wpleonardo commented on PR #1375: URL: https://github.com/apache/orc/pull/1375#issuecomment-1757658298
> @wpleonardo I tried, but still find no improvement > > ``` > orc file(snappy + unaligned) + avx512 > 0 rows in set. Elapsed: 3.478 sec. Processed 1.20 million rows, 539.37 MB (345.98 thousand rows/s., 155.08 MB/s.) > 0 rows in set. Elapsed: 3.424 sec. Processed 1.20 million rows, 539.37 MB (351.44 thousand rows/s., 157.53 MB/s.) > 0 rows in set. Elapsed: 3.444 sec. Processed 1.20 million rows, 539.37 MB (349.44 thousand rows/s., 156.63 MB/s.) > > > orc file (snappy + unaligned) + none > 0 rows in set. Elapsed: 3.362 sec. Processed 1.20 million rows, 539.37 MB (357.89 thousand rows/s., 160.42 MB/s.) > 0 rows in set. Elapsed: 3.535 sec. Processed 1.20 million rows, 539.37 MB (340.43 thousand rows/s., 152.59 MB/s.) > 0 rows in set. Elapsed: 3.370 sec. Processed 1.20 million rows, 539.37 MB (357.08 thousand rows/s., 160.06 MB/s.) > > > orc file (lz4 + unaligned) + avx512 > 0 rows in set. Elapsed: 3.075 sec. Processed 1.20 million rows, 1.90 GB (391.26 thousand rows/s., 618.31 MB/s.) > 0 rows in set. Elapsed: 3.082 sec. Processed 1.20 million rows, 1.90 GB (390.46 thousand rows/s., 617.05 MB/s.) > 0 rows in set. Elapsed: 3.014 sec. Processed 1.20 million rows, 1.90 GB (399.18 thousand rows/s., 630.82 MB/s.) > > > orc file (lz4 + unaligned) + none > rows in set. Elapsed: 2.973 sec. Processed 1.20 million rows, 1.90 GB (404.76 thousand rows/s., 639.64 MB/s.) > 0 rows in set. Elapsed: 3.070 sec. Processed 1.20 million rows, 1.90 GB (391.90 thousand rows/s., 619.32 MB/s.) > 0 rows in set. Elapsed: 2.903 sec. Processed 1.20 million rows, 1.90 GB (414.51 thousand rows/s., 655.05 MB/s.) > ``` Could you do a simple test first, for example, just select the int64 column instead of all columns? -- 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]
