Time Critical Computing

1
Time Critical Computing

• VRGroup, KAIST
• JunHyeok Heo

2
Contents

• Introduction
• Real-time System
• VR system as RTS
• Time Critical System
• Time critical rendering
• Time critical simulation
• Conclusion

3
Introduction

• Presence
• fast input output processing
• Performance Requirement
• more than 10 frames/sec
• less than 0.1 sec
• time budget

4
Introduction

• A choice with a very complex environment and a
  limited H/W
• Gracefully degradation in computation quality to
  meet the constraints
• Time Critical Computation Bryson97

5
Real-Time System

• What is time ?
• If no one asks me i know. But if I want to
  explain it to someone, I do not know it. by St.
  Augustine (A.D. 354-430)
• What is Real-Time System ?
• A system operated on the real-time operating
  mode.
• Real-Time Operating Mode
• - An operation of a computer system, in which
  the programs for the processing of incoming data
  are constantly operational so that the processing
  results are available within a given time
  interval. Depending on the application, the data
  may appear at random or at predetermined times.
• by German Industrial Standardization DIN44 300

6
Requirements for RTS

• Timeliness - relative time, absolute time
• Simultaneity - inputs from two or more
  independent sources
• Predictability - reactions are fully predictable
  w.r.t. the input from the environment
• prohibit the use of virtual memory or cache
• Dependability - correctness, robustness, readiness

7
Soft RTS Hard RTS

• Costs associated with a failure to meet specified
  timing constraints
• Soft Real-Time System
• a gradually increasing continuous function of
  time
• Hard Real-Time System
• a discontinous fast increasing function of time

8
VR System as RTS

• How much complex or dynamically changing in time
  the behavior of the displayed virtual environment
  is ?

Static Environment
Simple soft real-time system walk/fly through
systems
Soft real-time systems with the fixed number of
dynamic objects
Hard real-time system Augmented reality systems
Not real-time system systems with the variable
number of dynamic objects whose behaviors are
nondeterministic
Dynamic Environment
9
VR System as RTS

• Three kinds of time Bryson94
• Wall Time
• Display Time
• when objects are displayed
• Environment Time
• when objects are computed
• Runtime Architectures Bryson94,97
• Simulation loop
• Concurrent processes
• Separated Rendering/Simulation

10
(No Transcript)
11
(No Transcript)
12
VR System as RTS

• Time Constraints
• Rendering Task
• should render new image within 0.1 sec from the
  output time of previous rendered image
• User Input Task
• should render the image of newly updated V.E
  within 0.1 sec from the user input
• Simulation Tasks
• application dependent
• eg) a V.E. with a train and a stop

13
VR Systems as RTS

• Contrast with RTS
• Predictability
• Strictly known execution time of activated tasks
• Resource requirements
• Strong real-time scheduling of dynamic tasks
• Either time limits are met or they are not
• NP-hard even though the deadline and resource
  requirements are known at the initiated time
• VR systems
• Performance Constraint
• try to optimize the graceful degradation of
  performance
• Time Constraint between active objects

14
Time Critical System

• Time Critical Computation
• meet the soft deadline at the cost of gracefully
  degraded computation quality
• Scheduling Problem
• select the suitable LOD level of computation to
  meet time constraints
• difficult when time constraints by active objects
  exist
• very difficult in case of concurrent processes
  architecture

15
Time Critical Rendering

• Rendering Task
• Generate a image I with a virtual environment W
  and a synthetic view v within a given time
• Graceful degradation in image quality
• Approaches
• Geometry Based
• Remove or simplify perceptually unimportant parts
• Culling and LOD subtasks
• Image Based
• Image Pyramid
• Hybrid
• Cached image as a simplified level

16
Time Critical Rendering

• Culling Task
• invisible things equally does not contribute to
  image quality
• remove invisible things as many as possible
  within the time budget
• remove partly invisible things
• the number of remained polygons
• the image quality
• multi step task with LOD task

17

• LOD Task
• Discrete LOD
• select the suitable LOD for each object as
  follows Funkhouser93

Mark96
18

• Continuous LOD
• LOD Selection
• 1. Compute
• 2. If then
• select a set of objects, such that changing
  their graphical representation will produce a
  time saving of
• else if then
• select a set of objects, such that changing
  their graphical representation will produce an
  improvement in quality and consume at most time
• 3. Compute new graphical representation
• 4. Display the objects

19

• Tree Structure for Continuous LOD
• eg.) Vertex Tree Luebke97, Hoppe97
• A node representative vertex of child nodes
• A path simplification/subdivision sequence
• Parent node - low detail
• Children node - high detail
• While (TimeltTimeBudget)
• search down the tree
• refine the current result
• Progressive Refinement/Progressive Deepening

20
Time Critical Simulation

• Very application dependent
• Preemptive by time event
• Benefit of computation
• Accuracy of computation, quality of computation
• Cost of computation
• Execution time
• difficult if the execution paths are not well
  defined or dynamic

21

• simple feed back method
• assume all computational tasks are equal and
  divide the time budget equally by the number of
  tasks
• scale each tasks time budget by (desired frame
  time/actual frame time)
• do not apply too small or large scale to avoid
  oscillation, filtering out

Specified Total Frame Time
22
Time Critical Simulation

• LOD in computation
• discrete LOD
• prepare n alternative computations
• continuous LOD
• order of interpolation or integration
• reduce the computation to progressive refinement
  using tree structure (eg. time critical collision
  detection with hierarchical bounding sphere
  Hubbard95

23
(No Transcript)
24
Time Critical Simulation

• Simulation of active objects with time
  constraints between each other
• Specification of time constraints
• Real-time scheduling

25
Conclusion

• Real-time system
• VR system as RTS
• Time critical rendering
• geometry based rendering
• in case of including image as a rendering
  primitive
• Time critical simulation
• application dependent
• in case of active objects with time constraints
  between each other, need many experiences to
  develop many V.E.s with constraints based on a
  scenario or story
• good start point with the research results from
  real time system
• specification of time constraints
• real-time scheduling