You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@datafusion.apache.org by "adriangb (via GitHub)" <gi...@apache.org> on 2026/09/09 16:05:32 UTC
[PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
adriangb opened a new pull request, #25123:
URL: https://github.com/apache/datafusion/pull/25123
## Which issue does this PR close?
- Closes https://github.com/apache/datafusion/issues/25116.
## Rationale for this change
Some aggregate functions do not have a `GroupsAccumulator`. The built-in
examples are `covar_samp`, `covar_pop`, `regr_*`, `approx_percentile_cont`,
`approx_median`, `nth_value` and `any_value`. A user-defined aggregate that has
only an `Accumulator` is also an example. For all of these, DataFusion uses
`GroupsAccumulatorAdapter`. The adapter keeps one `Accumulator` for each group.
It sends the rows of each input batch to the correct accumulators.
If you group by a column that has many different values, these queries are slow.
This query on the ClickBench `hits` data is an example:
```sql
SELECT "UserID", covar_samp("ResolutionWidth", "ResolutionHeight") AS c
FROM hits
GROUP BY "UserID";
```
The cause is the adapter. For each input batch, the adapter looked at every
group that exists. It did this even when the batch had rows for only a few of
those groups. There are 17.6 million different `UserID` values, thus
approximately 1.5 million groups in each of the 12 partitions. With 8192 rows in
a batch, the adapter did approximately 180 steps for each input row before it
started the aggregate work.
The adapter also kept a scratch `Vec<u32>` of row indexes for each group. That
memory stayed for the full life of the group.
## What changes are included in this PR?
### 1. The adapter sorts the rows of a batch by group with a count
Each group now keeps one `u32` value in place of the scratch `Vec<u32>`. The
value is 0 between batches. For each batch, the adapter does three steps:
1. It counts the rows of each group. It also records a group the first time that
it finds a row for that group.
2. It adds those counts together to get the position where the rows of each
group start in the `take` index. Each count becomes that start position.
3. It writes each row index at the position for its group. The positions then go
back to 0 for the next batch.
Each step goes through the rows of the batch, or through the groups that the
batch has rows for. No step goes through all the groups that exist. The only
scratch memory that stays between batches is the list of groups with rows and
the list of start positions. Both are not larger than one batch. `size()` counts
both of them.
The rows of a group stay in the same order as in the input. Thus each
`Accumulator` gets the same rows, in the same order, as before this change.
This change also makes https://github.com/apache/datafusion/pull/24858
unnecessary. There is no more scratch memory for each group to account for. The
two memory limit tests in that PR show that the adapter spills because of that
memory, thus they do not apply after this change.
### 2. Two new `clickbench_extended` queries
Q14 groups `covar_samp` by `"UserID"` (17.6 million groups) and Q15 groups it by
`"RegionID"` (9,040 groups). The aggregate is cheap, thus the adapter is the
largest part of the time. An outer `MAX` keeps the result small.
### 3. A new benchmark for the adapter
`datafusion/functions-aggregate/benches/groups_accumulator_adapter.rs` moves the
group count from 64 to 1,000,000. It uses two accumulators, because the cost of
the adapter and the cost of the aggregate move in opposite directions as the
group count increases:
- `adapter_routing` uses an accumulator that only counts the rows that it gets.
What it measures is the adapter and nothing else.
- `adapter_covar_samp` uses a real aggregate that has no `GroupsAccumulator`. It
shows how much of that a query gets.
## Benchmark results
### ClickBench
`hits_partitioned`, 100 million rows, 12 partitions, warm page cache, release
build, M-series laptop with 12 cores. This machine moves by approximately 10%
between sequential runs. Thus the two binaries ran one after the other for each
single measurement, and the numbers below are medians of 12 measurements for
`"RegionID"` and 6 for `"UserID"`. `corr` on the same two columns is the control:
it has a `GroupsAccumulator`, thus this PR does not change it.
| Query | main | this PR | Ratio |
| ----------------------------------------- | ------- | ------- | ----- |
| `covar_samp` by `"UserID"` (17.6M groups) | 7278 ms | 5226 ms | 1.39x |
| `corr` by `"UserID"` (control) | 663 ms | 708 ms | 0.94x |
| `covar_samp` by `"RegionID"` (9k groups) | 267 ms | 275 ms | 0.97x |
| `corr` by `"RegionID"` (control) | 189 ms | 183 ms | 1.04x |
The two controls give a noise band of approximately 6%. The `"RegionID"` result
is inside that band, thus this PR does not make the low group count slower.
### The new adapter benchmark
| Groups | `adapter_routing` | `adapter_covar_samp` |
| --------- | ----------------- | -------------------- |
| 64 | 0.93x | 1.00x (p = 0.22) |
| 1,024 | 1.06x | 1.04x (p = 0.18) |
| 16,384 | 1.09x | 1.05x |
| 262,144 | 1.48x | 1.24x |
| 1,000,000 | 1.78x | 1.58x |
At 64 groups the adapter alone is approximately 7% slower. The old code copied
one contiguous block of row indexes for each group, and the new code writes each
row index on its own. A real accumulator hides that cost: `adapter_covar_samp`
at 64 and at 1,024 groups shows no difference that is statistically significant.
A second code path above a group count limit would remove those 7%. This PR does
not add one. One design is sufficient at both ends, and a second path adds a
value to tune.
## What is the testing strategy for this PR?
Two new unit tests in
`datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs`:
- `adapter_routes_rows_to_their_own_group` sends five batches, which include an
empty batch, through the adapter. The batches interleave the groups, they go
back to groups that they already used, and they leave some groups with no
rows. The test then compares all the groups against a result that it counts
itself.
- `adapter_routes_filtered_rows` shows that a filter keeps rows away from the
accumulators. This includes a group where the filter removes all of the rows.
To make sure that the first test catches an error, I removed the step that sets
the positions back to 0. The test failed. I then put the step back.
The full extended test suite is green: 11,042 tests pass.
## Are there any user-facing changes?
Queries that group by a column with many different values, and that use an
aggregate with no `GroupsAccumulator`, are faster. There are no changes to any
public API.
