee-bigdata2016Sarasi Sarangi
 
Lucas Theis - Compressing Images with Neural Networks - Creative AI meetup
Lucas Theis - Compressing Images with Neural Networks - Creative AI meetupLucas Theis - Compressing Images with Neural Networks - Creative AI meetup
Lucas Theis - Compressing Images with Neural Networks - Creative AI meetupLuba Elliott
 
Transformer in Vision
Transformer in VisionTransformer in Vision
Transformer in VisionSangmin Woo
 
Performance analysis of transformation and bogdonov chaotic substitution base...
Performance analysis of transformation and bogdonov chaotic substitution base...Performance analysis of transformation and bogdonov chaotic substitution base...
Performance analysis of transformation and bogdonov chaotic substitution base...IJECEIAES
 
Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...
Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...
Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...Anand Bhojan
 
Lossless Huffman coding image compression implementation in spatial domain by...
Lossless Huffman coding image compression implementation in spatial domain by...Lossless Huffman coding image compression implementation in spatial domain by...
Lossless Huffman coding image compression implementation in spatial domain by...IRJET Journal
 
Lec-08 Feature Aggregation II: Fisher Vector, AKULA and Super Vector
Lec-08 Feature Aggregation II: Fisher Vector, AKULA and Super VectorLec-08 Feature Aggregation II: Fisher Vector, AKULA and Super Vector
Lec-08 Feature Aggregation II: Fisher Vector, AKULA and Super VectorUnited States Air Force Academy
 
La question de la durabilité des technologies de calcul et de télécommunication
La question de la durabilité des technologies de calcul et de télécommunicationLa question de la durabilité des technologies de calcul et de télécommunication
La question de la durabilité des technologies de calcul et de télécommunicationAlexandre Monnin
 
EE660_Report_YaxinLiu_8448347171
EE660_Report_YaxinLiu_8448347171EE660_Report_YaxinLiu_8448347171
EE660_Report_YaxinLiu_8448347171Yaxin Liu
 

Similar to RAVEN: Perception-aware Optimization of Power Consumption for Mobile Games (20)

A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images
A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images
A Novel PSNR-B Approach for Evaluating the Quality of De-blocked Images
 
Comparison of image segmentation
Comparison of image segmentationComparison of image segmentation
Comparison of image segmentation
 
JPEG XR objective and subjective evaluations
JPEG XR objective and subjective evaluationsJPEG XR objective and subjective evaluations
JPEG XR objective and subjective evaluations
 
Survey paper on image compression techniques
Survey paper on image compression techniquesSurvey paper on image compression techniques
Survey paper on image compression techniques
 
Depth estimation do we need to throw old things away
Depth estimation do we need to throw old things awayDepth estimation do we need to throw old things away
Depth estimation do we need to throw old things away
 
Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...
Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...
Recommender Systems with Implicit Feedback Challenges, Techniques, and Applic...
 
Gpu based-image-quality-assessment-using-structural-similarity-(ssim)-index
Gpu based-image-quality-assessment-using-structural-similarity-(ssim)-indexGpu based-image-quality-assessment-using-structural-similarity-(ssim)-index
Gpu based-image-quality-assessment-using-structural-similarity-(ssim)-index
 
image compression ppt
image compression pptimage compression ppt
image compression ppt
 
Domainspecificsubgraph extraction ieee-bigdata2016
Domainspecificsubgraph extraction ieee-bigdata2016Domainspecificsubgraph extraction ieee-bigdata2016
Domainspecificsubgraph extraction ieee-bigdata2016
 
Domainspecificsubgraph extraction ieee-bigdata2016
Domainspecificsubgraph extraction ieee-bigdata2016Domainspecificsubgraph extraction ieee-bigdata2016
Domainspecificsubgraph extraction ieee-bigdata2016
 
Lucas Theis - Compressing Images with Neural Networks - Creative AI meetup
Lucas Theis - Compressing Images with Neural Networks - Creative AI meetupLucas Theis - Compressing Images with Neural Networks - Creative AI meetup
Lucas Theis - Compressing Images with Neural Networks - Creative AI meetup
 
TransNeRF
TransNeRFTransNeRF
TransNeRF
 
Transformer in Vision
Transformer in VisionTransformer in Vision
Transformer in Vision
 
Performance analysis of transformation and bogdonov chaotic substitution base...
Performance analysis of transformation and bogdonov chaotic substitution base...Performance analysis of transformation and bogdonov chaotic substitution base...
Performance analysis of transformation and bogdonov chaotic substitution base...
 
Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...
Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...
Energy Efficient Multi-player Smartphone Gaming using 3D Spatial Subdivisioni...
 
Lossless Huffman coding image compression implementation in spatial domain by...
Lossless Huffman coding image compression implementation in spatial domain by...Lossless Huffman coding image compression implementation in spatial domain by...
Lossless Huffman coding image compression implementation in spatial domain by...
 
Image compression .
Image compression .Image compression .
Image compression .
 
Lec-08 Feature Aggregation II: Fisher Vector, AKULA and Super Vector
Lec-08 Feature Aggregation II: Fisher Vector, AKULA and Super VectorLec-08 Feature Aggregation II: Fisher Vector, AKULA and Super Vector
Lec-08 Feature Aggregation II: Fisher Vector, AKULA and Super Vector
 
La question de la durabilité des technologies de calcul et de télécommunication
La question de la durabilité des technologies de calcul et de télécommunicationLa question de la durabilité des technologies de calcul et de télécommunication
La question de la durabilité des technologies de calcul et de télécommunication
 
EE660_Report_YaxinLiu_8448347171
EE660_Report_YaxinLiu_8448347171EE660_Report_YaxinLiu_8448347171
EE660_Report_YaxinLiu_8448347171
 

Recently uploaded

UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)
UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)
UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)Dr SOUNDIRARAJ N
 
An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...Chandu841456
 
Artificial-Intelligence-in-Electronics (K).pptx
Artificial-Intelligence-in-Electronics (K).pptxArtificial-Intelligence-in-Electronics (K).pptx
Artificial-Intelligence-in-Electronics (K).pptxbritheesh05
 
Instrumentation, measurement and control of bio process parameters ( Temperat...
Instrumentation, measurement and control of bio process parameters ( Temperat...Instrumentation, measurement and control of bio process parameters ( Temperat...
Instrumentation, measurement and control of bio process parameters ( Temperat...121011101441
 
computer application and construction management
computer application and construction managementcomputer application and construction management
computer application and construction managementMariconPadriquez1
 
Correctly Loading Incremental Data at Scale
Correctly Loading Incremental Data at ScaleCorrectly Loading Incremental Data at Scale
Correctly Loading Incremental Data at ScaleAlluxio, Inc.
 
8251 universal synchronous asynchronous receiver transmitter
8251 universal synchronous asynchronous receiver transmitter8251 universal synchronous asynchronous receiver transmitter
8251 universal synchronous asynchronous receiver transmitterShivangiSharma879191
 
complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...asadnawaz62
 
CCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdf
CCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdfCCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdf
CCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdfAsst.prof M.Gokilavani
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxJoão Esperancinha
 
Past, Present and Future of Generative AI
Past, Present and Future of Generative AIPast, Present and Future of Generative AI
Past, Present and Future of Generative AIabhishek36461
 
Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective IntroductionSachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective IntroductionDr.Costas Sachpazis
 
Comparative Analysis of Text Summarization Techniques
Comparative Analysis of Text Summarization TechniquesComparative Analysis of Text Summarization Techniques
Comparative Analysis of Text Summarization Techniquesugginaramesh
 
IVE Industry Focused Event - Defence Sector 2024
IVE Industry Focused Event - Defence Sector 2024IVE Industry Focused Event - Defence Sector 2024
IVE Industry Focused Event - Defence Sector 2024Mark Billinghurst
 
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...srsj9000
 
Architect Hassan Khalil Portfolio for 2024
Architect Hassan Khalil Portfolio for 2024Architect Hassan Khalil Portfolio for 2024
Architect Hassan Khalil Portfolio for 2024hassan khalil
 
Work Experience-Dalton Park.pptxfvvvvvvv
Work Experience-Dalton Park.pptxfvvvvvvvWork Experience-Dalton Park.pptxfvvvvvvv
Work Experience-Dalton Park.pptxfvvvvvvvLewisJB
 
Concrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptxConcrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptxKartikeyaDwivedi3
 

Recently uploaded (20)

UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)
UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)
UNIT III ANALOG ELECTRONICS (BASIC ELECTRONICS)
 
An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...An experimental study in using natural admixture as an alternative for chemic...
An experimental study in using natural admixture as an alternative for chemic...
 
