Using Generational Garbage Collection To Implement Cacheconscious Data Placement

1
Using Generational Garbage Collection To
Implement Cache-conscious Data Placement

• Trishul M. Chilimbi James R. Larus
• ???? ????? ???

2
Introduction

• Main memory access cost is increasing
• goal to improve cache locality
• introducing a technique for using a (copying)
  generational GC to reorganize data, so that
  objects with high temporal affinity are placed
  next to each other and thus are likely to reside
  in the same cache block

3
Contents

• background
• copying GC
• generational GC
• the method
• profiling instrumentation
• affinity graph
• algorithm steps
• results conclusions

4
Copying GC

• Two memory areas
• When FROMSPACE is full, moves all the live
  objects from FROMSPACE to TOSPACE

5
Copying GC
6
Copying GC (cheney algorithm)

• Breadth-first scan of the tree
• one continuous scan of TOSPACE

7
Copying GC (cheney algorithm)
8
Why Generational GC ?

• Most objects live a very short time, while a
  small percentage of them live much longer
• problem repeated copying of old objects

9
Generational GC

• segregating objects into multiple areas by age
• scavenge older objects less frequently
• copy long surviving objects to older generations

10
Real time data access profiling

• Real time profiling is more effective then an
  earlier training run
• must be low overhead
• done by modified compiler
• upon access , the object address is written into
  an object access buffer

11
Real time data access profiling
12
Profiling is low-overhead

• Implemented at object, not field, granularity
• most object accesses are not lightweight

13
Affinity graph

• Is based on object access buffer
• created prior to scavenge
• separate graph for each generation
• nodesobjects
• edgesaffinity between objects

14
ADFGDACCAFDGAC
A
1
D
15
ADFGDACCAFDGAC
A
1
F
1
1
D
16
ADFGDACCAFDGAC
A
1
F
1
1
1
G
D
1
17
ADFGDACCAFDGAC
A
1
F
1
1
2
G
D
2
18
Cache-conscious copying algorithm
Step1

• From the set of roots, pick the one with the
  highest affinity edge
• perform a greedy depth first traversal of the
  affinity graph starting from this node
• while traversing, copy each visited object to
  TOSPACE

19
(No Transcript)
20
Cache-conscious copying algorithm
Step2

• Process all objects between the unprocessed and
  free pointers, using Cheney algorithm

21
(No Transcript)
22
Cache-conscious copying algorithm
Step3

• Cleanup copy any roots not present to TOSPACE
  and process using Cheney algorithm

23
(No Transcript)
24

• this algorithm is not used in the youngest
  generation (where new objects are allocated and
  most of the garbage is generated)

25
Results

• Tested on 5 Cecil language programs
• on a Sun computer with 2GB memory, with 2 level
  cache, running Solaris

26
(No Transcript)
27
(No Transcript)
28
Conclusions

• This is an attractive technique that reduces
  cache miss rates by 21-42 and improves program
  performance by 14-37 , as compared to the
  commonly used alternative.