`GroupsAccumulatorAdapter::size()` reports a different number. It no longer
counts scratch memory for each group, because there is none. It now counts the
scratch memory of one batch. A memory pool sees a smaller number for the same
query.
🤖 Generated with [Claude Code](https://claude.com/claude-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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767718680
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676730)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767676730-2569-zjmbp 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 0576a0b400437ade5a6f3b102465d954a539f263 (0576a0b400437ade5a6f3b102465d954a539f263) to 0576a0b400437ade5a6f3b102465d954a539f263 [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..0576a0b400437ade5a6f3b102465d954a539f263)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
changed:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613943661
run benchmark clickbench_partitioned
baseline:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
changed:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605295426
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5605256201-2272-zfs4x 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607866038
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825245)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5607825245-2282-4xj8k 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607819406
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607779662)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5607779662-2281-kwngm 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767676204
run benchmark clickbench_extended
baseline:
ref: 0576a0b400437ade5a6f3b102465d954a539f263
changed:
ref: 0576a0b400437ade5a6f3b102465d954a539f263
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767892927
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767675578)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 835.22 ms │ 839.70 ms │ no change │
│ QQuery 1 │ 173.12 ms │ 173.90 ms │ no change │
│ QQuery 2 │ 447.18 ms │ 440.46 ms │ no change │
│ QQuery 3 │ 326.93 ms │ 327.80 ms │ no change │
│ QQuery 4 │ 1193.69 ms │ 1187.70 ms │ no change │
│ QQuery 5 │ 11133.17 ms │ 10428.65 ms │ +1.07x faster │
│ QQuery 6 │ 2.81 ms │ 2.73 ms │ no change │
│ QQuery 7 │ 775.95 ms │ 789.65 ms │ no change │
│ QQuery 8 │ 450.82 ms │ 447.54 ms │ no change │
│ QQuery 9 │ 2918.80 ms │ 3199.36 ms │ 1.10x slower │
│ QQuery 10 │ 638.92 ms │ 644.39 ms │ no change │
│ QQuery 11 │ 2202.49 ms │ 2057.20 ms │ +1.07x faster │
│ QQuery 12 │ 193.44 ms │ 193.73 ms │ no change │
│ QQuery 13 │ 540.51 ms │ 545.38 ms │ no change │
│ QQuery 14 │ 578.57 ms │ 592.89 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22411.62ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 21871.09ms │
│ Average Time (HEAD) │ 1494.11ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1458.07ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 12 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 835.22 / 859.65 ±12.47 / 869.95 ms │ 839.70 / 855.40 ±10.49 / 867.83 ms │ no change │
│ QQuery 1 │ 173.12 / 173.61 ±0.37 / 174.07 ms │ 173.90 / 174.24 ±0.32 / 174.83 ms │ no change │
│ QQuery 2 │ 447.18 / 449.60 ±1.60 / 451.64 ms │ 440.46 / 442.86 ±1.41 / 444.02 ms │ no change │
│ QQuery 3 │ 326.93 / 328.15 ±1.09 / 329.64 ms │ 327.80 / 329.91 ±2.27 / 333.48 ms │ no change │
│ QQuery 4 │ 1193.69 / 1199.22 ±4.39 / 1205.92 ms │ 1187.70 / 1194.22 ±5.81 / 1202.98 ms │ no change │
│ QQuery 5 │ 11133.17 / 11508.99 ±442.85 / 12376.80 ms │ 10428.65 / 10932.12 ±276.30 / 11213.55 ms │ +1.05x faster │
│ QQuery 6 │ 2.81 / 3.13 ±0.40 / 3.92 ms │ 2.73 / 3.06 ±0.34 / 3.69 ms │ no change │
│ QQuery 7 │ 775.95 / 824.43 ±45.94 / 910.14 ms │ 789.65 / 860.80 ±84.81 / 1023.50 ms │ no change │
│ QQuery 8 │ 450.82 / 457.69 ±10.21 / 477.82 ms │ 447.54 / 460.70 ±12.63 / 477.41 ms │ no change │
