================================================================================================
BYTE_STREAM_SPLIT INT32
================================================================================================

OpenJDK 64-Bit Server VM 25.0.4+7-LTS on Linux 6.17.0-1020-azure
Intel(R) Xeon(R) 6973P-C
BYTE_STREAM_SPLIT INT32:                  Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------
Spark vectorized readIntegers                         2              2           1        678.5           1.5       1.0X
Spark vectorized readIntegersAsLongs                  2              2           0        658.0           1.5       1.0X
Spark vectorized readIntegersAsDoubles                2              2           0        644.8           1.6       1.0X
parquet-mr readInteger (per-value)                    5              6           1        205.7           4.9       0.3X


================================================================================================
BYTE_STREAM_SPLIT INT64
================================================================================================

OpenJDK 64-Bit Server VM 25.0.4+7-LTS on Linux 6.17.0-1020-azure
Intel(R) Xeon(R) 6973P-C
BYTE_STREAM_SPLIT INT64:                  Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------
Spark vectorized readLongs                            5              6           1        206.6           4.8       1.0X
Spark vectorized readLongsAsInts                      5              6           1        205.0           4.9       1.0X
parquet-mr readLong (per-value)                      10             11           1        105.1           9.5       0.5X


================================================================================================
BYTE_STREAM_SPLIT FLOAT
================================================================================================

OpenJDK 64-Bit Server VM 25.0.4+7-LTS on Linux 6.17.0-1020-azure
Intel(R) Xeon(R) 6973P-C
BYTE_STREAM_SPLIT FLOAT:                  Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------
Spark vectorized readFloats                           2              2           0        670.0           1.5       1.0X
Spark vectorized readFloatsAsDoubles                  2              2           0        638.0           1.6       1.0X
parquet-mr readFloat (per-value)                      5              6           1        206.5           4.8       0.3X


================================================================================================
BYTE_STREAM_SPLIT DOUBLE
================================================================================================

OpenJDK 64-Bit Server VM 25.0.4+7-LTS on Linux 6.17.0-1020-azure
Intel(R) Xeon(R) 6973P-C
BYTE_STREAM_SPLIT DOUBLE:                 Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
------------------------------------------------------------------------------------------------------------------------
Spark vectorized readDoubles                          5              6           1        207.2           4.8       1.0X
parquet-mr readDouble (per-value)                    10             11           1        105.2           9.5       0.5X


================================================================================================
BYTE_STREAM_SPLIT FLBA
================================================================================================

OpenJDK 64-Bit Server VM 25.0.4+7-LTS on Linux 6.17.0-1020-azure
Intel(R) Xeon(R) 6973P-C
BYTE_STREAM_SPLIT FLBA:                                  Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
---------------------------------------------------------------------------------------------------------------------------------------
Spark vectorized readBinary (width=12)                              14             16           1         74.6          13.4       1.0X
Spark per-value readBinary (old updater path, width=12)             17             19           4         61.3          16.3       0.8X
parquet-mr readBinary (per-value, width=12)                         24             28           4         43.2          23.2       0.6X