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 2022/04/11 20:09:16 UTC

[GitHub] [spark] sunchao commented on a diff in pull request #36123: [SPARK-34863][SQL][FOLLOWUP] Add benchmark for Parquet & ORC nested column scan

sunchao commented on code in PR #36123:
URL: https://github.com/apache/spark/pull/36123#discussion_r847694826


##########
sql/core/benchmarks/DataSourceReadBenchmark-jdk11-results.txt:
##########
@@ -2,322 +2,430 @@
 SQL Single Numeric Column Scan
 ================================================================================================
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single BOOLEAN Column Scan:           Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           11809          12046         335          1.3         750.8       1.0X
-SQL Json                                           8588           8592           7          1.8         546.0       1.4X
-SQL Parquet Vectorized: DataPageV1                  140            162          18        112.0           8.9      84.1X
-SQL Parquet Vectorized: DataPageV2                  103            117          12        152.6           6.6     114.6X
-SQL Parquet MR: DataPageV1                         1634           1648          20          9.6         103.9       7.2X
-SQL Parquet MR: DataPageV2                         1495           1501           9         10.5          95.1       7.9X
-SQL ORC Vectorized                                  180            224          42         87.4          11.4      65.6X
-SQL ORC MR                                         1536           1576          57         10.2          97.7       7.7X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           11004          11065          86          1.4         699.6       1.0X
+SQL Json                                           7986           8011          35          2.0         507.7       1.4X
+SQL Parquet Vectorized: DataPageV1                  124            148          16        127.0           7.9      88.9X
+SQL Parquet Vectorized: DataPageV2                  101            115          12        155.0           6.5     108.4X
+SQL Parquet MR: DataPageV1                         1614           1620           8          9.7         102.6       6.8X
+SQL Parquet MR: DataPageV2                         1445           1446           2         10.9          91.9       7.6X
+SQL ORC Vectorized                                  163            204          41         96.2          10.4      67.3X
+SQL ORC MR                                         1407           1429          31         11.2          89.4       7.8X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single BOOLEAN Column Scan:   Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   109            114          10        144.3           6.9       1.0X
-ParquetReader Vectorized: DataPageV2                    90             93           3        175.3           5.7       1.2X
-ParquetReader Vectorized -> Row: DataPageV1             58             60           4        271.9           3.7       1.9X
-ParquetReader Vectorized -> Row: DataPageV2             39             41           3        404.0           2.5       2.8X
+ParquetReader Vectorized: DataPageV1                   123            140          14        128.3           7.8       1.0X
+ParquetReader Vectorized: DataPageV2                   105            114          11        150.3           6.7       1.2X
+ParquetReader Vectorized -> Row: DataPageV1             56             61           5        279.9           3.6       2.2X
+ParquetReader Vectorized -> Row: DataPageV2             39             43           4        399.4           2.5       3.1X
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single TINYINT Column Scan:           Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           14515          14526          16          1.1         922.8       1.0X
-SQL Json                                           9862           9863           2          1.6         627.0       1.5X
-SQL Parquet Vectorized: DataPageV1                  144            167          31        109.5           9.1     101.1X
-SQL Parquet Vectorized: DataPageV2                  139            159          27        113.4           8.8     104.6X
-SQL Parquet MR: DataPageV1                         1777           1780           3          8.8         113.0       8.2X
-SQL Parquet MR: DataPageV2                         1690           1691           2          9.3         107.4       8.6X
-SQL ORC Vectorized                                  201            238          46         78.3          12.8      72.2X
-SQL ORC MR                                         1513           1522          14         10.4          96.2       9.6X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           13262          13310          67          1.2         843.2       1.0X
+SQL Json                                           9104           9173          98          1.7         578.8       1.5X
+SQL Parquet Vectorized: DataPageV1                  136            172          31        115.4           8.7      97.3X
+SQL Parquet Vectorized: DataPageV2                  138            153          17        114.0           8.8      96.1X
+SQL Parquet MR: DataPageV1                         1789           1805          22          8.8         113.7       7.4X
+SQL Parquet MR: DataPageV2                         1631           1662          44          9.6         103.7       8.1X
