You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Auto merge of rust-lang#3194 - saethlin:remove-gc-heuristics, r=RalfJung
Remove Stacked Borrows GC heuristics
Removing these has no impact on our benchmarks. I think I initially added these because they have a significant impact on runtime at shorter GC intervals. But both these heuristics result in undesirable memory growth in real programs, especially `modified_since_last_gc`. I didn't realize at the time that required state becomes garbage as a result of changes to _other_ allocations.
I think this nets even primarily because we get better heap reuse. With this change I see almost all the mmap calls coming from our diagnostics infrastructure go away. Not that there were many to start with, but it's an indicator that our memory locality has improved.
Before:
```
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/backtraces/Cargo.toml"
Time (mean ± σ): 4.046 s ± 0.087 s [User: 3.952 s, System: 0.085 s]
Range (min … max): 3.952 s … 4.139 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/invalidate/Cargo.toml"
Time (mean ± σ): 6.271 s ± 0.073 s [User: 6.206 s, System: 0.054 s]
Range (min … max): 6.195 s … 6.365 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/mse/Cargo.toml"
Time (mean ± σ): 570.3 ms ± 6.7 ms [User: 505.5 ms, System: 61.8 ms]
Range (min … max): 559.6 ms … 576.0 ms 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/serde1/Cargo.toml"
Time (mean ± σ): 2.013 s ± 0.012 s [User: 1.938 s, System: 0.069 s]
Range (min … max): 1.994 s … 2.024 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/serde2/Cargo.toml"
Time (mean ± σ): 4.155 s ± 0.082 s [User: 4.078 s, System: 0.067 s]
Range (min … max): 4.011 s … 4.211 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/slice-get-unchecked/Cargo.toml"
Time (mean ± σ): 541.5 ms ± 3.9 ms [User: 477.3 ms, System: 60.0 ms]
Range (min … max): 536.1 ms … 545.1 ms 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/unicode/Cargo.toml"
Time (mean ± σ): 1.496 s ± 0.002 s [User: 1.442 s, System: 0.050 s]
Range (min … max): 1.494 s … 1.500 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/zip-equal/Cargo.toml"
Time (mean ± σ): 2.190 s ± 0.043 s [User: 2.109 s, System: 0.075 s]
Range (min … max): 2.114 s … 2.215 s 5 runs
```
After:
```
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/backtraces/Cargo.toml"
Time (mean ± σ): 3.954 s ± 0.057 s [User: 3.871 s, System: 0.075 s]
Range (min … max): 3.912 s … 4.052 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/invalidate/Cargo.toml"
Time (mean ± σ): 6.200 s ± 0.108 s [User: 6.129 s, System: 0.060 s]
Range (min … max): 6.072 s … 6.295 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/mse/Cargo.toml"
Time (mean ± σ): 575.3 ms ± 9.3 ms [User: 511.5 ms, System: 60.2 ms]
Range (min … max): 558.9 ms … 580.8 ms 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/serde1/Cargo.toml"
Time (mean ± σ): 1.966 s ± 0.007 s [User: 1.903 s, System: 0.058 s]
Range (min … max): 1.959 s … 1.974 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/serde2/Cargo.toml"
Time (mean ± σ): 4.138 s ± 0.040 s [User: 4.057 s, System: 0.072 s]
Range (min … max): 4.110 s … 4.207 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/slice-get-unchecked/Cargo.toml"
Time (mean ± σ): 541.8 ms ± 5.6 ms [User: 470.7 ms, System: 66.9 ms]
Range (min … max): 535.6 ms … 549.1 ms 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/unicode/Cargo.toml"
Time (mean ± σ): 1.490 s ± 0.021 s [User: 1.424 s, System: 0.060 s]
Range (min … max): 1.454 s … 1.505 s 5 runs
Benchmark 1: cargo miri run --manifest-path "/home/ben/miri/bench-cargo-miri/zip-equal/Cargo.toml"
Time (mean ± σ): 2.215 s ± 0.048 s [User: 2.113 s, System: 0.072 s]
Range (min … max): 2.183 s … 2.299 s 5 runs
```
0 commit comments