Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
waynexia commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4496880730 @codex review -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461331751 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4461199921) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark clickbench_partitioned.json ┏━━━┳━━━┳━━━┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇━━━╇━━━┩ │ QQuery 0 │ 1.23 / 4.76 ±6.92 / 18.59 ms │ 1.25 / 4.82 ±6.98 / 18.78 ms │ no change │ │ QQuery 1 │12.89 / 13.07 ±0.11 / 13.18 ms │12.70 / 13.12 ±0.24 / 13.37 ms │ no change │ │ QQuery 2 │36.34 / 36.86 ±0.36 / 37.29 ms │35.87 / 36.27 ±0.29 / 36.73 ms │ no change │ │ QQuery 3 │31.56 / 32.94 ±1.11 / 34.87 ms │31.81 / 32.26 ±0.38 / 32.77 ms │ no change │ │ QQuery 4 │ 240.65 / 245.56 ±2.93 / 249.81 ms │216.21 / 230.63 ±12.53 / 247.06 ms │ +1.06x faster │ │ QQuery 5 │ 287.84 / 290.17 ±1.33 / 291.97 ms │ 285.30 / 288.68 ±2.04 / 290.65 ms │ no change │ │ QQuery 6 │ 7.36 / 8.72 ±2.32 / 13.35 ms │ 7.48 / 7.96 ±0.32 / 8.38 ms │ +1.10x faster │ │ QQuery 7 │13.89 / 14.07 ±0.11 / 14.21 ms │14.14 / 14.36 ±0.20 / 14.71 ms │ no change │ │ QQuery 8 │ 325.36 / 330.01 ±2.78 / 332.85 ms │ 330.90 / 334.63 ±2.43 / 337.57 ms │ no change │ │ QQuery 9 │ 468.59 / 474.20 ±4.86 / 482.09 ms │ 461.37 / 471.01 ±5.91 / 478.72 ms │ no change │ │ QQuery 10 │71.79 / 72.31 ±0.49 / 72.93 ms │71.33 / 73.37 ±1.21 / 74.68 ms │ no change │ │ QQuery 11 │82.86 / 84.07 ±1.48 / 86.81 ms │81.88 / 83.69 ±2.48 / 88.56 ms │ no change │ │ QQuery 12 │ 286.67 / 290.60 ±3.29 / 295.31 ms │ 282.99 / 286.23 ±3.33 / 292.55 ms │ no change │ │ QQuery 13 │ 396.94 / 404.33 ±5.35 / 412.23 ms │392.21 / 404.07 ±11.13 / 423.66 ms │ no change │ │ QQuery 14 │288.16 / 297.60 ±11.73 / 319.92 ms │ 289.71 / 294.47 ±2.88 / 297.46 ms │ no change │ │ QQuery 15 │ 290.10 / 2
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461331710 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4461199760) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark clickbench_partitioned.json ┏━━━┳━━━┳━━━┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇━━━╇━━━┩ │ QQuery 0 │ 1.22 / 4.78 ±6.97 / 18.72 ms │ 1.26 / 4.86 ±6.98 / 18.83 ms │ no change │ │ QQuery 1 │12.62 / 13.14 ±0.26 / 13.30 ms │12.62 / 12.73 ±0.07 / 12.84 ms │ no change │ │ QQuery 2 │35.57 / 36.19 ±0.63 / 37.35 ms │35.36 / 35.68 ±0.19 / 35.90 ms │ no change │ │ QQuery 3 │31.76 / 32.70 ±0.88 / 34.11 ms │30.57 / 30.94 ±0.36 / 31.62 ms │ +1.06x faster │ │ QQuery 4 │ 243.39 / 248.77 ±3.63 / 253.57 ms │215.12 / 224.49 ±13.17 / 250.11 ms │ +1.11x faster │ │ QQuery 5 │ 290.34 / 292.86 ±1.75 / 295.41 ms │ 276.76 / 284.27 ±8.49 / 296.02 ms │ no change │ │ QQuery 6 │ 6.80 / 7.26 ±0.44 / 8.06 ms │ 7.46 / 7.80 ±0.26 / 8.15 ms │ 1.07x slower │ │ QQuery 7 │13.80 / 13.92 ±0.08 / 14.02 ms │14.33 / 14.58 ±0.31 / 15.08 ms │ no change │ │ QQuery 8 │ 332.63 / 335.70 ±2.83 / 340.65 ms │ 311.18 / 316.39 ±8.45 / 333.16 ms │ +1.06x faster │ │ QQuery 9 │464.09 / 475.89 ±10.56 / 489.86 ms │449.41 / 467.17 ±16.94 / 494.58 ms │ no change │ │ QQuery 10 │71.55 / 72.44 ±0.73 / 73.38 ms │70.73 / 71.41 ±0.77 / 72.87 ms │ no change │ │ QQuery 11 │80.97 / 82.35 ±0.93 / 83.74 ms │80.91 / 84.27 ±5.65 / 95.54 ms │ no change │ │ QQuery 12 │ 279.52 / 287.73 ±5.64 / 296.66 ms │270.69 / 283.13 ±13.30 / 300.92 ms │ no change │ │ QQuery 13 │397.90 / 410.36 ±10.45 / 425.89 ms │382.53 / 405.13 ±18.04 / 427.38 ms │ no change │ │ QQuery 14 │ 281.95 / 288.38 ±5.44 / 296.12 ms │ 292.10 / 297.68 ±4.96 / 307.00 ms │ no change │ │ QQuery 15 │ 281.93 / 2