+SQL ORC Vectorized                                  210            252          33         74.8          13.4      63.0X
+SQL ORC MR                                         1412           1437          36         11.1          89.7       9.4X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single TINYINT Column Scan:   Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   182            192          11         86.6          11.5       1.0X
-ParquetReader Vectorized: DataPageV2                   181            188           7         86.9          11.5       1.0X
-ParquetReader Vectorized -> Row: DataPageV1             96             99           4        163.3           6.1       1.9X
-ParquetReader Vectorized -> Row: DataPageV2             96             99           3        163.4           6.1       1.9X
+ParquetReader Vectorized: DataPageV1                   171            183          14         92.0          10.9       1.0X
+ParquetReader Vectorized: DataPageV2                   175            184           9         90.1          11.1       1.0X
+ParquetReader Vectorized -> Row: DataPageV1             88             95          12        179.0           5.6       1.9X
+ParquetReader Vectorized -> Row: DataPageV2             88             92           4        179.0           5.6       1.9X
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single SMALLINT Column Scan:          Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           15326          15437         156          1.0         974.4       1.0X
-SQL Json                                          10281          10290          13          1.5         653.7       1.5X
-SQL Parquet Vectorized: DataPageV1                  164            212          36         95.9          10.4      93.4X
-SQL Parquet Vectorized: DataPageV2                  230            244          11         68.5          14.6      66.7X
-SQL Parquet MR: DataPageV1                         2108           2111           4          7.5         134.0       7.3X
-SQL Parquet MR: DataPageV2                         1940           1963          33          8.1         123.3       7.9X
-SQL ORC Vectorized                                  229            279          34         68.7          14.6      66.9X
-SQL ORC MR                                         1903           1906           3          8.3         121.0       8.1X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           14022          14236         303          1.1         891.5       1.0X
+SQL Json                                           9763           9929         235          1.6         620.7       1.4X
+SQL Parquet Vectorized: DataPageV1                  173            226          38         90.7          11.0      80.9X
+SQL Parquet Vectorized: DataPageV2                  222            241          13         70.7          14.1      63.1X
+SQL Parquet MR: DataPageV1                         2069           2086          24          7.6         131.5       6.8X
+SQL Parquet MR: DataPageV2                         1771           1806          49          8.9         112.6       7.9X
+SQL ORC Vectorized                                  203            263          37         77.6          12.9      69.2X
+SQL ORC MR                                         1528           1552          34         10.3          97.2       9.2X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single SMALLINT Column Scan:  Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   253            262          10         62.2          16.1       1.0X
-ParquetReader Vectorized: DataPageV2                   323            327           9         48.8          20.5       0.8X
-ParquetReader Vectorized -> Row: DataPageV1            280            288           8         56.3          17.8       0.9X
-ParquetReader Vectorized -> Row: DataPageV2            301            314          21         52.2          19.1       0.8X
+ParquetReader Vectorized: DataPageV1                   246            256          11         63.9          15.6       1.0X
+ParquetReader Vectorized: DataPageV2                   301            313          17         52.3          19.1       0.8X
+ParquetReader Vectorized -> Row: DataPageV1            257            292          18         61.2          16.3       1.0X
+ParquetReader Vectorized -> Row: DataPageV2            296            318          25         53.1          18.8       0.8X
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single INT Column Scan:               Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           16756          16776          28          0.9        1065.3       1.0X
-SQL Json                                          10690          10692           3          1.5         679.6       1.6X
-SQL Parquet Vectorized: DataPageV1                  160            208          45         98.1          10.2     104.5X
-SQL Parquet Vectorized: DataPageV2                  390            423          23         40.3          24.8      43.0X
-SQL Parquet MR: DataPageV1                         2196           2201           8          7.2         139.6       7.6X
-SQL Parquet MR: DataPageV2                         2065           2072          10          7.6         131.3       8.1X
-SQL ORC Vectorized                                  323            338          10         48.7          20.5      51.9X
