cyb70289 commented on pull request #11817:
URL: https://github.com/apache/arrow/pull/11817#issuecomment-983284341
There's some regression from `FloatParsing`. I repeated the test locally
with similar result.
Below result is tested on Xeon gold 5218. `FloatParsing` sees 25%~30%
regression for clang, 12% regression for gcc.
I also tested on Arm neoverse N1, the regression is about 10% for both
compilers.
**clang-12**
`$ archery benchmark diff --suite-filter="arrow-value-parsing-benchmark"
--cc=clang --cxx=clang++`
```
-----------------------------------------------------------------------------------------
Non-regressions: (23)
-----------------------------------------------------------------------------------------
benchmark baseline
contender change %
HexParsing<Int16Type> 122.889M items/sec 146.914M
items/sec 19.551
IntegerParsing<Int16Type> 122.648M items/sec 130.271M
items/sec 6.215
IntegerFormatting<Int64Type> 50.467M items/sec 53.040M
items/sec 5.097
......
---------------------------------------------------------------------------------------
Regressions: (10)
---------------------------------------------------------------------------------------
benchmark baseline
contender change %
IntegerParsing<UInt32Type> 158.139M items/sec 149.383M
items/sec -5.537
IntegerParsing<UInt8Type> 193.646M items/sec 181.939M
items/sec -6.046
IntegerParsing<Int32Type> 96.648M items/sec 87.516M
items/sec -9.449
HexParsing<UInt8Type> 154.437M items/sec 139.302M
items/sec -9.800
HexParsing<Int8Type> 159.359M items/sec 143.353M
items/sec -10.044
HexParsing<UInt32Type> 104.878M items/sec 90.750M
items/sec -13.470
TimestampParsingISO8601<TimeUnit::NANO> 44.332M items/sec 37.728M
items/sec -14.897
FloatParsing<FloatType> 52.520M items/sec 38.632M
items/sec -26.444
FloatParsing<DoubleType> 59.252M items/sec 41.705M
items/sec -29.613
HexParsing<UInt16Type> 122.306M items/sec 84.441M
items/sec -30.959
```
**gcc-9.4**
`$ archery benchmark diff --suite-filter="arrow-value-parsing-benchmark"
--cc=gcc --cxx=g++`
```
-----------------------------------------------------------------------------------------
Non-regressions: (28)
-----------------------------------------------------------------------------------------
benchmark baseline
contender change %
HexParsing<UInt32Type> 117.514M items/sec 131.796M
items/sec 12.153
HexParsing<UInt8Type> 191.946M items/sec 213.739M
items/sec 11.354
TimestampParsingISO8601<TimeUnit::SECOND> 38.486M items/sec 41.766M
items/sec 8.523
IntegerParsing<Int8Type> 136.544M items/sec 147.524M
items/sec 8.042
HexParsing<UInt16Type> 150.635M items/sec 158.778M
items/sec 5.406
IntegerFormatting<UInt8Type> 405.355M items/sec 426.245M
items/sec 5.153
......
-----------------------------------------------------------------------------
Regressions: (5)
-----------------------------------------------------------------------------
benchmark baseline contender change %
IntegerParsing<UInt32Type> 167.434M items/sec 157.774M items/sec -5.769
IntegerFormatting<UInt16Type> 183.618M items/sec 172.128M items/sec -6.257
HexParsing<Int16Type> 142.974M items/sec 130.157M items/sec -8.965
FloatParsing<DoubleType> 45.377M items/sec 39.540M items/sec -12.864
FloatParsing<FloatType> 43.444M items/sec 37.624M items/sec -13.397
```
--
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]
