forked from TrueCloudLab/frostfs-sdk-go
Evgenii Stratonikov
665e5807bc
For big objects with known size we can optimize allocation patterns
by providing size hint. As with any hint, it does not affect transformer
functionality: slices with capacity > MaxSize are never allocated.
```
goos: linux
goarch: amd64
pkg: git.frostfs.info/TrueCloudLab/frostfs-sdk-go/object/transformer
cpu: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
│ out │
│ sec/op │
Transformer/small/no_size_hint-8 65.44µ ± 3%
Transformer/small/no_size_hint,_with_buffer-8 64.24µ ± 5%
Transformer/small/with_size_hint,_with_buffer-8 58.70µ ± 5%
Transformer/big/no_size_hint-8 367.8m ± 3%
Transformer/big/no_size_hint,_with_buffer-8 562.7m ± 0%
Transformer/big/with_size_hint,_with_buffer-8 385.6m ± 7%
geomean 5.197m
│ out │
│ B/op │
Transformer/small/no_size_hint-8 13.40Ki ± 0%
Transformer/small/no_size_hint,_with_buffer-8 13.40Ki ± 0%
Transformer/small/with_size_hint,_with_buffer-8 13.39Ki ± 0%
Transformer/big/no_size_hint-8 288.0Mi ± 0%
Transformer/big/no_size_hint,_with_buffer-8 1.390Gi ± 0%
Transformer/big/with_size_hint,_with_buffer-8 288.0Mi ± 0%
geomean 2.533Mi
│ out │
│ allocs/op │
Transformer/small/no_size_hint-8 92.00 ± 0%
Transformer/small/no_size_hint,_with_buffer-8 92.00 ± 0%
Transformer/small/with_size_hint,_with_buffer-8 92.00 ± 0%
Transformer/big/no_size_hint-8 546.5 ± 0%
Transformer/big/no_size_hint,_with_buffer-8 607.5 ± 0%
Transformer/big/with_size_hint,_with_buffer-8 545.5 ± 0%
geomean 228.1
```
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
|
||
|---|---|---|
| .. | ||
| channel.go | ||
| channel_test.go | ||
| hasher.go | ||
| size_hint_test.go | ||
| transformer.go | ||
| transformer_test.go | ||
| types.go | ||