-SQL ORC MR                                         1899           1906          11          8.3         120.7       8.8X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           16153          16252         140          1.0        1027.0       1.0X
+SQL Json                                          10406          10547         200          1.5         661.6       1.6X
+SQL Parquet Vectorized: DataPageV1                  159            207          33         99.1          10.1     101.8X
+SQL Parquet Vectorized: DataPageV2                  337            402          40         46.6          21.4      47.9X
+SQL Parquet MR: DataPageV1                         2160           2193          46          7.3         137.4       7.5X
+SQL Parquet MR: DataPageV2                         1892           1900          11          8.3         120.3       8.5X
+SQL ORC Vectorized                                  297            340          42         53.0          18.9      54.5X
+SQL ORC MR                                         1705           1732          38          9.2         108.4       9.5X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single INT Column Scan:       Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   278            285           9         56.6          17.7       1.0X
-ParquetReader Vectorized: DataPageV2                   514            518           2         30.6          32.7       0.5X
-ParquetReader Vectorized -> Row: DataPageV1            308            316          11         51.0          19.6       0.9X
-ParquetReader Vectorized -> Row: DataPageV2            498            525          27         31.6          31.6       0.6X
+ParquetReader Vectorized: DataPageV1                   251            262          10         62.6          16.0       1.0X
+ParquetReader Vectorized: DataPageV2                   418            431          13         37.7          26.6       0.6X
+ParquetReader Vectorized -> Row: DataPageV1            247            288          30         63.7          15.7       1.0X
+ParquetReader Vectorized -> Row: DataPageV2            412            455          39         38.1          26.2       0.6X
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single BIGINT Column Scan:            Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           21841          21851          14          0.7        1388.6       1.0X
-SQL Json                                          12828          12843          21          1.2         815.6       1.7X
-SQL Parquet Vectorized: DataPageV1                  241            279          19         65.2          15.3      90.6X
-SQL Parquet Vectorized: DataPageV2                  554            596          29         28.4          35.2      39.5X
-SQL Parquet MR: DataPageV1                         2404           2428          34          6.5         152.8       9.1X
-SQL Parquet MR: DataPageV2                         2153           2166          18          7.3         136.9      10.1X
-SQL ORC Vectorized                                  417            464          62         37.7          26.5      52.4X
-SQL ORC MR                                         2136           2146          14          7.4         135.8      10.2X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           19896          20026         183          0.8        1264.9       1.0X
+SQL Json                                          12540          12634         132          1.3         797.3       1.6X
+SQL Parquet Vectorized: DataPageV1                  221            271          30         71.3          14.0      90.1X
+SQL Parquet Vectorized: DataPageV2                  546            564          23         28.8          34.7      36.5X
+SQL Parquet MR: DataPageV1                         2196           2211          21          7.2         139.6       9.1X
+SQL Parquet MR: DataPageV2                         2085           2089           6          7.5         132.5       9.5X
+SQL ORC Vectorized                                  379            416          39         41.5          24.1      52.5X
+SQL ORC MR                                         1858           1859           2          8.5         118.1      10.7X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single BIGINT Column Scan:    Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   324            357          34         48.6          20.6       1.0X
-ParquetReader Vectorized: DataPageV2                   694            702          11         22.6          44.2       0.5X
-ParquetReader Vectorized -> Row: DataPageV1            378            385           8         41.6          24.0       0.9X
-ParquetReader Vectorized -> Row: DataPageV2            701            708           8         22.4          44.6       0.5X
+ParquetReader Vectorized: DataPageV1                   311            340          20         50.5          19.8       1.0X
+ParquetReader Vectorized: DataPageV2                   639            647          11         24.6          40.6       0.5X
+ParquetReader Vectorized -> Row: DataPageV1            359            376          13         43.9          22.8       0.9X
+ParquetReader Vectorized -> Row: DataPageV2            653            658           9         24.1          41.5       0.5X
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single FLOAT Column Scan:             Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           17238          17239           2          0.9        1096.0       1.0X
-SQL Json                                          12295          12307          18          1.3         781.7       1.4X