Artificial-Intelligence-in-Electronics (K).pptx
Artificial-Intelligence-in-Electronics (K).pptxArtificial-Intelligence-in-Electronics (K).pptx
Artificial-Intelligence-in-Electronics (K).pptx
 
Instrumentation, measurement and control of bio process parameters ( Temperat...
Instrumentation, measurement and control of bio process parameters ( Temperat...Instrumentation, measurement and control of bio process parameters ( Temperat...
Instrumentation, measurement and control of bio process parameters ( Temperat...
 
computer application and construction management
computer application and construction managementcomputer application and construction management
computer application and construction management
 
Correctly Loading Incremental Data at Scale
Correctly Loading Incremental Data at ScaleCorrectly Loading Incremental Data at Scale
Correctly Loading Incremental Data at Scale
 
8251 universal synchronous asynchronous receiver transmitter
8251 universal synchronous asynchronous receiver transmitter8251 universal synchronous asynchronous receiver transmitter
8251 universal synchronous asynchronous receiver transmitter
 
Call Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCR
Call Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCRCall Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCR
Call Us -/9953056974- Call Girls In Vikaspuri-/- Delhi NCR
 
complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...complete construction, environmental and economics information of biomass com...
complete construction, environmental and economics information of biomass com...
 
CCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdf
CCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdfCCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdf
CCS355 Neural Network & Deep Learning UNIT III notes and Question bank .pdf
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
 
Past, Present and Future of Generative AI
Past, Present and Future of Generative AIPast, Present and Future of Generative AI
Past, Present and Future of Generative AI
 
Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective IntroductionSachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
Sachpazis Costas: Geotechnical Engineering: A student's Perspective Introduction
 
Comparative Analysis of Text Summarization Techniques
Comparative Analysis of Text Summarization TechniquesComparative Analysis of Text Summarization Techniques
Comparative Analysis of Text Summarization Techniques
 
IVE Industry Focused Event - Defence Sector 2024
IVE Industry Focused Event - Defence Sector 2024IVE Industry Focused Event - Defence Sector 2024
IVE Industry Focused Event - Defence Sector 2024
 
young call girls in Rajiv Chowk🔝 9953056974 🔝 Delhi escort Service
young call girls in Rajiv Chowk🔝 9953056974 🔝 Delhi escort Serviceyoung call girls in Rajiv Chowk🔝 9953056974 🔝 Delhi escort Service
young call girls in Rajiv Chowk🔝 9953056974 🔝 Delhi escort Service
 
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
Gfe Mayur Vihar Call Girls Service WhatsApp -> 9999965857 Available 24x7 ^ De...
 
Architect Hassan Khalil Portfolio for 2024
Architect Hassan Khalil Portfolio for 2024Architect Hassan Khalil Portfolio for 2024
Architect Hassan Khalil Portfolio for 2024
 
Work Experience-Dalton Park.pptxfvvvvvvv
Work Experience-Dalton Park.pptxfvvvvvvvWork Experience-Dalton Park.pptxfvvvvvvv
Work Experience-Dalton Park.pptxfvvvvvvv
 
Concrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptxConcrete Mix Design - IS 10262-2019 - .pptx
Concrete Mix Design - IS 10262-2019 - .pptx
 

