You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2022/01/14 21:38:12 UTC
[GitHub] [spark] dongjoon-hyun commented on a change in pull request #35212: [SPARK-37912][SQL][TESTS] Add benchmark for vectorized delta binary packed encoding
dongjoon-hyun commented on a change in pull request #35212:
URL: https://github.com/apache/spark/pull/35212#discussion_r785184866
##########
File path: sql/core/benchmarks/DataSourceReadBenchmark-jdk17-results.txt
##########
@@ -2,269 +2,283 @@
SQL Single Numeric Column Scan
================================================================================================
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
SQL Single BOOLEAN Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-SQL CSV 11737 11812 106 1.3 746.2 1.0X
-SQL Json 7827 7904 109 2.0 497.6 1.5X
-SQL Parquet Vectorized 98 116 12 160.6 6.2 119.8X
-SQL Parquet MR 1529 1541 18 10.3 97.2 7.7X
-SQL ORC Vectorized 165 185 14 95.5 10.5 71.2X
-SQL ORC MR 1433 1440 9 11.0 91.1 8.2X
-
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV 12462 12576 162 1.3 792.3 1.0X
+SQL Json 7548 7554 8 2.1 479.9 1.7X
+SQL Parquet Vectorized 101 115 11 156.1 6.4 123.7X
+SQL Parquet MR 1989 2008 26 7.9 126.5 6.3X
+SQL Parquet Vectorized (Delta Binary) 68 77 7 230.5 4.3 182.6X
+SQL Parquet MR (Delta Binary) 1767 1767 0 8.9 112.3 7.1X
+SQL ORC Vectorized 154 160 8 102.3 9.8 81.1X
+SQL ORC MR 1543 1554 14 10.2 98.1 8.1X
+
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Parquet Reader Single BOOLEAN Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
-------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized 38 40 3 416.2 2.4 1.0X
-ParquetReader Vectorized -> Row 38 39 3 419.1 2.4 1.0X
+ParquetReader Vectorized 43 45 1 361.6 2.8 1.0X
+ParquetReader Vectorized -> Row 44 45 1 355.8 2.8 1.0X
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
SQL Single TINYINT Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-SQL CSV 13156 13192 51 1.2 836.4 1.0X
-SQL Json 8690 8784 133 1.8 552.5 1.5X
-SQL Parquet Vectorized 196 207 8 80.4 12.4 67.2X
-SQL Parquet MR 1831 1834 4 8.6 116.4 7.2X
-SQL ORC Vectorized 157 167 7 100.2 10.0 83.8X
-SQL ORC MR 1381 1387 8 11.4 87.8 9.5X
-
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV 14759 14802 61 1.1 938.4 1.0X
+SQL Json 8993 8994 1 1.7 571.8 1.6X
+SQL Parquet Vectorized 100 107 7 158.0 6.3 148.3X
+SQL Parquet MR 1969 1974 7 8.0 125.2 7.5X
+SQL Parquet Vectorized (Delta Binary) 98 104 5 160.9 6.2 150.9X
+SQL Parquet MR (Delta Binary) 1802 1814 18 8.7 114.5 8.2X
+SQL ORC Vectorized 152 156 4 103.2 9.7 96.9X
+SQL ORC MR 1419 1424 6 11.1 90.2 10.4X
+
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Parquet Reader Single TINYINT Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
-------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized 147 153 6 107.0 9.3 1.0X
-ParquetReader Vectorized -> Row 149 162 24 105.7 9.5 1.0X
+ParquetReader Vectorized 61 63 1 256.4 3.9 1.0X
+ParquetReader Vectorized -> Row 62 64 2 254.4 3.9 1.0X
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
SQL Single SMALLINT Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-SQL CSV 14024 14291 378 1.1 891.6 1.0X
-SQL Json 9777 9849 102 1.6 621.6 1.4X
-SQL Parquet Vectorized 153 175 18 102.9 9.7 91.8X
-SQL Parquet MR 1971 1979 11 8.0 125.3 7.1X
-SQL ORC Vectorized 193 211 15 81.4 12.3 72.5X
-SQL ORC MR 1665 1693 39 9.4 105.9 8.4X
-
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV 15622 15629 9 1.0 993.2 1.0X
+SQL Json 9446 9452 9 1.7 600.5 1.7X
+SQL Parquet Vectorized 128 135 8 123.0 8.1 122.1X
+SQL Parquet MR 2190 2211 30 7.2 139.2 7.1X
+SQL Parquet Vectorized (Delta Binary) 188 193 4 83.6 12.0 83.1X
+SQL Parquet MR (Delta Binary) 1932 1951 28 8.1 122.8 8.1X
+SQL ORC Vectorized 198 204 4 79.6 12.6 79.1X
+SQL ORC MR 1567 1571 6 10.0 99.6 10.0X
+
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
Parquet Reader Single SMALLINT Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
--------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized 217 227 7 72.6 13.8 1.0X
-ParquetReader Vectorized -> Row 214 216 2 73.5 13.6 1.0X
+ParquetReader Vectorized 192 195 2 82.0 12.2 1.0X
+ParquetReader Vectorized -> Row 188 202 12 83.6 12.0 1.0X
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1025-azure
+Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz
SQL Single INT Column Scan: Best Time(ms) Avg Time(ms) Stdev(ms) Rate(M/s) Per Row(ns) Relative
------------------------------------------------------------------------------------------------------------------------
-SQL CSV 15107 15205 139 1.0 960.5 1.0X
-SQL Json 9699 9773 104 1.6 616.7 1.6X
-SQL Parquet Vectorized 144 160 24 109.6 9.1 105.2X
-SQL Parquet MR 1903 1906 4 8.3 121.0 7.9X
-SQL ORC Vectorized 227 234 6 69.4 14.4 66.6X
-SQL ORC MR 1566 1578 17 10.0 99.5 9.6X
-
-OpenJDK 64-Bit Server VM 17.0.1+12-LTS on Linux 5.11.0-1022-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV 17258 17391 189 0.9 1097.2 1.0X
+SQL Json 9964 9968 6 1.6 633.5 1.7X
+SQL Parquet Vectorized 122 159 27 129.2 7.7 141.7X
+SQL Parquet MR 2193 2222 40 7.2 139.5 7.9X
+SQL Parquet Vectorized (Delta Binary) 297 323 26 52.9 18.9 58.1X
Review comment:
ditto. This looks like roughly 3 times slower than before.
--
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: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org