-SQL Parquet Vectorized: DataPageV1                  162            203          27         96.8          10.3     106.1X
-SQL Parquet Vectorized: DataPageV2                  157            194          32        100.4          10.0     110.0X
-SQL Parquet MR: DataPageV1                         2163           2165           3          7.3         137.5       8.0X
-SQL Parquet MR: DataPageV2                         2014           2014           1          7.8         128.0       8.6X
-SQL ORC Vectorized                                  458            462           5         34.4          29.1      37.7X
-SQL ORC MR                                         1984           1984           0          7.9         126.1       8.7X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           17338          17710         526          0.9        1102.3       1.0X
+SQL Json                                          11844          12121         392          1.3         753.0       1.5X
+SQL Parquet Vectorized: DataPageV1                  148            187          28        106.2           9.4     117.0X
+SQL Parquet Vectorized: DataPageV2                  147            183          31        106.8           9.4     117.7X
+SQL Parquet MR: DataPageV1                         2027           2033           9          7.8         128.9       8.6X
+SQL Parquet MR: DataPageV2                         1966           1981          21          8.0         125.0       8.8X
+SQL ORC Vectorized                                  399            425          25         39.4          25.4      43.4X
+SQL ORC MR                                         1748           1756          11          9.0         111.2       9.9X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single FLOAT Column Scan:     Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   252            259          10         62.3          16.0       1.0X
-ParquetReader Vectorized: DataPageV2                   252            256           9         62.3          16.0       1.0X
-ParquetReader Vectorized -> Row: DataPageV1            259            307          40         60.7          16.5       1.0X
-ParquetReader Vectorized -> Row: DataPageV2            260            295          25         60.5          16.5       1.0X
+ParquetReader Vectorized: DataPageV1                   226            240          15         69.6          14.4       1.0X
+ParquetReader Vectorized: DataPageV2                   225            237          15         69.9          14.3       1.0X
+ParquetReader Vectorized -> Row: DataPageV1            247            299          38         63.6          15.7       0.9X
+ParquetReader Vectorized -> Row: DataPageV2            245            296          25         64.1          15.6       0.9X
 
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 SQL Single DOUBLE Column Scan:            Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-SQL CSV                                           22485          22536          72          0.7        1429.5       1.0X
-SQL Json                                          16281          16286           8          1.0        1035.1       1.4X
-SQL Parquet Vectorized: DataPageV1                  232            288          35         67.9          14.7      97.1X
-SQL Parquet Vectorized: DataPageV2                  277            290           9         56.8          17.6      81.2X
-SQL Parquet MR: DataPageV1                         2331           2341          15          6.7         148.2       9.6X
-SQL Parquet MR: DataPageV2                         2216           2229          18          7.1         140.9      10.1X
-SQL ORC Vectorized                                  561            569           9         28.0          35.7      40.1X
-SQL ORC MR                                         2118           2137          27          7.4         134.6      10.6X
-
-OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.11.0-1028-azure
-Intel(R) Xeon(R) Platinum 8272CL CPU @ 2.60GHz
+SQL CSV                                           21052          21617         799          0.7        1338.4       1.0X
+SQL Json                                          15822          16049         321          1.0        1005.9       1.3X
+SQL Parquet Vectorized: DataPageV1                  266            286          19         59.0          16.9      79.0X
+SQL Parquet Vectorized: DataPageV2                  277            291          14         56.8          17.6      76.0X
+SQL Parquet MR: DataPageV1                         2267           2275          12          6.9         144.1       9.3X
+SQL Parquet MR: DataPageV2                         2046           2064          26          7.7         130.1      10.3X
+SQL ORC Vectorized                                  535            545          10         29.4          34.0      39.3X
+SQL ORC MR                                         1976           2000          34          8.0         125.6      10.7X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
 Parquet Reader Single DOUBLE Column Scan:    Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ---------------------------------------------------------------------------------------------------------------------------
-ParquetReader Vectorized: DataPageV1                   355            356           1         44.3          22.6       1.0X