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461330085 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4461199476) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark clickbench_partitioned.json ┏━━━┳━━━┳━━━┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇━━━╇━━━┩ │ QQuery 0 │ 1.20 / 4.71 ±6.88 / 18.47 ms │ 1.19 / 4.72 ±6.90 / 18.53 ms │ no change │ │ QQuery 1 │12.38 / 12.87 ±0.28 / 13.21 ms │12.44 / 12.86 ±0.27 / 13.21 ms │ no change │ │ QQuery 2 │35.67 / 35.96 ±0.42 / 36.79 ms │35.58 / 35.79 ±0.29 / 36.36 ms │ no change │ │ QQuery 3 │31.27 / 31.70 ±0.56 / 32.67 ms │30.66 / 31.03 ±0.26 / 31.31 ms │ no change │ │ QQuery 4 │ 234.24 / 239.69 ±3.43 / 244.76 ms │ 216.31 / 226.03 ±8.31 / 238.97 ms │ +1.06x faster │ │ QQuery 5 │ 280.91 / 283.29 ±1.59 / 285.20 ms │ 281.10 / 282.15 ±1.45 / 284.99 ms │ no change │ │ QQuery 6 │ 6.50 / 7.11 ±0.50 / 7.61 ms │ 6.02 / 6.81 ±0.68 / 7.82 ms │ no change │ │ QQuery 7 │13.82 / 14.75 ±1.61 / 17.96 ms │13.75 / 13.90 ±0.10 / 14.04 ms │ +1.06x faster │ │ QQuery 8 │ 317.14 / 321.46 ±2.44 / 324.73 ms │ 316.39 / 319.13 ±1.65 / 321.36 ms │ no change │ │ QQuery 9 │ 451.72 / 458.89 ±5.69 / 466.39 ms │ 457.81 / 463.01 ±3.20 / 466.31 ms │ no change │ │ QQuery 10 │70.90 / 71.34 ±0.40 / 71.86 ms │69.99 / 71.28 ±0.74 / 72.28 ms │ no change │ │ QQuery 11 │82.12 / 82.83 ±0.42 / 83.39 ms │80.82 / 82.81 ±2.81 / 88.41 ms │ no change │ │ QQuery 12 │ 276.10 / 280.50 ±3.10 / 284.56 ms │ 275.38 / 283.46 ±6.75 / 292.93 ms │ no change │ │ QQuery 13 │ 386.72 / 394.42 ±8.23 / 409.41 ms │ 391.69 / 396.27 ±4.01 / 403.78 ms │ no change │ │ QQuery 14 │ 281.94 / 286.41 ±4.90 / 295.37 ms │284.65 / 292.80 ±11.96 / 316.44 ms │ no change │ │ QQuery 15 │ 278.97 / 2
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461222138 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4461199476) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4461199476-146-xrlgw 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: clickbench_partitioned Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461223378 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4461199760) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4461199760-148-2qfd9 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: clickbench_partitioned Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461222334 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4461199921) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4461199921-147-v2d9c 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: clickbench_partitioned Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461199760 run benchmark clickbench_partitioned -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461199921 run benchmark clickbench_partitioned -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4461199476 run benchmark clickbench_partitioned -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4458365838 > rerun another round and those regression disappear 😼 Some of the clickbench queries can be noisy as sometimes they can prune out more IO / decoding due to tightening filters of topk / min / max being non-deterministic (sometimes it gets a better bound earlier by chance). I think they probably are not related at all to this code. -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
waynexia commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4457859499 rerun another round and those regression disappear 😼 -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
waynexia commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4457644142 I rerun it use an aws c6a.4xlarge instance. Improvement on q4 is reproduced, but still has lots of regressions and some of them seem to be real. I'll verify them on my local environment next Details ``` Comparing HEAD and q4 Benchmark clickbench_partitioned.json | Query | HEAD | q4 |Change | |---|||---| | QQuery 0 | 1.36 / 4.94 ±7.12 / 19.19 ms | 1.34 / 4.96 ±7.17 / 19.31 ms | no change | | QQuery 1 | 16.62 / 19.50 ±5.12 / 29.72 ms | 16.31 / 18.93 ±4.47 / 27.86 ms | no change | | QQuery 2 |49.02 / 56.98 ±13.18 / 83.28 ms |49.78 / 59.22 ±12.82 / 84.56 ms | no change | | QQuery 3 |53.17 / 413.82 ±718.84 / 1851.51 ms |59.77 / 344.50 ±565.86 / 1476.22 ms | 1.12x slower | | QQuery 4 | 554.10 / 563.18 ±8.72 / 578.19 ms |439.19 / 580.67 ±129.14 / 769.30 ms | +1.26x faster | | QQuery 5 | 516.88 / 862.42 ±659.98 / 2182.05 ms | 520.97 / 985.10 ±801.77 / 