[SPARK-51045][TESTS][4.0] Regenerate benchmark results after upgrading to Scala 2.13.16 #49741
Conversation
dongjoon-hyun
changed the title
| AMD EPYC 7763 64-Core Processor | ||
| Test contains use empty Set: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative | ||
| ------------------------------------------------------------------------------------------------------------------------ | ||
| Use HashSet 3 3 0 291.9 3.4 1.0X | ||
| Use EnumSet 4 4 0 227.7 4.4 0.8X | ||
| Use HashSet 1 1 0 1390.7 0.7 1.0X |
There was a problem hiding this comment.
Although the ratio changes in both Java 17 and 21. The measure time is too small, 1ms. So, I believe we can ignore because it means it gets faster.
| Legacy 9203 9215 8 0.1 9203.3 1.0X | ||
| New 813 816 2 1.2 813.1 11.3X | ||
| Legacy 6862 6873 7 0.1 6861.7 1.0X | ||
| New 796 821 9 1.3 795.6 8.6X |
There was a problem hiding this comment.
Both becomes faster and New is still faster than Legacy.
| Spark 3987 3988 1 0.3 3986.6 1.2X | ||
| Spark Binary 2762 2766 3 0.4 2761.6 1.8X | ||
| Common Codecs 5040 5043 3 0.2 5039.8 1.0X | ||
| Java 3137 3141 6 0.3 3136.7 1.6X |
There was a problem hiding this comment.
Java becomes faster then Spark.
| Spark 3482 3488 10 0.3 3482.0 1.4X | ||
| Spark Binary 2638 2639 0 0.4 2638.3 1.9X | ||
| Common Codecs 4955 4983 43 0.2 4955.2 1.0X | ||
| Java 5790 5815 28 0.2 5790.1 0.9X |
There was a problem hiding this comment.
Java is still the slowest one in Java 21.
| SQL Json 8374 8384 14 1.9 532.4 1.2X | ||
| SQL Json with UnsafeRow 9509 9511 3 1.7 604.6 1.1X | ||
| SQL Parquet Vectorized: DataPageV1 83 93 6 189.4 5.3 123.2X | ||
| SQL Parquet Vectorized: DataPageV2 208 217 7 75.7 13.2 49.2X |
There was a problem hiding this comment.
Apparently, the ratio between DataPage1 and DataPageV2 is changed too big in this case.
| ConstantColumnVector 892 892 1 459.3 2.2 1.0X | ||
| OnHeapColumnVector 1020 1021 1 401.5 2.5 0.9X | ||
| OffHeapColumnVector 892 893 1 459.0 2.2 1.0X | ||
| ConstantColumnVector 0 0 0 13274135.5 0.0 1.0X |
There was a problem hiding this comment.
This looks strange because the value is 0.
There was a problem hiding this comment.
The 0s are strange. Do we also get 0s for master branch for this benchmark result?
| ConstantColumnVector 765 766 1 535.4 1.9 1.0X | ||
| OnHeapColumnVector 774 774 1 529.3 1.9 1.0X | ||
| OffHeapColumnVector 830 831 2 493.6 2.0 0.9X | ||
| ConstantColumnVector 0 0 0 3321170.8 0.0 1.0X |
There was a problem hiding this comment.
This also becomes 0.
|
There are a few places to spot, but in general, when we cross check both Java 17/21. There seems to be no regression in Scala 2.13.16. |
|
Could you review this PR when you have some time, @huaxingao ? |
|
To @huaxingao , yes, it happens even in this PR twice for
For the record, |
|
Thank you for spending your time here to help, @huaxingao ! |
…g to Scala 2.13.16 ### What changes were proposed in this pull request? This PR aims to regenerate benchmark results of `branch-4.0` after upgrading to Scala 2.13.16. - #49478 ### Why are the changes needed? To check a regression again - #49411 (comment) ### Does this PR introduce _any_ user-facing change? No, this updates only test results. ### How was this patch tested? Manual review. ### Was this patch authored or co-authored using generative AI tooling? No. Closes #49741 from dongjoon-hyun/bm_40. Authored-by: Dongjoon Hyun <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
|
Merged to branch-4.0. |
What changes were proposed in this pull request?
This PR aims to regenerate benchmark results of
branch-4.0after upgrading to Scala 2.13.16.Why are the changes needed?
To check a regression again
from_json#49411 (comment)Does this PR introduce any user-facing change?
No, this updates only test results.
How was this patch tested?
Manual review.
Was this patch authored or co-authored using generative AI tooling?
No.