-ParquetReader Vectorized: DataPageV2                   355            356           1         44.3          22.6       1.0X
-ParquetReader Vectorized -> Row: DataPageV1            379            386           9         41.5          24.1       0.9X
-ParquetReader Vectorized -> Row: DataPageV2            379            389          10         41.5          24.1       0.9X
+ParquetReader Vectorized: DataPageV1                   314            337          25         50.1          20.0       1.0X
+ParquetReader Vectorized: DataPageV2                   309            323          14         50.8          19.7       1.0X
+ParquetReader Vectorized -> Row: DataPageV1            331            348          13         47.5          21.1       0.9X
+ParquetReader Vectorized -> Row: DataPageV2            332            347          11         47.4          21.1       0.9X
+
+
+================================================================================================
+SQL Single Numeric Column Scan in Struct
+================================================================================================
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
+SQL Single TINYINT Column Scan in Struct:                    Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
+-------------------------------------------------------------------------------------------------------------------------------------------
+SQL ORC MR                                                            2193           2196           4          7.2         139.5       1.0X
+SQL ORC Vectorized (Nested Column Disabled)                           2211           2222          16          7.1         140.6       1.0X
+SQL ORC Vectorized (Nested Column Enabled)                             268            310          32         58.7          17.0       8.2X
+SQL Parquet MR: DataPageV1                                            2243           2280          53          7.0         142.6       1.0X
+SQL Parquet Vectorized: DataPageV1 (Nested Column Disabled)           2747           2758          16          5.7         174.6       0.8X
+SQL Parquet Vectorized: DataPageV1 (Nested Column Enabled)             155            174          22        101.7           9.8      14.2X
+SQL Parquet MR: DataPageV2                                            2193           2203          13          7.2         139.5       1.0X
+SQL Parquet Vectorized: DataPageV2 (Nested Column Disabled)           2709           2733          33          5.8         172.3       0.8X
+SQL Parquet Vectorized: DataPageV2 (Nested Column Enabled)             150            174          27        104.7           9.6      14.6X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
+SQL Single SMALLINT Column Scan in Struct:                   Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
+-------------------------------------------------------------------------------------------------------------------------------------------
+SQL ORC MR                                                            2322           2391          97          6.8         147.6       1.0X
+SQL ORC Vectorized (Nested Column Disabled)                           2362           2374          17          6.7         150.2       1.0X
+SQL ORC Vectorized (Nested Column Enabled)                             412            419           9         38.2          26.2       5.6X
+SQL Parquet MR: DataPageV1                                            2393           2400          10          6.6         152.1       1.0X
+SQL Parquet Vectorized: DataPageV1 (Nested Column Disabled)           2919           2922           4          5.4         185.6       0.8X
+SQL Parquet Vectorized: DataPageV1 (Nested Column Enabled)             228            281          54         69.0          14.5      10.2X
+SQL Parquet MR: DataPageV2                                            2223           2240          25          7.1         141.3       1.0X
+SQL Parquet Vectorized: DataPageV2 (Nested Column Disabled)           2692           2712          28          5.8         171.2       0.9X
+SQL Parquet Vectorized: DataPageV2 (Nested Column Enabled)             341            361          31         46.1          21.7       6.8X
+
+OpenJDK 64-Bit Server VM 11.0.14+9-LTS on Linux 5.13.0-1021-azure
+Intel(R) Xeon(R) Platinum 8171M CPU @ 2.60GHz
+SQL Single INT Column Scan in Struct:                        Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
+-------------------------------------------------------------------------------------------------------------------------------------------
+SQL ORC MR                                                            2376           2380           6          6.6         151.0       1.0X
+SQL ORC Vectorized (Nested Column Disabled)                           2333           2378          64          6.7         148.4       1.0X
+SQL ORC Vectorized (Nested Column Enabled)                             430            451          20         36.6          27.3       5.5X
+SQL Parquet MR: DataPageV1                                            2485           2501          22          6.3         158.0       1.0X
+SQL Parquet Vectorized: DataPageV1 (Nested Column Disabled)           3017           3062          65          5.2         191.8       0.8X

Review Comment:
   I'm not sure. This part is calling the row-based Parquet reader from `parquet-mr` which I'm not very familiar with. Overall its performance is not very good.



-- 
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