2584.76 ms | no change | | QQuery 6 |3.21 / 3.61 ±0.66 / 4.94 ms | 3.22 / 3.63 ±0.50 / 4.60 ms | no change | | QQuery 7 | 17.74 / 17.93 ±0.21 / 18.32 ms | 17.66 / 17.87 ±0.20 / 18.11 ms | no change | | QQuery 8 | 686.29 / 696.36 ±5.83 / 703.40 ms | 754.21 / 1027.06 ±161.68 / 1242.22 ms | 1.10x slower | | QQuery 9 | 728.66 / 774.36 ±46.60 / 853.92 ms | 757.98 / 773.33 ±27.82 / 828.94 ms | no change | | QQuery 10 | 131.60 / 136.60 ±3.78 / 142.90 ms | 132.43 / 143.93 ±16.42 / 176.37 ms | no change | | QQuery 11 | 154.82 / 157.99 ±3.68 / 164.70 ms | 154.22 / 157.11 ±2.27 / 160.61 ms | no change | | QQuery 12 | 556.54 / 580.66 ±16.57 / 600.00 ms | 572.20 / 581.34 ±5.28 / 585.67 ms | no change | | QQuery 13 | 910.76 / 925.92 ±11.40 / 942.69 ms | 917.47 / 1089.09 ±312.06 / 1711.62 ms | no change | | QQuery 14 | 592.23 / 607.12 ±16.48 / 629.28 ms | 589.55 / 604.00 ±15.24 / 632.15 ms | no change | | QQuery 15 | 660.31 / 668.69 ±7.57 / 680.85 ms | 660.65 / 698.46 ±57.26 / 811.65 ms | no change | | QQuery 16 | 1379.80 / 1539.88 ±180.32 / 1779.23 ms | 1886.79 / 2404.33 ±589.74 / 3522.30 ms | 1.37x slower | | QQuery 17 | 1354.58 / 1382.99 ±16.74 / 1399.85 ms | 1384.44 / 1492.35 ±176.97 / 1845.19 ms | no change | | QQuery 18 | 3025.36 / 3872.99 ±884.61 / 5536.77 ms |3199.56 / 4494.24 ±1372.71 / 7031.04 ms | 1.06x slower | | QQuery 19 | 49.55 / 81.69 ±61.68 / 205.05 ms | 52.03 / 132.06 ±99.55 / 287.15 ms | no change | | QQuery 20 |819.85 / 4346.66 ±7038.47 / 18423.59 ms |802.50 / 4260.38 ±6900.87 / 18062.11 ms | no change | | QQuery 21 | 1000.17 / 1016.61 ±10.11 / 1031.33 ms | 1010.43 / 1068.85 ±45.61 / 1122.61 ms | no change | | QQuery 22 | 2156.11 / 4996.83 ±5654.14 / 16305.09 ms | 2167.60 / 4980.96 ±5607.84 / 16196.63 ms | no change | | QQuery 23 | 8584.84 / 24247.23 ±20717.36 / 60235.10 ms | 8519.57 / 17462.29 ±13228.65 / 42692.78 ms | no change | | QQuery 24 | 74.87 / 79.44 ±3.41 / 85.18 ms | 72.72 / 81.10 ±6.23 / 88.68 ms | no change | | QQuery 25 | 178.34 / 183.73 ±7.06 / 197.70 ms | 169.92 / 173.68 ±2.34 / 177.19 ms | no change | | QQuery 26 | 71.38 / 72.57 ±1.21 / 74.72 ms | 73.72 / 75.13 ±1.16 / 76.90 ms | no change | | QQuery 27 | 1192.07 / 1217.48 ±20.19 / 1242.67 ms | 1167.96 / 1187.36 ±21.54 / 1224.20 ms | no change | | QQuery 28 | 3712.95 / 3738.42 ±30.59 / 3797.21 ms | 3625.79 / 3669.24 ±67.10 / 3801.25 ms | no change | | QQuery 29 | 53.46 / 65.66 ±21.75 / 108.99 ms | 55.63 / 59.62 ±5.05 / 68.61 ms | no change | | QQuery 30 | 567.78 / 580.97 ±8.15 / 589.39 ms | 574.09 / 579.27 ±3.67 / 584.37 ms | no change | | QQuery 31 | 703.23 / 716.17 ±19.01 / 753.59 ms | 704.20 / 714.94 ±6.34 / 721.92 ms | no change | | QQuery 32 | 2717.58 / 3374.99 ±571.57 / 4437.74 ms | 3369.42 / 3940.48 ±562.65 / 4949.74 ms | 1.24x slower | | QQuery 33 | 3161.04 / 3230.02 ±69.97 / 3363.47 ms | 3168.24 / 3310.67 ±178.75 / 3659.44 ms | no change | | QQuery 34 | 3207.08 / 3409.06 ±234.38 / 3788.10 ms | 3190.32 / 3281.
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4450179783 > ``` > │ QQuery 40 │13.86 / 14.30 ±0.31 / 14.65 ms │13.83 / 17.12 ±3.59 / 22.13 ms │ 1.20x slower │ > │ QQuery 42 │13.11 / 13.56 ±0.63 / 14.79 ms │12.73 / 16.60 ±7.03 / 30.64 ms │ 1.22x slower │ > ``` > > Some queries regress a lot... This is likely noise, best is to look at the lowest reported times in those cases. -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
waynexia commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449382845 ``` │ QQuery 40 │13.86 / 14.30 ±0.31 / 14.65 ms │13.83 / 17.12 ±3.59 / 22.13 ms │ 1.20x slower │ │ QQuery 42 │13.11 / 13.56 ±0.63 / 14.79 ms │12.73 / 16.60 ±7.03 / 30.64 ms │ 1.22x slower │ ``` Some queries regress a lot... -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449218936 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4449065667) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark clickbench_extended.json ┏━━━┳━━━┳━━━┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇━━━╇━━━┩ │ QQuery 0 │ 786.54 / 797.82 ±6.35 / 805.19 ms │791.70 / 806.98 ±12.49 / 821.08 ms │ no change │ │ QQuery 1 │ 186.79 / 187.19 ±0.32 / 187.60 ms │ 187.98 / 190.58 ±3.69 / 197.92 ms │ no change │ │ QQuery 2 │ 457.81 / 460.77 ±2.79 / 465.83 ms │ 459.38 / 461.64 ±1.38 / 463.23 ms │ no change │ │ QQuery 3 │ 307.80 / 308.83 ±0.94 / 310.50 ms │ 312.75 / 314.05 ±1.03 / 314.98 ms │ no change │ │ QQuery 4 │ 666.17 / 673.37 ±8.65 / 689.97 ms │665.11 / 675.12 ±11.04 / 695.84 ms │ no change │ │ QQuery 5 │ 10396.71 / 10611.51 ±134.12 / 10771.39 ms │ 10253.91 / 10471.92 ±174.88 / 10786.78 ms │ no change │ │ QQuery 6 │ 998.24 / 1010.63 ±13.40 / 1034.10 ms │ 998.57 / 1051.73 ±73.05 / 1194.45 ms │ no change │ │ QQuery 7 │729.77 / 749.10 ±12.93 / 770.08 ms │737.80 / 763.53 ±16.01 / 787.16 ms │ no change │ │ QQuery 8 │377.14 / 401.10 ±40.03 / 481.02 ms │374.73 / 386.89 ±13.48 / 405.40 ms │ no change │ │ QQuery 9 │ 2841.62 / 2925.22 ±61.99 / 3031.73 ms │ 2805.11 / 2947.75 ±76.26 / 3022.04 ms │ no change │ │ QQuery 10 │632.15 / 652.21 ±19.94 / 687.11 ms │633.44 / 666.65 ±27.04 / 708.89 ms │ no change │ │ QQuery 11 │ 2374.98 / 2393.55 ±18.22 / 2419.52 ms │ 2214.87 / 2241.66 ±17.95 / 2267.82 ms │ +1.07x faster │ │ QQuery 12 │186.51 / 207.94 ±30.34 / 268.00 ms │189.17 / 207.94 ±32.86 / 273.60 ms │ no change │ │ QQuery 13 │550.13 / 571.96 ±29.38 / 629.55 ms │ 554.15 / 560.94 ±6.80 / 571.02 ms │ no change │ └───
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449184157 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4449065667) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark clickbench_partitioned.json ┏━━━┳━━━┳━━━┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇━━━╇━━━┩ │ QQuery 0 │ 1.20 / 4.57 ±6.68 / 17.93 ms │ 1.22 / 4.55 ±6.62 / 17.79 ms │ no change │ │ QQuery 1 │12.20 / 12.51 ±0.23 / 12.79 ms │12.53 / 12.73 ±0.12 / 12.89 ms │ no change │ │ QQuery 2 │35.18 / 36.25 ±1.07 / 38.21 ms │35.34 / 35.60 ±0.23 / 35.96 ms │ no change │ │ QQuery 3 │30.34 / 31.24 ±0.82 / 32.75 ms │30.33 / 30.57 ±0.23 / 30.95 ms │ no change │ │ QQuery 4 │ 228.33 / 231.92 ±2.33 / 235.39 ms │ 219.18 / 228.03 ±6.79 / 235.14 ms │ no change │ │ QQuery 5 │ 276.24 / 279.14 ±2.32 / 282.16 ms │ 274.67 / 276.65 ±1.31 / 278.13 ms │ no change │ │ QQuery 6 │ 5.92 / 6.70 ±0.64 / 7.73 ms │ 6.39 / 7.06 ±0.64 / 7.89 ms │ 1.05x slower │ │ QQuery 7 │13.69 / 13.90 ±0.29 / 14.45 ms │13.86 / 14.07 ±0.19 / 14.34 ms │ no change │ │ QQuery 8 │ 308.78 / 311.83 ±1.84 / 314.29 ms │ 308.36 / 311.51 ±1.96 / 314.23 ms │ no change │ │ QQuery 9 │ 442.03 / 449.85 ±6.87 / 461.74 ms │ 440.53 / 447.78 ±8.15 / 461.74 ms │ no change │ │ QQuery 10 │70.02 / 70.75 ±0.61 / 71.48 ms │69.82 / 70.15 ±0.21 / 70.43 ms │ no change │ │ QQuery 11 │80.95 / 81.82 ±0.49 / 82.29 ms │80.31 / 82.69 ±3.62 / 89.88 ms │ no change │ │ QQuery 12 │ 267.99 / 272.24 ±3.01 / 277.36 ms │ 268.09 / 271.76 ±2.52 / 275.04 ms │ no change │ │ QQuery 13 │ 382.56 / 389.75 ±8.05 / 405.28 ms │ 376.60 / 380.50 ±2.63 / 384.07 ms │ no change │ │ QQuery 14 │ 276.50 / 279.68 ±5.51 / 290.64 ms │ 276.23 / 279.34 ±2.19 / 282.87 ms │ no change │ │ QQuery 15 │ 272.62 / 2
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on code in PR #22145:
URL: https://github.com/apache/datafusion/pull/22145#discussion_r3240189863
##
datafusion/physical-plan/src/aggregates/group_values/single_group_by/primitive.rs:
##
@@ -79,147 +79,334 @@ hash_float!(f16, f32, f64);
///
/// This specialization is significantly faster than using the more general
/// purpose `Row`s format
+enum GroupValuesPrimitiveState {
+GroupIds {
+/// Stores the `(group_index, hash)` based on the hash of its value
+///
+/// We also store `hash` is for reducing cost of rehashing. Such cost
+/// is obvious in high cardinality group by situation.
+/// More details can see:
+///
+map: HashTable<(usize, u64)>,
Review Comment:
I found `HashTable<(usize, T::Native)>` also to be a bit faster than this
current approach (and also saving memory). We can build the `Vec`
when emitting data.
Could be part of some future PR.
