Lecture 15 Advanced Garbage Collection

Transcription

1 Lecture 15 Advanced Garbage Collection I. Break Up GC in Time (Incremental) II. Break Up GC in Space (Partial) Readings: Ch CS243: Advanced Garbage Collection 1

2 Trace-Based GC: Memory Life-Cycle CS243: Advanced Garbage Collection 2

3 Incremental GC Interleaves GC with mutator action to reduce pause time Stop during trace: Ideal = (R New) Lost CS243: Advanced Garbage Collection 3

4 Effects of Mutation Reachable set changes as mutator runs R: set of reachable objects before the mutator runs Ideal: set of reachable objects at the end of the GC cycle New: set of newly created objects Lost: set of objects that become unreachable in the interim Ideal = (R New) - Lost Ideal: Very expensive Conservative Incremental GC: May misclassify some unreachable as reachable should not include objects unreachable before GC starts guarantees that garbage will be eliminated in the next round Ideal = (R New) - Lost Answer (R New) CS243: Advanced Garbage Collection 4

5 Algorithm Proposal 1 Initial condition Scanned, Unscanned lists from before To resume GC Find root sets Place newly reached objects in unscanned list Continue to trace reachability without redoing scanned objects Did we find all reachable objects? CS243: Advanced Garbage Collection 5

6 Missed Reachable Objects All reaching pointers are found in scanned objects Requires the occurrence of a 3-step sequence in the mutator: CS243: Advanced Garbage Collection 6

7 Solution Intercept p in any of the three-step sequence Treat pointee of p as unscanned CS243: Advanced Garbage Collection 7

8 Efficiency of Different Barriers Most efficient: Write barrier less instances than read barrier includes less unreachable objects than overwrite barriers Implementing Write Barriers Remember all locations that are written Divide memory into "cards", set a bit for each card Use MMU to protect memory pages containing scanned objects CS243: Advanced Garbage Collection 8

9 II. Partial GC Reduces pause time by collecting only objects in the target area: Algorithm New root set = original root set + pointers from Stable to Target set Change mutator to intercept all writes to Stable set Never misclassify reachable as unreachable May misclassify unreachable as reachable CS243: Advanced Garbage Collection 9

10 Generational GC Observation: objects die young 80-98% die within a few million instructions or before 1 MB has been allocated Generational GC: trace newly allocated objects more often P 3 P 2 P 1 ith generation new root set = original root set + all pointers from generations j to i (j > i) When 1st generation fills up GC copies reachable objects into 2nd generation, and so on. P 0 CS243: Advanced Garbage Collection 10

11 Generational GC Example more mature P 2 When P k fills, move reachable in P 0...P k to P 1...P k+1 and reset P 0. "remembered set" P 1 When P 1 fills, move reachable in P 0 +P 1 to P 1 +P 2 and reset P 0. less mature "remembered set" P 0 New objects go in P 0. When P 0 fills, move reachable in P 0 to P 1 and reset P 0. CS243: Advanced Garbage Collection 11

12 Properties Never misclassify reachable as unreachable Misclassify unreachable as reachable when pointers in earlier generations are overwritten eventually collect all garbage as generations get larger Effective: time spent on partitions with lots of garbage GC of mature objects takes longer Size of target set increases Eventually a full GC is performed CS243: Advanced Garbage Collection 12

13 The Train Algorithm Problem with generational GC: 1. Occasional total collection (last partition). 2. Long-lived objects move many times. Train algorithm useful for long-lived objects. Replaces higher-numbered partitions in generational GC. I.e., first partition is not part of the train, and is used for new objects just as in the previous algorithm. 13

14 Partitions = Cars Train 1 Car 11 Car 12 Car 13 Train 2... Car 21 Car Car 2k Train n Car n1 Car n2 14

15 Organization of Heap There can be any number of trains, and each train can have any number of cars. You need to decide on a policy that gives a reasonable number of each. Objects entering the system can: Be placed in the last car of the last train. If no room, start a new car at the end of the last train. If train too long, start a new train (which becomes the last). 15

16 Remembered Sets Each car has a remembered set of references from 1. Higher-numbered trains and 2. Higher-numbered cars of the same train. Important: since we only garbage-collect first cars and first trains, we never need to worry about forward references (to later trains or later cars of the same train). 16

17 Garbage-Collection Steps 1. Collect the first car of the first train. Like we collect P 0 in generational GC. 2. Collect the entire first train if there are no references from the root set or other trains. This is how we find and eliminate large, cyclic garbage structures. 17

18 Collecting the First Car of the First Train 1. Do a partial collection as before, using every other car/train as the stable set. 2. Move all Scanned (reachable) objects of the first car somewhere else. 3. Get rid of the car. 18

19 Moving Reachable Objects If object o has a reference from another train, pick one such train and move o to that train. Same car as reference, if possible, else make new car. If references only from root set or first train, move o to another car of first train, or create new car. 19

20 Panic Mode The problem: It is possible that when collecting the first car, nothing is garbage. We may then create a new last car of the first train with the same objects as the old first car. Situation is rare, but possible, requiring: 1. A cyclic, but nongarbage structure in the first train. 2. A mutator that changes references to the first train with perfect timing so that we never see anything in the remembered sets for the first car from outside the first train. 20

21 Panic Mode (2) If that happens, we go into panic mode, which requires that: 1. If a reference to any object in the first train is rewritten, we make the new reference a dummy member of the root set. 2. During panic-mode GC, if we encounter a reference from the dummy root set, we move the referenced object to another train. Eventually, all reachable objects will leave the first train, leaving cyclic garbage, and we can delete the first train. 21

22 Conclusions Trace-based GC: find all reachable objects, complement to get unreachable 4 states: free, unreached, unscanned, scanned break up reachability analysis in time (incremental) in space (partial: generational) CS243: Advanced Garbage Collection 22