│ QQuery 9 │ 2918.80 / 2999.61 ±81.40 / 3149.94 ms │ 3199.36 / 3260.78 ±39.99 / 3316.02 ms │ 1.09x slower │
│ QQuery 10 │ 638.92 / 675.53 ±47.21 / 765.76 ms │ 644.39 / 692.91 ±76.52 / 844.45 ms │ no change │
│ QQuery 11 │ 2202.49 / 2229.25 ±26.89 / 2274.85 ms │ 2057.20 / 2144.68 ±62.22 / 2243.06 ms │ no change │
│ QQuery 12 │ 193.44 / 221.11 ±41.69 / 303.53 ms │ 193.73 / 214.26 ±29.19 / 270.58 ms │ no change │
│ QQuery 13 │ 540.51 / 549.10 ±6.65 / 556.91 ms │ 545.38 / 571.80 ±22.53 / 611.17 ms │ no change │
│ QQuery 14 │ 578.57 / 618.21 ±42.56 / 698.86 ms │ 592.89 / 629.42 ±33.42 / 683.02 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23097.27ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22767.17ms │
│ Average Time (HEAD) │ 1539.82ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1517.81ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 12.0 GiB |
| Avg memory | 4.9 GiB |
| CPU user | 1060.2s |
| CPU sys | 49.7s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 115.0s |
| Peak memory | 11.4 GiB |
| Avg memory | 5.0 GiB |
| CPU user | 1037.9s |
| CPU sys | 51.5s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5614114792
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943661)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.39 ms │ 1.26 ms │ +1.11x faster │
│ QQuery 1 │ 12.49 ms │ 11.69 ms │ +1.07x faster │
│ QQuery 2 │ 37.09 ms │ 35.98 ms │ no change │
│ QQuery 3 │ 32.40 ms │ 30.70 ms │ +1.06x faster │
│ QQuery 4 │ 261.06 ms │ 218.05 ms │ +1.20x faster │
│ QQuery 5 │ 282.88 ms │ 273.05 ms │ no change │
│ QQuery 6 │ 1.25 ms │ 1.23 ms │ no change │
│ QQuery 7 │ 13.27 ms │ 13.01 ms │ no change │
│ QQuery 8 │ 329.96 ms │ 321.81 ms │ no change │
│ QQuery 9 │ 455.38 ms │ 454.79 ms │ no change │
│ QQuery 10 │ 71.31 ms │ 70.44 ms │ no change │
│ QQuery 11 │ 82.89 ms │ 81.11 ms │ no change │
│ QQuery 12 │ 270.66 ms │ 265.80 ms │ no change │
│ QQuery 13 │ 403.44 ms │ 372.18 ms │ +1.08x faster │
│ QQuery 14 │ 291.75 ms │ 281.64 ms │ no change │
│ QQuery 15 │ 266.01 ms │ 267.07 ms │ no change │
│ QQuery 16 │ 614.41 ms │ 653.26 ms │ 1.06x slower │
│ QQuery 17 │ 614.17 ms │ 714.91 ms │ 1.16x slower │
│ QQuery 18 │ 1270.10 ms │ 1266.55 ms │ no change │
│ QQuery 19 │ 27.48 ms │ 27.08 ms │ no change │
│ QQuery 20 │ 514.58 ms │ 523.69 ms │ no change │
│ QQuery 21 │ 510.39 ms │ 513.05 ms │ no change │
│ QQuery 22 │ 980.24 ms │ 976.96 ms │ no change │
│ QQuery 23 │ 3086.43 ms │ 3084.57 ms │ no change │
│ QQuery 24 │ 42.96 ms │ 40.94 ms │ no change │
│ QQuery 25 │ 117.90 ms │ 110.10 ms │ +1.07x faster │
│ QQuery 26 │ 43.72 ms │ 40.94 ms │ +1.07x faster │
│ QQuery 27 │ 530.76 ms │ 517.19 ms │ no change │
│ QQuery 28 │ 2932.28 ms │ 2913.38 ms │ no change │
│ QQuery 29 │ 42.39 ms │ 40.91 ms │ no change │
│ QQuery 30 │ 300.09 ms │ 303.21 ms │ no change │
│ QQuery 31 │ 277.61 ms │ 294.14 ms │ 1.06x slower │
│ QQuery 32 │ 912.70 ms │ 1100.49 ms │ 1.21x slower │
│ QQuery 33 │ 1447.38 ms │ 1477.63 ms │ no change │
│ QQuery 34 │ 1420.90 ms │ 1492.89 ms │ 1.05x slower │
│ QQuery 35 │ 276.86 ms │ 282.32 ms │ no change │
│ QQuery 36 │ 65.99 ms │ 66.15 ms │ no change │
│ QQuery 37 │ 35.26 ms │ 35.29 ms │ no change │
│ QQuery 38 │ 41.67 ms │ 42.33 ms │ no change │
│ QQuery 39 │ 136.52 ms │ 141.90 ms │ no change │
│ QQuery 40 │ 15.33 ms │ 14.91 ms │ no change │
│ QQuery 41 │ 15.08 ms │ 14.31 ms │ +1.05x faster │
│ QQuery 42 │ 14.33 ms │ 14.15 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19100.74ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19403.05ms │
│ Average Time (HEAD) │ 444.20ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 451.23ms │
│ Queries Faster │ 8 │
│ Queries Slower │ 5 │
│ Queries with No Change │ 30 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.39 / 4.44 ±5.87 / 16.17 ms │ 1.26 / 4.18 ±5.71 / 15.61 ms │ +1.06x faster │
│ QQuery 1 │ 12.49 / 12.80 ±0.38 / 13.49 ms │ 11.69 / 11.96 ±0.20 / 12.15 ms │ +1.07x faster │
