feat: Create benchmarks comet cast by coderfender · Pull Request #2932 · apache/datafusion-comet

coderfender · 2025-12-18T00:47:43Z

Which issue does this PR close?

Create mechanism to benchmark CometCast operation with ANSI support (Similar to CometAggregatedBenchmark) to have a baseline understanding of native cast performance

Closes #.

Rationale for this change

What changes are included in this PR?

How are these changes tested?

codecov-commenter · 2025-12-18T01:36:15Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 59.63%. Comparing base (f09f8af) to head (ed9d181).
⚠️ Report is 782 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #2932      +/-   ##
============================================
+ Coverage     56.12%   59.63%   +3.50%     
- Complexity      976     1379     +403     
============================================
  Files           119      167      +48     
  Lines         11743    15459    +3716     
  Branches       2251     2559     +308     
============================================
+ Hits           6591     9219    +2628     
- Misses         4012     4946     +934     
- Partials       1140     1294     +154

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

coderfender · 2025-12-19T01:03:34Z

================================================================================================
Running benchmark BooleanType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : BooleanType , ansi mode enabled : false:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : BOOLEAN , ansi mode enabled : false            172            186          14         60.9          16.4       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : BOOLEAN , ansi mode enabled : false            224            253          22         46.7          21.4       0.8X


================================================================================================
Running benchmark BooleanType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : BooleanType , ansi mode enabled : true:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : BOOLEAN , ansi mode enabled : true            170            190          23         61.7          16.2       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : BOOLEAN , ansi mode enabled : true            211            228          12         49.7          20.1       0.8X


================================================================================================
Running benchmark ByteType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : ByteType , ansi mode enabled : false:            Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : TINYINT , ansi mode enabled : false            174            216          23         60.2          16.6       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : TINYINT , ansi mode enabled : false            213            237          21         49.2          20.3       0.8X


================================================================================================
Running benchmark ByteType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : ByteType , ansi mode enabled : true:            Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : TINYINT , ansi mode enabled : true             35             57          23        298.3           3.4       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : TINYINT , ansi mode enabled : true             27             55          27        393.5           2.5       1.3X


================================================================================================
Running benchmark ShortType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : ShortType , ansi mode enabled : false:            Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : SMALLINT , ansi mode enabled : false            175            186          12         60.0          16.7       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : SMALLINT , ansi mode enabled : false            198            223          16         52.9          18.9       0.9X


================================================================================================
Running benchmark ShortType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : ShortType , ansi mode enabled : true:            Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : SMALLINT , ansi mode enabled : true             40             56          17        261.4           3.8       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : SMALLINT , ansi mode enabled : true             27             56          20        394.4           2.5       1.5X


================================================================================================
Running benchmark IntegerType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : IntegerType , ansi mode enabled : false:     Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
--------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : INT , ansi mode enabled : false            174            186          16         60.2          16.6       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : INT , ansi mode enabled : false            203            241          24         51.5          19.4       0.9X


================================================================================================
Running benchmark IntegerType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : IntegerType , ansi mode enabled : true:     Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : INT , ansi mode enabled : true             40             58          21        263.3           3.8       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : INT , ansi mode enabled : true             24             59          27        429.0           2.3       1.6X


================================================================================================
Running benchmark LongType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : LongType , ansi mode enabled : false:           Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : BIGINT , ansi mode enabled : false            175            206          17         59.9          16.7       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : BIGINT , ansi mode enabled : false            163            206          29         64.5          15.5       1.1X


================================================================================================
Running benchmark LongType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : LongType , ansi mode enabled : true:           Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : BIGINT , ansi mode enabled : true            175            193          15         60.0          16.7       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : BIGINT , ansi mode enabled : true            155            197          36         67.6          14.8       1.1X


================================================================================================
Running benchmark FloatType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : FloatType , ansi mode enabled : false:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : FLOAT , ansi mode enabled : false            174            183          12         60.4          16.6       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : FLOAT , ansi mode enabled : false            176            243          29         59.5          16.8       1.0X


================================================================================================
Running benchmark FloatType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : FloatType , ansi mode enabled : true:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
---------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : FLOAT , ansi mode enabled : true            177            199           8         59.3          16.8       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : FLOAT , ansi mode enabled : true            181            250          27         57.9          17.3       1.0X


================================================================================================
Running benchmark DoubleType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : DoubleType , ansi mode enabled : false:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : DOUBLE , ansi mode enabled : false            175            186          12         60.1          16.7       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : DOUBLE , ansi mode enabled : false            225            257          15         46.7          21.4       0.8X


================================================================================================
Running benchmark DoubleType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : DoubleType , ansi mode enabled : true:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : DOUBLE , ansi mode enabled : true            186            213          23         56.5          17.7       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : DOUBLE , ansi mode enabled : true            189            245          32         55.5          18.0       1.0X


================================================================================================
Running benchmark DecimalType(10,2))
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : DecimalType(10,2) , ansi mode enabled : false:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : DECIMAL(10,2) , ansi mode enabled : false            543            559          19         19.3          51.8       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : DECIMAL(10,2) , ansi mode enabled : false            285            331          36         36.8          27.2       1.9X


================================================================================================
Running benchmark DecimalType(10,2))
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : DecimalType(10,2) , ansi mode enabled : true:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : DECIMAL(10,2) , ansi mode enabled : true             33             59          14        317.3           3.2       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : DECIMAL(10,2) , ansi mode enabled : true             24             59          35        433.6           2.3       1.4X


================================================================================================
Running benchmark StringType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : StringType , ansi mode enabled : false:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
-----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : STRING , ansi mode enabled : false            567            568           1         18.5          54.1       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : STRING , ansi mode enabled : false            485            514          36         21.6          46.3       1.2X


================================================================================================
Running benchmark StringType)
================================================================================================

OpenJDK 64-Bit Server VM 17.0.16+8-LTS on Mac OS X 16.0
Apple M2 Max
Cast function from : LongType to : StringType , ansi mode enabled : true:         Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
----------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL Parquet - Spark Cast expr from BIGINT to : STRING , ansi mode enabled : true            570            582          21         18.4          54.4       1.0X
SQL Parquet - Comet Cast expr from BIGINT to : STRING , ansi mode enabled : true            445            477          33         23.6          42.4       1.3X

coderfender · 2025-12-20T00:01:03Z

Addressed review comments

andygrove

Thanks @coderfender

coderfender · 2025-12-20T03:57:42Z

Thank you for the approval @andygrove

B Vadlamani added 2 commits December 17, 2025 16:15

comet_cast_init

633aca7

comet_cast_benchmark_suite

03b59a6

andygrove reviewed Dec 18, 2025

View reviewed changes