--
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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449088660 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4449065667) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4449065667-82-5w7wp 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: clickbench_extended Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449086767 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4449065667) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4449065667-81-7rccx 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: clickbench_partitioned Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449065667 run benchmark clickbench_partitioned clickbench_extended -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4449064007 ``` │ QQuery 4 │ 232.65 / 234.21 ±1.59 / 236.52 ms │ 213.86 / 215.85 ±1.58 / 218.49 ms │ +1.09x faster │ ``` Seems like the query is faster though (and memory is ~halved) so seems like a great optimization - I expect it for higher cardinality Benchmark runner might have a bit more cache and thus might not hit RAM, etc. -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
waynexia commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448846680 Hmmm, the result does not seem good 🤔 Let me dig deeper -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448817345 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark clickbench_partitioned.json ┏━━━┳━━━┳━━━┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇━━━╇━━━┩ │ QQuery 0 │ 1.20 / 4.69 ±6.86 / 18.40 ms │ 1.21 / 4.71 ±6.87 / 18.45 ms │ no change │ │ QQuery 1 │12.48 / 12.86 ±0.25 / 13.26 ms │12.45 / 12.95 ±0.26 / 13.23 ms │ no change │ │ QQuery 2 │35.38 / 35.85 ±0.33 / 36.13 ms │35.75 / 36.02 ±0.27 / 36.53 ms │ no change │ │ QQuery 3 │30.36 / 30.99 ±0.81 / 32.58 ms │30.72 / 30.88 ±0.19 / 31.23 ms │ no change │ │ QQuery 4 │ 232.65 / 234.21 ±1.59 / 236.52 ms │ 213.86 / 215.85 ±1.58 / 218.49 ms │ +1.09x faster │ │ QQuery 5 │ 279.97 / 283.25 ±2.42 / 286.26 ms │ 277.39 / 279.05 ±1.95 / 281.47 ms │ no change │ │ QQuery 6 │ 6.59 / 7.11 ±0.33 / 7.56 ms │ 7.02 / 7.25 ±0.22 / 7.61 ms │ no change │ │ QQuery 7 │13.73 / 13.78 ±0.05 / 13.87 ms │13.73 / 13.81 ±0.12 / 14.04 ms │ no change │ │ QQuery 8 │ 316.64 / 321.16 ±4.22 / 328.38 ms │ 313.53 / 318.65 ±4.98 / 327.97 ms │ no change │ │ QQuery 9 │ 451.14 / 456.79 ±5.49 / 467.00 ms │ 451.46 / 456.03 ±3.45 / 460.95 ms │ no change │ │ QQuery 10 │69.19 / 69.88 ±0.41 / 70.33 ms │69.58 / 70.36 ±0.68 / 71.55 ms │ no change │ │ QQuery 11 │81.29 / 81.89 ±0.63 / 82.76 ms │80.36 / 81.29 ±0.59 / 81.84 ms │ no change │ │ QQuery 12 │ 273.58 / 276.74 ±3.33 / 282.70 ms │ 271.47 / 277.61 ±5.58 / 287.44 ms │ no change │ │ QQuery 13 │ 380.50 / 390.57 ±9.15 / 407.47 ms │ 376.60 / 385.16 ±8.29 / 399.18 ms │ no change │ │ QQuery 14 │ 279.39 / 281.34 ±1.95 / 284.38 ms │ 275.76 / 280.05 ±3.48 / 285.77 ms │ no change │ │ QQuery 15 │ 277.20 / 2
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448801710 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark tpcds_sf1.json ┏━━━┳━━━┳┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇━━━╇╇━━━┩ │ QQuery 1 │ 6.24 / 6.75 ±0.84 / 8.44 ms │6.28 / 6.76 ±0.83 / 8.42 ms │ no change │ │ QQuery 2 │81.98 / 82.54 ±0.29 / 82.77 ms │ 82.42 / 83.36 ±0.93 / 84.67 ms │ no change │ │ QQuery 3 │28.90 / 29.39 ±0.29 / 29.78 ms │ 29.19 / 29.56 ±0.26 / 29.96 ms │ no change │ │ QQuery 4 │531.79 / 540.08 ±11.18 / 562.05 ms │ 516.91 / 542.94 ±25.90 / 575.22 ms │ no change │ │ QQuery 5 │54.13 / 54.60 ±0.36 / 55.22 ms │ 54.69 / 55.45 ±0.73 / 56.83 ms │ no change │ │ QQuery 6 │37.33 / 37.59 ±0.18 / 37.82 ms │ 36.73 / 37.51 ±0.53 / 38.15 ms │ no change │ │ QQuery 7 │ 112.42 / 114.08 ±1.51 / 115.91 ms │ 110.35 / 113.40 ±2.12 / 115.56 ms │ no change │ │ QQuery 8 │40.61 / 40.92 ±0.27 / 41.34 ms │ 40.53 / 41.24 ±0.38 / 41.64 ms │ no change │ │ QQuery 9 │52.35 / 54.97 ±1.71 / 57.63 ms │ 54.01 / 55.36 ±0.85 / 56.61 ms │ no change │ │ QQuery 10 │81.69 / 83.30 ±1.46 / 85.78 ms │ 81.74 / 83.38 ±1.77 / 86.31 ms │ no change │ │ QQuery 11 │321.59 / 353.39 ±17.77 / 371.79 ms │ 315.49 / 349.39 ±19.12 / 369.07 ms │ no change │ │ QQuery 12 │28.80 / 29.29 ±0.62 / 30.48 ms │ 29.16 / 29.39 ±0.23 / 29.77 ms │ no change │ │ QQuery 13 │ 132.08 / 133.43 ±1.05 / 134.89 ms │ 132.51 / 133.41 ±0.67 / 134.28 ms │ no change │ │ QQuery 14 │ 521.23 / 523.63 ±1.79 / 525.73 ms │ 521.41 / 524.74 ±2.07 / 527.16 ms │ no change │ │ QQuery 15 │65.87 / 67.23 ±0.84 / 68.19 ms │ 66.33 / 67.08 ±0.92 / 68.75 ms │ no change │ │ QQuery 16 │