│ QQuery 2 │ 37.09 / 37.50 ±0.32 / 38.01 ms │ 35.98 / 36.52 ±0.53 / 37.31 ms │ no change │
│ QQuery 3 │ 32.40 / 33.24 ±0.70 / 34.11 ms │ 30.70 / 31.10 ±0.46 / 31.99 ms │ +1.07x faster │
│ QQuery 4 │ 261.06 / 266.78 ±3.38 / 270.11 ms │ 218.05 / 222.89 ±2.94 / 226.76 ms │ +1.20x faster │
│ QQuery 5 │ 282.88 / 308.75 ±13.09 / 318.06 ms │ 273.05 / 274.74 ±1.33 / 277.08 ms │ +1.12x faster │
│ QQuery 6 │ 1.25 / 1.40 ±0.22 / 1.83 ms │ 1.23 / 1.39 ±0.23 / 1.84 ms │ no change │
│ QQuery 7 │ 13.27 / 13.47 ±0.12 / 13.59 ms │ 13.01 / 13.10 ±0.09 / 13.25 ms │ no change │
│ QQuery 8 │ 329.96 / 341.71 ±8.96 / 356.58 ms │ 321.81 / 326.92 ±4.09 / 333.22 ms │ no change │
│ QQuery 9 │ 455.38 / 464.13 ±5.19 / 469.49 ms │ 454.79 / 459.95 ±6.48 / 472.67 ms │ no change │
│ QQuery 10 │ 71.31 / 74.19 ±2.88 / 79.43 ms │ 70.44 / 71.19 ±0.95 / 72.77 ms │ no change │
│ QQuery 11 │ 82.89 / 83.85 ±0.83 / 85.12 ms │ 81.11 / 81.97 ±0.76 / 83.27 ms │ no change │
│ QQuery 12 │ 270.66 / 284.64 ±15.70 / 315.16 ms │ 265.80 / 272.51 ±5.89 / 280.21 ms │ no change │
│ QQuery 13 │ 403.44 / 425.10 ±16.67 / 453.57 ms │ 372.18 / 383.95 ±8.94 / 398.03 ms │ +1.11x faster │
│ QQuery 14 │ 291.75 / 299.41 ±8.53 / 309.99 ms │ 281.64 / 288.26 ±6.41 / 296.92 ms │ no change │
│ QQuery 15 │ 266.01 / 275.69 ±5.34 / 282.23 ms │ 267.07 / 279.36 ±6.77 / 285.72 ms │ no change │
│ QQuery 16 │ 614.41 / 645.91 ±40.24 / 723.75 ms │ 653.26 / 707.81 ±28.08 / 733.36 ms │ 1.10x slower │
│ QQuery 17 │ 614.17 / 633.88 ±16.98 / 664.08 ms │ 714.91 / 728.38 ±7.11 / 735.29 ms │ 1.15x slower │
│ QQuery 18 │ 1270.10 / 1329.98 ±67.03 / 1456.53 ms │ 1266.55 / 1341.29 ±62.01 / 1443.15 ms │ no change │
│ QQuery 19 │ 27.48 / 33.55 ±11.60 / 56.74 ms │ 27.08 / 33.44 ±11.84 / 57.10 ms │ no change │
│ QQuery 20 │ 514.58 / 521.89 ±6.71 / 532.56 ms │ 523.69 / 536.70 ±8.94 / 551.73 ms │ no change │
│ QQuery 21 │ 510.39 / 515.80 ±3.96 / 520.24 ms │ 513.05 / 516.78 ±3.66 / 523.53 ms │ no change │
│ QQuery 22 │ 980.24 / 988.96 ±6.82 / 1001.10 ms │ 976.96 / 987.01 ±6.77 / 996.82 ms │ no change │
│ QQuery 23 │ 3086.43 / 3119.25 ±21.43 / 3154.00 ms │ 3084.57 / 3143.28 ±43.39 / 3217.95 ms │ no change │
│ QQuery 24 │ 42.96 / 50.62 ±6.86 / 62.50 ms │ 40.94 / 43.80 ±5.46 / 54.72 ms │ +1.16x faster │
│ QQuery 25 │ 117.90 / 118.59 ±0.91 / 120.36 ms │ 110.10 / 111.32 ±1.26 / 113.51 ms │ +1.07x faster │
│ QQuery 26 │ 43.72 / 46.85 ±3.12 / 51.91 ms │ 40.94 / 43.06 ±2.50 / 47.57 ms │ +1.09x faster │
│ QQuery 27 │ 530.76 / 540.60 ±7.13 / 552.81 ms │ 517.19 / 524.52 ±6.48 / 534.26 ms │ no change │
│ QQuery 28 │ 2932.28 / 2989.84 ±58.23 / 3084.25 ms │ 2913.38 / 2932.40 ±25.68 / 2983.34 ms │ no change │
│ QQuery 29 │ 42.39 / 50.94 ±16.69 / 84.31 ms │ 40.91 / 41.15 ±0.26 / 41.66 ms │ +1.24x faster │
│ QQuery 30 │ 300.09 / 322.60 ±33.96 / 389.94 ms │ 303.21 / 308.05 ±5.39 / 318.56 ms │ no change │
│ QQuery 31 │ 277.61 / 289.37 ±9.00 / 302.84 ms │ 294.14 / 314.73 ±15.86 / 337.83 ms │ 1.09x slower │
│ QQuery 32 │ 912.70 / 965.38 ±53.34 / 1056.02 ms │ 1100.49 / 1116.59 ±15.10 / 1135.70 ms │ 1.16x slower │
│ QQuery 33 │ 1447.38 / 1513.85 ±69.91 / 1648.99 ms │ 1477.63 / 1551.98 ±88.57 / 1723.00 ms │ no change │
│ QQuery 34 │ 1420.90 / 1489.55 ±57.33 / 1578.75 ms │ 1492.89 / 1563.64 ±87.02 / 1729.07 ms │ no change │
│ QQuery 35 │ 276.86 / 301.20 ±34.97 / 370.20 ms │ 282.32 / 310.02 ±22.69 / 337.39 ms │ no change │
│ QQuery 36 │ 65.99 / 72.01 ±4.79 / 79.83 ms │ 66.15 / 72.51 ±5.12 / 80.96 ms │ no change │
│ QQuery 37 │ 35.26 / 36.01 ±0.66 / 36.90 ms │ 35.29 / 42.01 ±8.79 / 58.35 ms │ 1.17x slower │
│ QQuery 38 │ 41.67 / 46.75 ±8.60 / 63.91 ms │ 42.33 / 43.70 ±1.19 / 45.14 ms │ +1.07x faster │
│ QQuery 39 │ 136.52 / 151.23 ±10.43 / 168.62 ms │ 141.90 / 152.83 ±9.92 / 167.51 ms │ no change │
│ QQuery 40 │ 15.33 / 15.77 ±0.44 / 16.62 ms │ 14.91 / 15.44 ±0.53 / 16.45 ms │ no change │
