pitrou commented on issue #42951:
URL: https://github.com/apache/arrow/issues/42951#issuecomment-3323605827
For reference, the current benchmark results on my machine (AMD Zen 2 CPU,
Ubuntu 24.04):
```
BM_PlainDecodingSpacedFloat/32768/1
3347 ns 3347 ns 198613 bytes_per_second=36.4711Gi/s
null_percent=0.01
BM_PlainDecodingSpacedFloat/32768/100
6131 ns 6130 ns 113455 bytes_per_second=19.9137Gi/s
null_percent=1
BM_PlainDecodingSpacedFloat/32768/1000
29780 ns 29776 ns 22998 bytes_per_second=4.09963Gi/s
null_percent=10
BM_PlainDecodingSpacedFloat/32768/5000
79986 ns 79971 ns 8705 bytes_per_second=1.52643Gi/s
null_percent=50
BM_PlainDecodingSpacedFloat/32768/10000
981 ns 981 ns 713115 bytes_per_second=124.382Gi/s
null_percent=100
BM_PlainDecodingSpacedDouble/32768/1
6623 ns 6622 ns 107008 bytes_per_second=36.8692Gi/s
null_percent=0.01
BM_PlainDecodingSpacedDouble/32768/100
10482 ns 10480 ns 68795 bytes_per_second=23.2951Gi/s
null_percent=1
BM_PlainDecodingSpacedDouble/32768/1000
33411 ns 33405 ns 20686 bytes_per_second=7.30841Gi/s
null_percent=10
BM_PlainDecodingSpacedDouble/32768/5000
86465 ns 86449 ns 8093 bytes_per_second=2.8241Gi/s
null_percent=50
BM_PlainDecodingSpacedDouble/32768/10000
2135 ns 2135 ns 327316 bytes_per_second=114.361Gi/s
null_percent=100
```
We see that when the null proportion is close to 0% or 100%, performance is
excellent. It's still relatively good when null proportion is in the middle.
(also, this randomly-generated data is really a worst case because nulls are
distributed evenly across the data, which might not be the case for real-world
data)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
