Power Analysis of Mobile 3D Graphics

About This Presentation

Title:

Power Analysis of Mobile 3D Graphics

Description:

Scan-line conversion. Pixel rendering. Texturing. Texturing. 10. Workload Across Applications ... Scan-line. conversion. Per-pixel. Operations. Transform ... – PowerPoint PPT presentation

Number of Views:26

Avg rating:3.0/5.0

Slides: 29

Provided by: brenmo

Category:

more less

Transcript and Presenter's Notes

Title: Power Analysis of Mobile 3D Graphics

1
(No Transcript)
2
-Based Workload
Estimation for Mobile 3D Graphics

Bren Mochocki, Kanishka Lahiri, Srihari
Cadambi, Xiaobo Sharon Hu
NEC Laboratories America, University of Notre
Dame

DAC 2006
3
Mobile Graphics Technology

Increasing resource load
Performance (Speed)
Lifetime (Energy)

Graphics Technology
Advanced 3D
Basic 3D
Video clips
2D color
1997
Time
4
Meeting Performance/Lifetime Requirements
Hardware Solutions

Woo, 04
low-power 3D ASIC

Kameyama, 03
low-power 3D ASIC

Gu, Chakraborty, Ooi, 06
Games are up for DVFS

Akenine-Moller, 03
Texture compression
for mobile terminals

Mochocki, Lahiri, Cadambi, 06
DVFS for mobile 3D graphics

Graphics Algorithms
System - Level Optimizations

Tack, 04
LoD control for mobile terminals

Accurate workload prediction is critical
5
Mobile 3D Workload Estimation

Why?
Adapt architectural parameters
Adapt application parameters
Better on-line resource management
Desirable properties
Speed must be performed on-line
Accuracy
Compact

6
Workload-Estimation Spectrum
General Purpose Simplicity
Application specific Accuracy
History-Based Predictors
Analytical Predictors

General purpose history-based predictors provide
poor prediction accuracy for rapidly changing
workloads
Highly accurate analytical schemes are too
complex for use at run time

7
Workload-Estimation Spectrum
General Purpose Simplicity
Application specific Accuracy
Signature-Based Predictor

Uses combination of history and
application-specific parameters (the signature)
to predict future workload
Strikes a balance between simplicity and accuracy
Preserves both cause AND effect
Preserves substantial history

8
Outline

Introduction and Motivation
Background
3D-pipeline Basics
Challenges in workload Estimation
Signature-Based Workload Prediction
Experimental Results
Conclusions

9
3D Pipeline Basics

3D representation ? 2D image

Geometry
Setup
Rendering
World View
Camera View
Raster View
Frame Buffer

Transformations
Lighting

Clipping
Scan-line conversion

Pixel rendering
Texturing

10
Workload Across Applications
12
RoomRev
TexCube
10
8
6
Execution Cycles (ARM, x107)
4
2
0
Benchmark

Workload varies significantly between
applications
Prediction scheme must be flexible

11
Workload Within an Application

Workload can change rapidly between frames

Race
geometry
render
Execution Cycles (ARM, x107)
setup
1
16
31
46
61
76
91
106
121
136
151
166
181
196
Frame
12
Outline

Introduction and Motivation
Background
Signature-Based Workload Prediction
Signature Generation
Method Overview
Pipeline Modifications
Experimental Results
Conclusions

13
Example
Signature ltvertex count, avg. areagt
3D Pipeline
end frame

Frame Buffer
Application
extract
extract signature
measure workload
1.0e4
lt6, 2.5gt
Signature Workload

Default
Signature Table
Workload Prediction
14
Example
Signature ltvertex count, avg. areagt
3D Pipeline
end frame

Frame Buffer
Application
extract
extract signature
measure workload
1.0e4
lt6, 2.5gt
Signature Workload
lt6, 2.5gt

Signature Table
1.0e4
1.0e4
Workload Prediction
15
Example
Signature ltvertex count, avg. areagt
3D Pipeline
end frame

Frame Buffer
Application
extract
No overlap (render all pixels)
extract signature
measure workload
1.2e4
lt6, 2.5gt
Signature Workload
lt6, 2.5gt

Signature Table
1.0e4
1.0e4
Workload Prediction
16
Partitioning the 3D pipeline
ORIGINAL
GEOMETRY
SETUP
RENDER
Application
Display
Transform
Clipping
Transform
Clipping
Lighting
Scan-line conversion
Per-pixel Operations
Lighting
Scan-line conversion
Per-pixel Operations
PARTITIONED
Application
Display
Transform Clipping
Scan-line conversion
Per-pixel Operations
Lighting
Buffer
Pre-Buffer
Post Buffer
17
Pipeline Workload

Pre-buffer workload is less than 10 of the total
workload
Pre-buffer variation is small
Post-buffer workload is large with significant
variation

18
Signature Composition

Can vary by application
May include
Average Triangle Area
Average Triangle Height
Total vertex count
Lit vertex count
Number of lights
Any measurable parameter
Larger signatures ? more accurate
Smaller signatures ? less time space

19
Outline

Introduction Background
Experimental Framework
Signature-Based Workload Prediction
Experimental Results
Evaluation Framework
Signature length vs. accuracy
Frame Rate
Energy
Conclusions

20
Architectural View

pre-buffer
signature extraction

post-buffer
Applications Processor
Programmable 3D Graphics Engine
Performance counter
System-level Communication Architecture
Frame Buffer
Memory
measure workload

buffer
signature table

output
21
Evaluation Framework
OpenGL/ES library Instrumented with pipeline
stage triggers Hans-Martin Will
Vincent
Fast, cycle-accurate Simulation W. Qin
3D application
Cross Compiler ARM g
Simit-ARM
OpenGL/ES 1.0 3D application
Trace simulator of mobile 3D pipeline
Triangle, Instruction, Trigger traces
Workload prediction scheme
Trace Simulator
Architecture Model
Processor Energy Model
3D pipeline Performance/power
Simulation output
22
Workload Accuracy
gt 2 fps error at peaks
Average Error (normalized)
Peaks lt 1 fps
ltagt 2 bytes
lta,bgt 6 bytes
lta,b,cgt 10 bytes
lta,b,c,dgt 14 bytes
Signature Complexity
ltagt triangle count, ltbgt avg. area, ltcgt avg.
height, ltdgt vertex count
23
Frame Rate
High peaks result in wasted energy
Target
Low valleys result in poor visual quality
24
Workload prediction for DVFS
32 energy reduction
25
Outline