│ QQuery 41 │ 15.08 / 18.60 ±6.56 / 31.72 ms │ 14.31 / 15.16 ±1.12 / 17.33 ms │ +1.23x faster │
│ QQuery 42 │ 14.33 / 16.15 ±3.29 / 22.73 ms │ 14.15 / 14.28 ±0.15 / 14.49 ms │ +1.13x faster │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19752.26ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19971.88ms │
│ Average Time (HEAD) │ 459.35ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 464.46ms │
│ Queries Faster │ 13 │
│ Queries Slower │ 5 │
│ Queries with No Change │ 25 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 10.3 GiB |
| Avg memory | 4.3 GiB |
| CPU user | 1015.1s |
| CPU sys | 69.8s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 11.4 GiB |
| Avg memory | 4.3 GiB |
| CPU user | 1022.1s |
| CPU sys | 73.9s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607825438
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611504239
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611481841-2292-t5j65 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611505890
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611481841-2291-fjjnt 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611643761
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611499243)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611499243-2296-phgp8 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpcds
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611733982
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611499243)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpcds
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 5.63 ms │ 5.73 ms │ no change │
│ QQuery 2 │ 80.15 ms │ 80.61 ms │ no change │
│ QQuery 3 │ 28.95 ms │ 29.52 ms │ no change │
│ QQuery 4 │ 481.17 ms │ 496.66 ms │ no change │
│ QQuery 5 │ 51.74 ms │ 51.70 ms │ no change │
│ QQuery 6 │ 35.59 ms │ 36.48 ms │ no change │
│ QQuery 7 │ 74.92 ms │ 75.25 ms │ no change │
│ QQuery 8 │ 36.38 ms │ 36.50 ms │ no change │
│ QQuery 9 │ 51.85 ms │ 53.37 ms │ no change │
│ QQuery 10 │ 63.29 ms │ 63.32 ms │ no change │
│ QQuery 11 │ 296.87 ms │ 293.46 ms │ no change │
│ QQuery 12 │ 28.52 ms │ 28.52 ms │ no change │
│ QQuery 13 │ 117.36 ms │ 117.95 ms │ no change │
│ QQuery 14 │ 416.29 ms │ 413.32 ms │ no change │
│ QQuery 15 │ 57.65 ms │ 56.77 ms │ no change │
│ QQuery 16 │ 6.63 ms │ 6.68 ms │ no change │
│ QQuery 17 │ 79.38 ms │ 79.40 ms │ no change │
│ QQuery 18 │ 103.75 ms │ 104.54 ms │ no change │
│ QQuery 19 │ 41.36 ms │ 41.21 ms │ no change │
│ QQuery 20 │ 35.47 ms │ 35.27 ms │ no change │
│ QQuery 21 │ 17.23 ms │ 16.87 ms │ no change │
│ QQuery 22 │ 63.52 ms │ 64.09 ms │ no change │
│ QQuery 23 │ 309.07 ms │ 317.98 ms │ no change │
│ QQuery 24 │ 193.80 ms │ 207.26 ms │ 1.07x slower │
│ QQuery 25 │ 109.61 ms │ 110.92 ms │ no change │
│ QQuery 26 │ 49.07 ms │ 50.07 ms │ no change │
│ QQuery 27 │ 6.23 ms │ 6.36 ms │ no change │
│ QQuery 28 │ 56.28 ms │ 57.33 ms │ no change │
│ QQuery 29 │ 97.13 ms │ 98.02 ms │ no change │
│ QQuery 30 │ 32.28 ms │ 33.09 ms │ no change │
│ QQuery 31 │ 109.51 ms │ 112.13 ms │ no change │
│ QQuery 32 │ 19.98 ms │ 20.83 ms │ no change │
│ QQuery 33 │ 38.04 ms │ 38.60 ms │ no change │
│ QQuery 34 │ 10.09 ms │ 10.28 ms │ no change │
│ QQuery 35 │ 72.49 ms │ 73.29 ms │ no change │
│ QQuery 36 │ 5.71 ms │ 5.70 ms │ no change │
│ QQuery 37 │ 6.73 ms │ 6.83 ms │ no change │
│ QQuery 38 │ 61.23 ms │ 62.00 ms │ no change │
│ QQuery 39 │ 89.40 ms │ 89.32 ms │ no change │
│ QQuery 40 │ 23.64 ms │ 24.13 ms │ no change │
│ QQuery 41 │ 11.20 ms │ 11.15 ms │ no change │
│ QQuery 42 │ 23.31 ms │ 23.63 ms │ no change │
│ QQuery 43 │ 5.24 ms │ 5.06 ms │ no change │
│ QQuery 44 │ 9.39 ms │ 9.19 ms │ no change │
│ QQuery 45 │ 38.52 ms │ 39.05 ms │ no change │
│ QQuery 46 │ 11.78 ms │ 11.59 ms │ no change │
│ QQuery 47 │ 225.70 ms │ 223.59 ms │ no change │
│ QQuery 48 │ 95.50 ms │ 96.15 ms │ no change │
│ QQuery 49 │ 71.99 ms │ 70.10 ms │ no change │