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448786692 🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Details ``` Comparing HEAD and q4 Benchmark tpch_sf1.json ┏━━━┳┳┳━━━┓ ┃ Query ┃ HEAD ┃ q4 ┃Change ┃ ┡━━━╇╇╇━━━┩ │ QQuery 1 │ 39.13 / 40.83 ±1.23 / 42.17 ms │ 39.11 / 39.75 ±0.93 / 41.59 ms │ no change │ │ QQuery 2 │ 20.69 / 21.03 ±0.32 / 21.42 ms │ 20.40 / 20.71 ±0.28 / 21.12 ms │ no change │ │ QQuery 3 │ 35.67 / 37.57 ±1.02 / 38.63 ms │ 34.95 / 35.68 ±0.64 / 36.86 ms │ +1.05x faster │ │ QQuery 4 │ 17.39 / 17.50 ±0.12 / 17.73 ms │ 17.21 / 17.65 ±0.48 / 18.58 ms │ no change │ │ QQuery 5 │ 42.64 / 45.30 ±1.37 / 46.57 ms │ 43.14 / 44.66 ±1.26 / 46.38 ms │ no change │ │ QQuery 6 │ 16.56 / 17.52 ±0.89 / 19.12 ms │ 16.57 / 16.63 ±0.07 / 16.75 ms │ +1.05x faster │ │ QQuery 7 │ 49.69 / 50.56 ±0.54 / 51.37 ms │ 50.47 / 52.59 ±1.32 / 53.75 ms │ no change │ │ QQuery 8 │ 46.10 / 46.32 ±0.14 / 46.52 ms │ 46.08 / 46.45 ±0.33 / 47.03 ms │ no change │ │ QQuery 9 │ 51.16 / 52.22 ±0.68 / 53.21 ms │ 51.90 / 52.10 ±0.18 / 52.40 ms │ no change │ │ QQuery 10 │ 64.31 / 64.53 ±0.12 / 64.65 ms │ 64.06 / 64.29 ±0.16 / 64.55 ms │ no change │ │ QQuery 11 │ 13.47 / 14.11 ±0.83 / 15.72 ms │ 13.51 / 13.97 ±0.61 / 15.17 ms │ no change │ │ QQuery 12 │ 24.86 / 25.15 ±0.25 / 25.61 ms │ 25.30 / 25.62 ±0.19 / 25.84 ms │ no change │ │ QQuery 13 │ 35.89 / 36.81 ±0.88 / 38.05 ms │ 35.09 / 35.82 ±0.54 / 36.54 ms │ no change │ │ QQuery 14 │ 25.87 / 26.75 ±1.24 / 29.21 ms │ 25.78 / 26.05 ±0.18 / 26.34 ms │ no change │ │ QQuery 15 │ 31.76 / 32.39 ±0.63 / 33.23 ms │ 31.47 / 31.93 ±0.48 / 32.73 ms │ no change │ │ QQuery 16 │ 15.01 / 15.16 ±0.14 / 15.40 ms │ 15.18 / 15.31 ±0.11 / 15.51 ms │ no change │ │ QQuery 17 │ 74.55 / 76.87 ±1.59 / 78.89 ms │ 75.82 / 77.14 ±1.57 / 79.58 ms │ no change │ │ QQuery 18 │ 69.33 / 71.14 ±2.04 / 74.95 ms │ 69.82 / 70.70 ±1.15 / 72.94 ms │ no change
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448699138 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4448673356-78-lmv8g 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: clickbench_partitioned Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448700954 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4448673356-79-zc7ct 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: tpcds Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448694970 🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356) **Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4448673356-80-zdxf9 6.12.68+ #1 SMP Wed Apr 1 02:23:28 UTC 2026 aarch64 GNU/Linux` CPU Details (lscpu) ``` Architecture:aarch64 CPU op-mode(s): 64-bit Byte Order: Little Endian CPU(s): 16 On-line CPU(s) list: 0-15 Vendor ID: ARM Model name: Neoverse-V2 Model: 1 Thread(s) per core: 1 Core(s) per cluster: 16 Socket(s): - Cluster(s): 1 Stepping:r0p1 BogoMIPS:2000.00 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti L1d cache: 1 MiB (16 instances) L1i cache: 1 MiB (16 instances) L2 cache:32 MiB (16 instances) L3 cache:80 MiB (1 instance) NUMA node(s):1 NUMA node0 CPU(s): 0-15 Vulnerability Gather data sampling: Not affected Vulnerability Indirect target selection: Not affected Vulnerability Itlb multihit: Not affected Vulnerability L1tf: Not affected Vulnerability Mds: Not affected Vulnerability Meltdown: Not affected Vulnerability Mmio stale data: Not affected Vulnerability Reg file data sampling:Not affected Vulnerability Retbleed: Not affected Vulnerability Spec rstack overflow: Not affected Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl Vulnerability Spectre v1:Mitigation; __user pointer sanitization Vulnerability Spectre v2:Mitigation; CSV2, BHB Vulnerability Srbds: Not affected Vulnerability Tsa: Not affected Vulnerability Tsx async abort: Not affected Vulnerability