RAVEN: Perception-aware Optimization of Power Consumption for Mobile Games

  • 1. RAVEN: Perception-aware Optimization of Power Consumption for Mobile Games Published in MobiCom’17 Chanyou Hwang et al. 1 Gabin An, Jinhan Kim 2018.10.31
  • 3. Background ● A mobile GPU constitutes a major portion of power consumption on the devices. ● Visually appealing mobile games => High energy cost GPU Battery 3
  • 4. Motivation 1 ● No matter the states, the game apps maintain high frame rates. 4 In-Game Lobby Crossy Road (길건너친구들)
  • 5. Motivation 2 ● Large portions of adjacent frames show strong similarity. 5
  • 6. Goal ● To reduce power consumption of mobile game playing without degrading user experiences. 6
  • 7. Goal ● To reduce power consumption of mobile game playing without degrading user experiences. 7 Recap.
  • 8. Challenges 1. The perceptual similarity should be measured efficiently. 2. The degradation of user experiences should be minimal. 3. The system should be transparent to the apps and avoid low-level system optimization (since mobile gaming apps and GPU drivers are closed source). 8 RAVEN RAVEN is a DC comics character whose ability is ‘sense and alter.’
  • 13. RAVEN: System Overview ● Three major components for scaling game-frame rendering ○ F-Tracker (FrameDiff Tracker) ○ R-Regulator (Rate Regulator) ○ R-Injector (Rate Injector) ● Small virtual display ○ Low resolution to reduce overhead 13
  • 14. RAVEN: Virtual Display (1) Mobile gameplay (1920 x 1080) Mobile gameplay (80 x 45) Virtual Display 14 piggyback
  • 15. RAVEN: F-Tracker (2) Virtual Display F-Tracker Down-scaled pixels Current Frame (fN ) Recently Rendered Frame (fN-k ) Calculate Y-Diff 15
  • 16. Determine how many frames will be skipped (No-Skipping | 1-Skipping | 3-Skipping) RAVEN: R-Regulator (3) R-Regulator predict the similarity scores with future frames Y-Diff(fN-k , fN ) 16 EPS: Estimated Perceptual Similarity
  • 17. RAVEN: R-Injector (4) R-Injector Inject a delay into the rendering loop of game app The # of frames to be skipped eglSwapBuffer() Frames with delay SurfaceFlinger 17
  • 18. How can we calculate similarity between frames? ● SSIM [1] ○ Existing method to measure user's perception of visual changes ■ SSIM > 0.97 ● A strong level of similarity between two images in mobile games [2] ■ Computationally expensive 😭 ● YUV Color Space ○ Color encoding system like RGB ○ Leverage Y (luminance) values to approximate SSIM efficiently [1] Wang, Zhou, et al. "Image quality assessment: from error visibility to structural similarity." IEEE transactions on image processing 13.4 (2004): 600-612. [2] Cuervo, E., Wolman, A., Cox, L. P., Lebeck, K., Razeen, A., Saroiu, S., & Musuvathi, M. (2015, May). Kahawai: High-quality mobile gaming using gpu offload. In Proceedings of the 13th Annual International Conference on Mobile Systems, Applications, and Services (pp. 121-135). ACM. 18
  • 19. Y-Diff ● Y−Diff is the difference in Y values of two images in the YUV color space. ○ luminance (Y): brightness ○ chrominance (UV): color ● Computation cost: O(N) ○ N = the number of pixels 19 UV space (Y′ value = 0.5)
  • 20. Would it be reasonable to use Y-Diff? the relation between Y−Diff and SSIM 20
  • 21. Would it be reasonable to use Y-Diff? the relation between Y−Diff and SSIM 21 Pearson’s correlation coefficient of -0.926
  • 22. Would it be reasonable to use Y-Diff? the relation between Y−Diff and SSIM 22 Pearson’s correlation coefficient of -0.926 SSIM and Y−Diff values are strongly correlated and show a linear relationship.
  • 23. Can we predict similarity of future frames? ● X-axis: Similarity between previous-current frames ● Y-axis: Similarity between current-future frames 23
  • 24. Can we predict similarity of future frames? ● X-axis: Similarity between previous-current frames ● Y-axis: Similarity between current-future frames 24 Pearson’s correlation coefficient of 0.93
  • 25. Can we predict similarity of future frames? ● X-axis: Similarity between previous-current frames ● Y-axis: Similarity between current-future frames 25 Pearson’s correlation coefficient of 0.93 The similarity with a future frame can be predicted from the similarity with a previous one.
  • 26. Perceptual Similarity Prediction Y-Diff between current frames and recently rendered frame 26 ● EPS estimates SSIM using Y-Diff (regression)
  • 27. Perceptual Similarity Prediction Constant Set depending on app and index k of previous frame (1, 2, 4) 27 ● EPS estimates SSIM using Y-Diff (regression) Constant
