Dandandan commented on PR #21830: URL: https://github.com/apache/datafusion/pull/21830#issuecomment-4338819858
> This seems like a reasonable change to me if we see a meaningful difference in the benchmarks. But the results I see look mixed / minor? The TPCH-10 results look pretty good (see below). I think it mainly has effect on (larger) partitioned hash joins. I am going to run https://github.com/apache/datafusion/pull/21900 ``` Comparing HEAD and repartition-fastrange -------------------- Benchmark tpch_sf10.json -------------------- ┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓ ┃ Query ┃ HEAD ┃ repartition-fastrange ┃ Change ┃ ┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩ │ QQuery 1 │ 325.71 / 327.48 ±1.16 / 328.72 ms │ 325.07 / 326.22 ±0.74 / 327.03 ms │ no change │ │ QQuery 2 │ 133.72 / 135.69 ±1.69 / 138.03 ms │ 123.54 / 125.22 ±1.37 / 127.08 ms │ +1.08x faster │ │ QQuery 3 │ 289.39 / 295.52 ±4.76 / 304.05 ms │ 268.85 / 276.86 ±4.28 / 280.85 ms │ +1.07x faster │ │ QQuery 4 │ 155.34 / 157.92 ±3.80 / 165.46 ms │ 136.18 / 137.76 ±1.05 / 139.23 ms │ +1.15x faster │ │ QQuery 5 │ 422.60 / 427.12 ±3.99 / 434.57 ms │ 385.64 / 390.16 ±3.51 / 394.93 ms │ +1.09x faster │ │ QQuery 6 │ 133.01 / 134.42 ±1.49 / 137.19 ms │ 132.00 / 132.76 ±0.41 / 133.26 ms │ no change │ │ QQuery 7 │ 529.51 / 542.42 ±8.17 / 555.27 ms │ 493.39 / 502.99 ±6.48 / 510.73 ms │ +1.08x faster │ │ QQuery 8 │ 463.27 / 467.15 ±3.80 / 472.78 ms │ 432.67 / 435.82 ±2.93 / 439.85 ms │ +1.07x faster │ │ QQuery 9 │ 640.69 / 651.23 ±9.02 / 664.10 ms │ 601.04 / 610.93 ±6.84 / 622.08 ms │ +1.07x faster │ │ QQuery 10 │ 326.78 / 335.39 ±4.77 / 340.55 ms │ 315.74 / 330.89 ±10.93 / 345.75 ms │ no change │ │ QQuery 11 │ 106.64 / 108.39 ±1.49 / 110.96 ms │ 93.22 / 96.80 ±2.88 / 100.71 ms │ +1.12x faster │ │ QQuery 12 │ 201.34 / 207.36 ±6.13 / 218.89 ms │ 191.39 / 194.88 ±4.48 / 203.52 ms │ +1.06x faster │ │ QQuery 13 │ 300.56 / 309.36 ±5.89 / 318.53 ms │ 295.81 / 303.60 ±6.75 / 312.45 ms │ no change │ │ QQuery 14 │ 181.08 / 185.28 ±5.31 / 195.70 ms │ 179.48 / 183.64 ±5.18 / 193.42 ms │ no change │ │ QQuery 15 │ 329.96 / 332.66 ±2.07 / 336.26 ms │ 324.95 / 330.09 ±3.89 / 335.19 ms │ no change │ │ QQuery 16 │ 78.74 / 80.38 ±1.58 / 83.07 ms │ 72.58 / 78.30 ±6.51 / 90.58 ms │ no change │ │ QQuery 17 │ 743.07 / 748.14 ±5.14 / 757.76 ms │ 681.43 / 695.49 ±9.96 / 711.85 ms │ +1.08x faster │ │ QQuery 18 │ 757.33 / 781.48 ±16.94 / 801.12 ms │ 702.59 / 724.97 ±15.98 / 752.39 ms │ +1.08x faster │ │ QQuery 19 │ 268.01 / 279.33 ±12.32 / 301.73 ms │ 265.86 / 277.68 ±20.81 / 319.27 ms │ no change │ │ QQuery 20 │ 308.04 / 317.20 ±9.03 / 333.08 ms │ 296.19 / 316.59 ±14.29 / 338.65 ms │ no change │ │ QQuery 21 │ 808.99 / 817.11 ±7.90 / 827.32 ms │ 737.60 / 741.17 ±3.71 / 747.63 ms │ +1.10x faster │ │ QQuery 22 │ 80.85 / 83.74 ±2.62 / 88.20 ms │ 74.27 / 75.77 ±1.06 / 77.40 ms │ +1.11x faster │ └───────────┴────────────────────────────────────┴────────────────────────────────────┴───────────────┘ ┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓ ┃ Benchmark Summary ┃ ┃ ┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩ │ Total Time (HEAD) │ 7724.76ms │ │ Total Time (repartition-fastrange) │ 7288.57ms │ │ Average Time (HEAD) │ 351.13ms │ │ Average Time (repartition-fastrange) │ 331.30ms │ │ Queries Faster │ 13 │ │ Queries Slower │ 0 │ │ Queries with No Change │ 9 │ │ Queries with Failure │ 0 │ └──────────────────────────────────────┴───────────┘ ``` -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