│ QQuery 50 │ 59.04 ms │ 58.09 ms │ no change │
│ QQuery 51 │ 91.41 ms │ 92.19 ms │ no change │
│ QQuery 52 │ 24.07 ms │ 23.75 ms │ no change │
│ QQuery 53 │ 29.20 ms │ 28.77 ms │ no change │
│ QQuery 54 │ 55.11 ms │ 53.97 ms │ no change │
│ QQuery 55 │ 23.35 ms │ 23.13 ms │ no change │
│ QQuery 56 │ 38.82 ms │ 38.55 ms │ no change │
│ QQuery 57 │ 176.02 ms │ 176.06 ms │ no change │
│ QQuery 58 │ 111.98 ms │ 112.68 ms │ no change │
│ QQuery 59 │ 117.56 ms │ 118.02 ms │ no change │
│ QQuery 60 │ 39.51 ms │ 39.51 ms │ no change │
│ QQuery 61 │ 12.51 ms │ 12.40 ms │ no change │
│ QQuery 62 │ 46.03 ms │ 46.00 ms │ no change │
│ QQuery 63 │ 29.05 ms │ 29.50 ms │ no change │
│ QQuery 64 │ 368.03 ms │ 373.81 ms │ no change │
│ QQuery 65 │ 120.63 ms │ 127.18 ms │ 1.05x slower │
│ QQuery 66 │ 79.96 ms │ 79.74 ms │ no change │
│ QQuery 67 │ 237.32 ms │ 257.03 ms │ 1.08x slower │
│ QQuery 68 │ 11.92 ms │ 12.23 ms │ no change │
│ QQuery 69 │ 57.93 ms │ 58.30 ms │ no change │
│ QQuery 70 │ 105.42 ms │ 106.94 ms │ no change │
│ QQuery 71 │ 35.33 ms │ 35.69 ms │ no change │
│ QQuery 72 │ 1859.96 ms │ 1845.46 ms │ no change │
│ QQuery 73 │ 10.74 ms │ 10.11 ms │ +1.06x faster │
│ QQuery 74 │ 185.12 ms │ 182.63 ms │ no change │
│ QQuery 75 │ 149.51 ms │ 147.67 ms │ no change │
│ QQuery 76 │ 35.85 ms │ 35.60 ms │ no change │
│ QQuery 77 │ 62.50 ms │ 61.60 ms │ no change │
│ QQuery 78 │ 230.29 ms │ 225.34 ms │ no change │
│ QQuery 79 │ 66.74 ms │ 66.83 ms │ no change │
│ QQuery 80 │ 100.77 ms │ 99.01 ms │ no change │
│ QQuery 81 │ 26.78 ms │ 26.49 ms │ no change │
│ QQuery 82 │ 16.81 ms │ 16.73 ms │ no change │
│ QQuery 83 │ 34.86 ms │ 34.77 ms │ no change │
│ QQuery 84 │ 30.13 ms │ 30.02 ms │ no change │
│ QQuery 85 │ 104.88 ms │ 103.32 ms │ no change │
│ QQuery 86 │ 26.18 ms │ 25.87 ms │ no change │
│ QQuery 87 │ 64.81 ms │ 64.02 ms │ no change │
│ QQuery 88 │ 63.73 ms │ 62.79 ms │ no change │
│ QQuery 89 │ 35.84 ms │ 35.38 ms │ no change │
│ QQuery 90 │ 17.68 ms │ 17.80 ms │ no change │
│ QQuery 91 │ 45.37 ms │ 46.84 ms │ no change │
│ QQuery 92 │ 29.78 ms │ 30.51 ms │ no change │
│ QQuery 93 │ 50.82 ms │ 52.44 ms │ no change │
│ QQuery 94 │ 39.00 ms │ 39.41 ms │ no change │
│ QQuery 95 │ 80.82 ms │ 83.37 ms │ no change │
│ QQuery 96 │ 23.82 ms │ 24.95 ms │ no change │
│ QQuery 97 │ 51.99 ms │ 55.20 ms │ 1.06x slower │
│ QQuery 98 │ 42.48 ms │ 44.36 ms │ no change │
│ QQuery 99 │ 69.64 ms │ 71.73 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 9383.73ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 9444.68ms │
│ Average Time (HEAD) │ 94.79ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 95.40ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 4 │
│ Queries with No Change │ 94 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 5.63 / 6.22 ±0.97 / 8.16 ms │ 5.73 / 6.30 ±0.92 / 8.13 ms │ no change │
│ QQuery 2 │ 80.15 / 80.41 ±0.18 / 80.63 ms │ 80.61 / 80.93 ±0.38 / 81.44 ms │ no change │
│ QQuery 3 │ 28.95 / 29.14 ±0.15 / 29.39 ms │ 29.52 / 29.67 ±0.13 / 29.87 ms │ no change │
│ QQuery 4 │ 481.17 / 482.87 ±2.55 / 487.92 ms │ 496.66 / 512.36 ±10.95 / 524.90 ms │ 1.06x slower │
│ QQuery 5 │ 51.74 / 55.02 ±4.99 / 64.96 ms │ 51.70 / 52.40 ±0.47 / 53.09 ms │ no change │
│ QQuery 6 │ 35.59 / 36.51 ±0.56 / 37.09 ms │ 36.48 / 36.77 ±0.26 / 37.21 ms │ no change │
│ QQuery 7 │ 74.92 / 75.70 ±0.44 / 76.20 ms │ 75.25 / 75.80 ±0.72 / 77.17 ms │ no change │
│ QQuery 8 │ 36.38 / 36.74 ±0.41 / 37.50 ms │ 36.50 / 38.20 ±2.34 / 42.85 ms │ no change │
│ QQuery 9 │ 51.85 / 56.21 ±6.00 / 68.08 ms │ 53.37 / 54.11 ±0.53 / 54.59 ms │ no change │
│ QQuery 10 │ 63.29 / 63.62 ±0.21 / 63.81 ms │ 63.32 / 63.59 ±0.19 / 63.76 ms │ no change │
│ QQuery 11 │ 296.87 / 300.86 ±4.13 / 307.60 ms │ 293.46 / 303.58 ±6.76 / 312.16 ms │ no change │
│ QQuery 12 │ 28.52 / 28.97 ±0.30 / 29.43 ms │ 28.52 / 29.06 ±0.40 / 29.75 ms │ no change │