  • 28. Perceptual Similarity Prediction 28 ● EPS estimates SSIM using Y-Diff (regression) Average of recent Y-Diff scores over window size w (reflecting past history) w = 1, 2, 5, and 10 Best
  • 29. SSIM vs. EPS 29 Candy Crush Soda Saga Archery Master 3D
  • 30. SSIM vs. EPS 30 Candy Crush Soda Saga Archery Master 3D The predictions were more accurate within the potential range of threshold values (> 0.97)
  • 31. Determining the Scaling Factors (1) 31 1-Skipping 3-Skipping
  • 32. Determining the Scaling Factors (2) ● False transition ○ Skipping of a frame which should have not been skipped ○ EPS > τ1 or τ3 , but SSIM < 0.97 ● Increasing the τ1 and τ3 reduces the false transition ratio. ● For example, ○ when τ3 =0.97, the ratio of false transition is 27% for 3-Skipping ○ when τ3 =0.9945, the ratio of false transition is 2.7% for 3-Skipping 32 below the strong level of similaritya skip occur
  • 33. Cloning the Primary Display: Virtual Display ● A virtual display could be allocated and maintained by SurfaceFlinger of Android. ● The overhead of a virtual display is primarily proportional to the resolution of the display. 33
  • 34. Cloning the Primary Display: Why a 0.9989 Pearson’s correlation 34
  • 35. Cloning the Primary Display: Why 35
  • 36. Cloning the Primary Display: Why 36 The Y−Diff results with 80×45 and 1920×1080 are nearly identical with the selected game-play data, scoring a 0.9989 Pearson’s correlation coefficient
  • 38. RAVEN: Virtual Display (1) Mobile gameplay (1920 x 1080) Mobile gameplay (80 x 45) Virtual Display 38 piggyback
  • 39. RAVEN: F-Tracker (2) Virtual Display F-Tracker Down-scaled pixels Current Frame (fN ) Recently Rendered Frame (fN-k ) Calculate Y-Diff 39
  • 40. Determine how many frames will be skipped (No-Skipping | 1-Skipping | 3-Skipping) RAVEN: R-Regulator (3) R-Regulator predict the similarity scores with future frames Y-Diff 40 EPS: Estimated Perceptual Similarity
  • 41. RAVEN: R-Injector (4) R-Injector Inject a delay into the rendering loop of game app The # of frames to be skipped eglSwapBuffer() Frames with delay SurfaceFlinger 41
  • 43. - Two different settings for the evaluation of RAVEN - PAS : τ1 = 0.9975, τ3 = 0.9993 (more rigorous) - PAS++: τ1 = 0.975, τ3 = 0.9983 (more generous) - The lower the thresholds, the more the skips. Evaluation 1-Skipping 3-Skipping 43
  • 44. Four Different Settings to Compare ● PAS ● PAS++ ● 30-FPS (without RAVEN) ● 60-FPS (without RAVEN) 44
  • 45. Game Apps used for Experiments 13 commercial game apps with various graphical characteristics A. Static ○ Simple graphic effects ex) puzzles and board games B. Dynamic ○ Continuous graphical changes C. Hybrid ○ Clear separation between the input and the response phase 45 Solitaire (Static) Cookie Run (Dynamic) Angry Birds (Hybrid)
  • 46. Baseline of Video Quality ● Perceptual similarity w.r.t. the original sequence of frames ○ SSIM ○ VMAF [3]: A state-of-the-art video quality metric 46 [3] Li, Z., Aaron, A., Katsavounidis, I., Moorthy, A., & Manohara, M. (2016). Toward a practical perceptual video quality metric. The Netflix Tech Blog, 6.
  • 48. Result (1): Comparison of PAS and PAS++ 48
  • 49. Result (1): Comparison of PAS and PAS++ Summary ➔ PAS++ skipped more frames than PAS. ➔ In 7 out of the 12 cases, PAS++ skipped as many frames as 30-FPS setting. 49
  • 51. Result (2): Video Quality Scores Summary ● PAS and PAS++ achieved high video-quality scores. ○ PAS > PAS++ for any cases ● PAS and PAS++ produced better results than 30-FPS, especially for the games in the Dynamic group. 51
  • 52. Result (3): Energy Saving on Nexus 5X 52
  • 53. Result (3): Energy Saving Summary ➔ PAS does not save substantial amount of energy in the dynamic games. Except those games, PAS saves at least half the amount of energy saved by 30-FPS. ➔ PAS++ saves at least 10% energy consumption. on Nexus 5X 53
  • 54. User Study ● 12 participants (7 males and 5 females, in ages from 20 to 30) ● Assessment Protocol ○ Double Stimulus Impairment Scale (DSIS) ○ Double Stimulus Continuous Quality Scale (DSCQS) 54
  • 56. Conclusion ● RAVEN: a new, on-the-fly perception-aware rate scaling technique ○ A light-weight frame comparison technique using Y-Diff ○ A low resolution virtual display ● Up-to 35% of total device energy saving ○ w/o any substantial user experience degradation 56
  • 57. Discussion ● Online-learning of the coefficients in the EPS regression equation ○ As RAVEN learns coefficients ( dfs and ) using recorded gameplay video in prior, it would be better to learn them while users are playing game. ● Application to other domains ○ Watching a video on Youtube or Netflix ○ It consumes GPU as much as gameplay. 57