The R-LRPD Test: Speculative Parallelization of Partially Parallel Loops

1
The R-LRPD TestSpeculative Parallelization of
Partially Parallel Loops

• Francis Dang, Hao Yu, and Lawrence Rauchwerger
• Department of Computer Science
• Texas AM University

2
Motivation

• To maximize performance, extract the maximum
  available parallelism from loops.
• Static compiler methods may be insufficient.
• Access patterns may be too complex.
• Required information is only available at
  runtime.
• Run-time methods needed to extract loop
  parallelism
• Inspector/Executor
• Speculative Parallelization

3
Speculative Parallelization LRPD Test

• Main Idea
• Execute a loop as a DOALL.
• Record memory references during execution.
• Check for data dependences.
• If there was a dependence, re-execute the loop
  sequentially.
• Disadvantages
• One data dependence can invalidate speculative
  parallelization.
• Slowdown is proportional to speculative parallel
  execution time.
• Partial parallelism is not exploited.

4
Partially Parallel Loop Example
do i 1, 8 z AKi ALi z Ci end do K18 1,2,3,1,4,2,1,1 L18 4,5,5,4,3,5,3,3
iter 1 2 3 4 5 6 7 8
A()
1 R R R R
2 R R
3 R W W W
4 W W R
5 W W W
5
The Recursive LRPD

• Main Idea
• Transform a partially parallel loop into a
  sequence of fully parallel, block-scheduled
  loops.
• Iterations before the first data dependence are
  correct and committed.
• Re-apply the LRPD test on the remaining
  iterations.
• Worst case
• Sequential time plus testing overhead

6
Algorithm
7
Implementation

• Implemented in run-time pass in Polaris and
  additional hand-inserted code.
• Privatization with copy-in/copy-out for arrays
  under test.
• Replicated buffers for reductions.
• Backup arrays for checkpointing.

8
Recursive LRPD Example
do i 1, 8 z AKi ALi z Ci end do K18 1,2,3,1,4,2,1,1 L18 4,5,5,4,2,5,3,3
9
Heuristics

• Work Redistribution
• Sliding Window Approach
• Data Dependence Graph Extraction

10
Work Redistribution

• Redistribute remaining iterations across
  processors.
• Execution time for each stage will decrease.
• Disadvantages
• May uncover new dependences across processors.
• May incur remote cache misses from data
  redistribution.

11
Work Redistribution Example
do i 1, 8 z AKi ALi z Ci end do K18 1,2,3,1,4,2,1,1 L18 4,5,5,4,2,5,3,3
12
Redistribution Model

• Redistribution may not always be beneficial.
• Stop redistribution if
• The cost of data redistribution outweighs the
  benefit from work redistribution.
• Synthetic loop to model this adaptive method.

13
Redistribution Model
14
Sliding Window R-LRPD

• R-LRPD can generate a sequential schedule for
  long dependence distributions.
• Strip-mine the speculative execution.
• Apply the R-LRPD on a contiguous block of
  iterations.
• Only dependences within the window cause
  failures.
• Adds more global synchronizations and test
  overhead.

15
DDG Extraction

• R-LRPD can generate sequential schedules for
  complex dependence distributions.
• Use the SW R-LRPD scheme to extract the data
  dependence graph (DDG).
• Generate an optimized schedule from the DDG.
• Obtains the DDG for loops from which a proper
  inspector cannot be extracted.

16
Performance Issues

• Performance issues
• Blocked scheduling potential cause for load
  imbalance.
• Checkpointing can be expensive.
• Feedback guided blocked scheduling
• Use the timing information from the previous
  instantiation (Bull, EuroPar 98)
• Estimate the processor chunk sizes for minimal
  load imbalance.
• On-Demand Checkpointing
• Checkpoint only data modified during execution.

17
Experiments

• Setup
• 16 processor HP V-Class
• 4 GB memory
• HP-UX 11.0

18
Experimental Results Input Profiles
19
Experimental Results - TRACK
20
Experimental Results - TRACK
21
Experimental Results - TRACK
22
Experimental Results - TRACK
23
Experimental Results Sliding Window
24
Experimental Results Sliding Window
25
Experimental Results FMA3D
26
Experimental Results SPICE 2G6
27
Conclusion

• Contribution
• Can speculatively parallelize any loop.
• Concern is now optimizing the parallelization and
  not when to parallelize.
• Future work
• Use dependence distribution information for
  adaptive redistribution and scheduling.