│ QQuery 13 │ 117.36 / 118.58 ±0.72 / 119.60 ms │ 117.95 / 119.02 ±1.59 / 122.19 ms │ no change │
│ QQuery 14 │ 416.29 / 418.50 ±2.91 / 424.12 ms │ 413.32 / 418.09 ±3.89 / 425.20 ms │ no change │
│ QQuery 15 │ 57.65 / 59.56 ±2.25 / 63.88 ms │ 56.77 / 58.33 ±0.83 / 59.18 ms │ no change │
│ QQuery 16 │ 6.63 / 6.91 ±0.21 / 7.25 ms │ 6.68 / 6.86 ±0.19 / 7.21 ms │ no change │
│ QQuery 17 │ 79.38 / 81.20 ±1.30 / 82.78 ms │ 79.40 / 79.99 ±0.55 / 80.97 ms │ no change │
│ QQuery 18 │ 103.75 / 106.36 ±1.89 / 109.26 ms │ 104.54 / 105.96 ±1.48 / 108.67 ms │ no change │
│ QQuery 19 │ 41.36 / 41.62 ±0.23 / 41.91 ms │ 41.21 / 41.51 ±0.24 / 41.85 ms │ no change │
│ QQuery 20 │ 35.47 / 35.69 ±0.12 / 35.82 ms │ 35.27 / 35.62 ±0.26 / 36.07 ms │ no change │
│ QQuery 21 │ 17.23 / 18.20 ±1.03 / 19.99 ms │ 16.87 / 17.85 ±1.10 / 19.99 ms │ no change │
│ QQuery 22 │ 63.52 / 64.26 ±0.72 / 65.46 ms │ 64.09 / 64.89 ±0.47 / 65.45 ms │ no change │
│ QQuery 23 │ 309.07 / 313.26 ±3.81 / 320.36 ms │ 317.98 / 321.66 ±3.58 / 328.08 ms │ no change │
│ QQuery 24 │ 193.80 / 197.81 ±4.71 / 206.83 ms │ 207.26 / 211.05 ±4.54 / 219.88 ms │ 1.07x slower │
│ QQuery 25 │ 109.61 / 112.18 ±1.64 / 114.41 ms │ 110.92 / 113.42 ±1.64 / 115.41 ms │ no change │
│ QQuery 26 │ 49.07 / 49.56 ±0.28 / 49.84 ms │ 50.07 / 50.37 ±0.22 / 50.67 ms │ no change │
│ QQuery 27 │ 6.23 / 6.39 ±0.27 / 6.93 ms │ 6.36 / 6.55 ±0.14 / 6.80 ms │ no change │
│ QQuery 28 │ 56.28 / 59.42 ±1.79 / 60.86 ms │ 57.33 / 60.21 ±2.27 / 62.78 ms │ no change │
│ QQuery 29 │ 97.13 / 100.23 ±4.23 / 108.53 ms │ 98.02 / 99.73 ±1.27 / 101.23 ms │ no change │
│ QQuery 30 │ 32.28 / 32.57 ±0.31 / 33.16 ms │ 33.09 / 33.41 ±0.36 / 33.92 ms │ no change │
│ QQuery 31 │ 109.51 / 112.24 ±3.24 / 118.24 ms │ 112.13 / 115.23 ±2.70 / 119.22 ms │ no change │
│ QQuery 32 │ 19.98 / 20.33 ±0.37 / 20.92 ms │ 20.83 / 21.13 ±0.18 / 21.34 ms │ no change │
│ QQuery 33 │ 38.04 / 38.95 ±1.00 / 40.88 ms │ 38.60 / 38.89 ±0.19 / 39.09 ms │ no change │
│ QQuery 34 │ 10.09 / 10.31 ±0.17 / 10.57 ms │ 10.28 / 10.41 ±0.08 / 10.51 ms │ no change │
│ QQuery 35 │ 72.49 / 73.10 ±0.39 / 73.44 ms │ 73.29 / 74.38 ±1.33 / 76.91 ms │ no change │
│ QQuery 36 │ 5.71 / 5.84 ±0.19 / 6.21 ms │ 5.70 / 5.81 ±0.19 / 6.20 ms │ no change │
│ QQuery 37 │ 6.73 / 6.92 ±0.16 / 7.19 ms │ 6.83 / 6.94 ±0.10 / 7.12 ms │ no change │
│ QQuery 38 │ 61.23 / 62.45 ±0.92 / 63.83 ms │ 62.00 / 62.73 ±0.65 / 63.56 ms │ no change │
│ QQuery 39 │ 89.40 / 90.20 ±0.47 / 90.63 ms │ 89.32 / 89.63 ±0.31 / 90.18 ms │ no change │
│ QQuery 40 │ 23.64 / 23.91 ±0.25 / 24.34 ms │ 24.13 / 26.27 ±2.54 / 31.23 ms │ 1.10x slower │
│ QQuery 41 │ 11.20 / 11.31 ±0.11 / 11.50 ms │ 11.15 / 11.26 ±0.12 / 11.48 ms │ no change │
│ QQuery 42 │ 23.31 / 23.63 ±0.29 / 24.09 ms │ 23.63 / 24.88 ±1.19 / 26.79 ms │ 1.05x slower │
│ QQuery 43 │ 5.24 / 6.26 ±1.89 / 10.04 ms │ 5.06 / 5.22 ±0.18 / 5.56 ms │ +1.20x faster │
│ QQuery 44 │ 9.39 / 9.63 ±0.19 / 9.98 ms │ 9.19 / 9.37 ±0.12 / 9.57 ms │ no change │
│ QQuery 45 │ 38.52 / 38.96 ±0.35 / 39.55 ms │ 39.05 / 39.50 ±0.53 / 40.43 ms │ no change │
│ QQuery 46 │ 11.78 / 12.41 ±0.58 / 13.38 ms │ 11.59 / 11.85 ±0.25 / 12.31 ms │ no change │
│ QQuery 47 │ 225.70 / 228.47 ±1.46 / 229.64 ms │ 223.59 / 226.97 ±2.04 / 229.08 ms │ no change │
│ QQuery 48 │ 95.50 / 95.92 ±0.38 / 96.51 ms │ 96.15 / 97.53 ±1.40 / 100.10 ms │ no change │
│ QQuery 49 │ 71.99 / 73.31 ±1.49 / 76.16 ms │ 70.10 / 70.99 ±0.70 / 71.89 ms │ no change │
│ QQuery 50 │ 59.04 / 60.45 ±1.66 / 63.13 ms │ 58.09 / 58.91 ±0.57 / 59.84 ms │ no change │
│ QQuery 51 │ 91.41 / 95.47 ±4.14 / 102.78 ms │ 92.19 / 94.76 ±2.31 / 97.78 ms │ no change │
│ QQuery 52 │ 24.07 / 24.33 ±0.19 / 24.58 ms │ 23.75 / 24.17 ±0.29 / 24.56 ms │ no change │
│ QQuery 53 │ 29.20 / 30.32 ±1.28 / 32.76 ms │ 28.77 / 29.15 ±0.34 / 29.74 ms │ no change │
│ QQuery 54 │ 55.11 / 55.40 ±0.34 / 56.04 ms │ 53.97 / 54.64 ±0.51 / 55.53 ms │ no change │