Vmscape: Not affected ``` Comparing q4 (65980f0637f7d62b28d8bed9770b3ce92051e666) to 7f2f78d (merge-base) [diff](https://github.com/apache/datafusion/compare/7f2f78d48b6d3d6aee2ce2fd29910bb4c11b1012..65980f0637f7d62b28d8bed9770b3ce92051e666) using: tpch Results will be posted here when complete --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448673356 run benchmarks -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448556915 Hi @waynexia, thanks for the request (https://github.com/apache/datafusion/pull/22145#issuecomment-4448556762). Only whitelisted users can trigger benchmarks. Allowed users: [Dandandan](https://github.com/Dandandan), [Fokko](https://github.com/Fokko), [Jefffrey](https://github.com/Jefffrey), [Omega359](https://github.com/Omega359), [adriangb](https://github.com/adriangb), [alamb](https://github.com/alamb), [asubiotto](https://github.com/asubiotto), [brunal](https://github.com/brunal), [buraksenn](https://github.com/buraksenn), [cetra3](https://github.com/cetra3), [codephage2020](https://github.com/codephage2020), [coderfender](https://github.com/coderfender), [comphead](https://github.com/comphead), [erenavsarogullari](https://github.com/erenavsarogullari), [etseidl](https://github.com/etseidl), [friendlymatthew](https://github.com/friendlymatthew), [gabotechs](https://github.com/gabotechs), [geoffreyclaude](https://github.com/geoffreyclaude), [grtlr](https://github.com/grtlr), [h aohuaijin](https://github.com/haohuaijin), [jonathanc-n](https://github.com/jonathanc-n), [kevinjqliu](https://github.com/kevinjqliu), [klion26](https://github.com/klion26), [kosiew](https://github.com/kosiew), [kumarUjjawal](https://github.com/kumarUjjawal), [kunalsinghdadhwal](https://github.com/kunalsinghdadhwal), [liamzwbao](https://github.com/liamzwbao), [mbutrovich](https://github.com/mbutrovich), [mkleen](https://github.com/mkleen), [mzabaluev](https://github.com/mzabaluev), [neilconway](https://github.com/neilconway), [rluvaton](https://github.com/rluvaton), [sdf-jkl](https://github.com/sdf-jkl), [timsaucer](https://github.com/timsaucer), [xudong963](https://github.com/xudong963), [zhuqi-lucas](https://github.com/zhuqi-lucas). --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
waynexia commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448556762 run benchmarks -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448525486 Benchmark for [this request](https://github.com/apache/datafusion/pull/22145#issuecomment-4448508770) failed. Last 20 lines of output: Click to expand ``` * [new tag] 48.0.0-rc1 -> 48.0.0-rc1 * [new tag] 48.0.0-rc2 -> 48.0.0-rc2 * [new tag] 5.0.0-> 5.0.0 * [new tag] 5.0.0-rc1-> 5.0.0-rc1 * [new tag] 5.0.0-rc3-> 5.0.0-rc3 * [new tag] 6.0.0-> 6.0.0 * [new tag] 6.0.0-rc0-> 6.0.0-rc0 * [new tag] 7.0.0-> 7.0.0 * [new tag] 7.0.0-rc2-> 7.0.0-rc2 * [new tag] 8.0.0-> 8.0.0 * [new tag] 8.0.0-rc1-> 8.0.0-rc1 * [new tag] 8.0.0-rc2-> 8.0.0-rc2 * [new tag] 9.0.0-> 9.0.0 * [new tag] 9.0.0-rc1-> 9.0.0-rc1 * [new tag] ballista-0.5.0 -> ballista-0.5.0 * [new tag] ballista-0.6.0 -> ballista-0.6.0 * [new tag] ballista-0.7.0 -> ballista-0.7.0 * [new tag] python-0.3.0 -> python-0.3.0 * [new tag] python-0.4.0 -> python-0.4.0 Switched to branch 'q4' ``` --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448524681 Benchmark for [this request](https://github.com/apache/datafusion/pull/22145#issuecomment-4448508770) failed. Last 20 lines of output: Click to expand ``` * [new tag] 48.0.0-rc1 -> 48.0.0-rc1 * [new tag] 48.0.0-rc2 -> 48.0.0-rc2 * [new tag] 5.0.0-> 5.0.0 * [new tag] 5.0.0-rc1-> 5.0.0-rc1 * [new tag] 5.0.0-rc3-> 5.0.0-rc3 * [new tag] 6.0.0-> 6.0.0 * [new tag] 6.0.0-rc0-> 6.0.0-rc0 * [new tag] 7.0.0-> 7.0.0 * [new tag] 7.0.0-rc2-> 7.0.0-rc2 * [new tag] 8.0.0-> 8.0.0 * [new tag] 8.0.0-rc1-> 8.0.0-rc1 * [new tag] 8.0.0-rc2-> 8.0.0-rc2 * [new tag] 9.0.0-> 9.0.0 * [new tag] 9.0.0-rc1-> 9.0.0-rc1 * [new tag] ballista-0.5.0 -> ballista-0.5.0 * [new tag] ballista-0.6.0 -> ballista-0.6.0 * [new tag] ballista-0.7.0 -> ballista-0.7.0 * [new tag] python-0.3.0 -> python-0.3.0 * [new tag] python-0.4.0 -> python-0.4.0 Switched to branch 'q4' ``` --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448524428 Benchmark for [this request](https://github.com/apache/datafusion/pull/22145#issuecomment-4448508770) failed. Last 20 lines of output: Click to expand ``` * [new tag] 48.0.0-rc1 -> 48.0.0-rc1 * [new tag] 48.0.0-rc2 -> 48.0.0-rc2 * [new tag] 5.0.0-> 5.0.0 * [new tag] 5.0.0-rc1-> 5.0.0-rc1 * [new tag] 5.0.0-rc3-> 5.0.0-rc3 * [new tag] 6.0.0-> 6.0.0 * [new tag] 6.0.0-rc0-> 6.0.0-rc0 * [new tag] 7.0.0-> 7.0.0 * [new tag] 7.0.0-rc2-> 7.0.0-rc2 * [new tag] 8.0.0-> 8.0.0 * [new tag] 8.0.0-rc1-> 8.0.0-rc1 * [new tag] 8.0.0-rc2-> 8.0.0-rc2 * [new tag] 9.0.0-> 9.0.0 * [new tag] 9.0.0-rc1-> 9.0.0-rc1 * [new tag] ballista-0.5.0 -> ballista-0.5.0 * [new tag] ballista-0.6.0 -> ballista-0.6.0 * [new tag] ballista-0.7.0 -> ballista-0.7.0 * [new tag] python-0.3.0 -> python-0.3.0 * [new tag] python-0.4.0 -> python-0.4.0 Switched to branch 'q4' ``` --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4448508770 run benchmarks -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4440476186 Benchmark for [this request](https://github.com/apache/datafusion/pull/22145#issuecomment-4440449362) failed. Last 20 lines of output: Click to expand ``` * [new tag] 48.0.0-rc1 -> 48.0.0-rc1 * [new tag] 48.0.0-rc2 -> 48.0.0-rc2 * [new tag] 5.0.0-> 5.0.0 * [new tag] 5.0.0-rc1-> 5.0.0-rc1 * [new tag] 5.0.0-rc3-> 5.0.0-rc3 * [new tag] 6.0.0-> 6.0.0 * [new tag] 6.0.0-rc0-> 6.0.0-rc0 * [new tag] 7.0.0-> 7.0.0 * [new tag] 7.0.0-rc2-> 7.0.0-rc2 * [new tag] 8.0.0-> 8.0.0 * [new tag] 8.0.0-rc1-> 8.0.0-rc1 * [new tag] 8.0.0-rc2-> 8.0.0-rc2 * [new tag] 9.0.0-> 9.0.0 * [new tag] 9.0.0-rc1-> 9.0.0-rc1 * [new tag] ballista-0.5.0 -> ballista-0.5.0 * [new tag] ballista-0.6.0 -> ballista-0.6.0 * [new tag] ballista-0.7.0 -> ballista-0.7.0 * [new tag] python-0.3.0 -> python-0.3.0 * [new tag] python-0.4.0 -> python-0.4.0 Switched to branch 'q4' ``` --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4440471735 Benchmark for [this request](https://github.com/apache/datafusion/pull/22145#issuecomment-4440449362) failed. Last 20 lines of output: Click to expand ``` * [new tag] 48.0.0-rc1 -> 48.0.0-rc1 * [new tag] 48.0.0-rc2 -> 48.0.0-rc2 * [new tag] 5.0.0-> 5.0.0 * [new tag] 5.0.0-rc1-> 5.0.0-rc1 * [new tag] 5.0.0-rc3-> 5.0.0-rc3 * [new tag] 6.0.0-> 6.0.0 * [new tag] 6.0.0-rc0-> 6.0.0-rc0 * [new tag] 7.0.0-> 7.0.0 * [new tag] 7.0.0-rc2-> 7.0.0-rc2 * [new tag] 8.0.0-> 8.0.0 * [new tag] 8.0.0-rc1-> 8.0.0-rc1 * [new tag] 8.0.0-rc2-> 8.0.0-rc2 * [new tag] 9.0.0-> 9.0.0 * [new tag] 9.0.0-rc1-> 9.0.0-rc1 * [new tag] ballista-0.5.0 -> ballista-0.5.0 * [new tag] ballista-0.6.0 -> ballista-0.6.0 * [new tag] ballista-0.7.0 -> ballista-0.7.0 * [new tag] python-0.3.0 -> python-0.3.0 * [new tag] python-0.4.0 -> python-0.4.0 Switched to branch 'q4' ``` --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
adriangbot commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4440470558 Benchmark for [this request](https://github.com/apache/datafusion/pull/22145#issuecomment-4440449362) failed. Last 20 lines of output: Click to expand ``` * [new tag] 48.0.0-rc1 -> 48.0.0-rc1 * [new tag] 48.0.0-rc2 -> 48.0.0-rc2 * [new tag] 5.0.0-> 5.0.0 * [new tag] 5.0.0-rc1-> 5.0.0-rc1 * [new tag] 5.0.0-rc3-> 5.0.0-rc3 * [new tag] 6.0.0-> 6.0.0 * [new tag] 6.0.0-rc0-> 6.0.0-rc0 * [new tag] 7.0.0-> 7.0.0 * [new tag] 7.0.0-rc2-> 7.0.0-rc2 * [new tag] 8.0.0-> 8.0.0 * [new tag] 8.0.0-rc1-> 8.0.0-rc1 * [new tag] 8.0.0-rc2-> 8.0.0-rc2 * [new tag] 9.0.0-> 9.0.0 * [new tag] 9.0.0-rc1-> 9.0.0-rc1 * [new tag] ballista-0.5.0 -> ballista-0.5.0 * [new tag] ballista-0.6.0 -> ballista-0.6.0 * [new tag] ballista-0.7.0 -> ballista-0.7.0 * [new tag] python-0.3.0 -> python-0.3.0 * [new tag] python-0.4.0 -> python-0.4.0 Switched to branch 'q4' ``` --- [File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner -- 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]
Re: [PR] [PoC] perf: optimize group-only group-by case for primitive cases (clickbench q4) [datafusion]
Dandandan commented on PR #22145: URL: https://github.com/apache/datafusion/pull/22145#issuecomment-4440449362 run benchmarks -- 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]