│ QQuery 55 │ 23.35 / 23.72 ±0.22 / 23.95 ms │ 23.13 / 23.35 ±0.23 / 23.73 ms │ no change │
│ QQuery 56 │ 38.82 / 39.32 ±0.32 / 39.65 ms │ 38.55 / 39.00 ±0.40 / 39.69 ms │ no change │
│ QQuery 57 │ 176.02 / 179.92 ±3.29 / 185.45 ms │ 176.06 / 177.42 ±1.34 / 179.68 ms │ no change │
│ QQuery 58 │ 111.98 / 114.82 ±2.54 / 119.41 ms │ 112.68 / 113.54 ±0.76 / 114.45 ms │ no change │
│ QQuery 59 │ 117.56 / 117.92 ±0.21 / 118.19 ms │ 118.02 / 118.88 ±1.20 / 121.26 ms │ no change │
│ QQuery 60 │ 39.51 / 40.36 ±0.88 / 42.04 ms │ 39.51 / 40.22 ±0.61 / 41.23 ms │ no change │
│ QQuery 61 │ 12.51 / 12.63 ±0.12 / 12.83 ms │ 12.40 / 12.52 ±0.15 / 12.79 ms │ no change │
│ QQuery 62 │ 46.03 / 46.66 ±0.82 / 48.24 ms │ 46.00 / 46.57 ±0.51 / 47.50 ms │ no change │
│ QQuery 63 │ 29.05 / 29.34 ±0.18 / 29.56 ms │ 29.50 / 29.82 ±0.18 / 29.97 ms │ no change │
│ QQuery 64 │ 368.03 / 371.40 ±4.11 / 379.26 ms │ 373.81 / 376.38 ±3.35 / 382.67 ms │ no change │
│ QQuery 65 │ 120.63 / 123.90 ±1.67 / 125.07 ms │ 127.18 / 128.78 ±1.23 / 130.77 ms │ no change │
│ QQuery 66 │ 79.96 / 80.96 ±1.37 / 83.65 ms │ 79.74 / 80.68 ±0.57 / 81.33 ms │ no change │
│ QQuery 67 │ 237.32 / 242.58 ±2.94 / 245.96 ms │ 257.03 / 260.15 ±2.86 / 264.81 ms │ 1.07x slower │
│ QQuery 68 │ 11.92 / 12.13 ±0.18 / 12.46 ms │ 12.23 / 12.38 ±0.10 / 12.51 ms │ no change │
│ QQuery 69 │ 57.93 / 59.83 ±3.11 / 66.03 ms │ 58.30 / 58.95 ±0.55 / 59.93 ms │ no change │
│ QQuery 70 │ 105.42 / 108.00 ±2.86 / 111.83 ms │ 106.94 / 112.49 ±5.49 / 122.39 ms │ no change │
│ QQuery 71 │ 35.33 / 35.63 ±0.25 / 36.06 ms │ 35.69 / 37.46 ±2.10 / 41.42 ms │ 1.05x slower │
│ QQuery 72 │ 1859.96 / 1889.98 ±41.72 / 1972.03 ms │ 1845.46 / 1920.05 ±50.97 / 1977.12 ms │ no change │
│ QQuery 73 │ 10.74 / 11.10 ±0.22 / 11.37 ms │ 10.11 / 10.57 ±0.71 / 11.98 ms │ no change │
│ QQuery 74 │ 185.12 / 189.48 ±3.37 / 194.19 ms │ 182.63 / 184.37 ±1.76 / 187.50 ms │ no change │
│ QQuery 75 │ 149.51 / 149.93 ±0.74 / 151.40 ms │ 147.67 / 149.78 ±2.22 / 152.50 ms │ no change │
│ QQuery 76 │ 35.85 / 38.16 ±2.52 / 42.98 ms │ 35.60 / 36.10 ±0.67 / 37.39 ms │ +1.06x faster │
│ QQuery 77 │ 62.50 / 63.97 ±1.75 / 67.23 ms │ 61.60 / 62.14 ±0.39 / 62.72 ms │ no change │
│ QQuery 78 │ 230.29 / 237.13 ±7.16 / 250.18 ms │ 225.34 / 233.01 ±5.68 / 240.05 ms │ no change │
│ QQuery 79 │ 66.74 / 67.39 ±0.44 / 67.83 ms │ 66.83 / 67.26 ±0.53 / 68.21 ms │ no change │
│ QQuery 80 │ 100.77 / 103.43 ±3.09 / 109.46 ms │ 99.01 / 103.74 ±5.94 / 115.45 ms │ no change │
│ QQuery 81 │ 26.78 / 27.46 ±0.58 / 28.38 ms │ 26.49 / 26.76 ±0.31 / 27.33 ms │ no change │
│ QQuery 82 │ 16.81 / 17.11 ±0.26 / 17.45 ms │ 16.73 / 17.03 ±0.23 / 17.33 ms │ no change │
│ QQuery 83 │ 34.86 / 35.03 ±0.16 / 35.31 ms │ 34.77 / 35.90 ±1.21 / 38.23 ms │ no change │
│ QQuery 84 │ 30.13 / 31.51 ±2.52 / 36.54 ms │ 30.02 / 30.16 ±0.16 / 30.38 ms │ no change │
│ QQuery 85 │ 104.88 / 107.34 ±3.21 / 113.54 ms │ 103.32 / 104.92 ±1.84 / 108.49 ms │ no change │
│ QQuery 86 │ 26.18 / 26.62 ±0.37 / 27.14 ms │ 25.87 / 26.85 ±1.17 / 29.12 ms │ no change │
│ QQuery 87 │ 64.81 / 65.24 ±0.34 / 65.55 ms │ 64.02 / 64.63 ±0.48 / 65.41 ms │ no change │
│ QQuery 88 │ 63.73 / 65.90 ±3.08 / 71.97 ms │ 62.79 / 64.39 ±1.97 / 68.18 ms │ no change │
│ QQuery 89 │ 35.84 / 36.49 ±0.66 / 37.73 ms │ 35.38 / 35.98 ±0.52 / 36.90 ms │ no change │
│ QQuery 90 │ 17.68 / 17.86 ±0.22 / 18.26 ms │ 17.80 / 19.00 ±1.64 / 22.20 ms │ 1.06x slower │
│ QQuery 91 │ 45.37 / 45.94 ±0.52 / 46.88 ms │ 46.84 / 47.99 ±0.95 / 49.69 ms │ no change │
│ QQuery 92 │ 29.78 / 30.69 ±0.77 / 31.87 ms │ 30.51 / 31.48 ±0.87 / 33.08 ms │ no change │
│ QQuery 93 │ 50.82 / 52.57 ±1.64 / 55.67 ms │ 52.44 / 53.59 ±1.20 / 55.74 ms │ no change │
│ QQuery 94 │ 39.00 / 39.62 ±0.41 / 40.02 ms │ 39.41 / 40.09 ±0.53 / 41.02 ms │ no change │
│ QQuery 95 │ 80.82 / 82.11 ±1.29 / 84.50 ms │ 83.37 / 86.11 ±1.95 / 89.02 ms │ no change │
│ QQuery 96 │ 23.82 / 23.98 ±0.16 / 24.25 ms │ 24.95 / 25.08 ±0.12 / 25.23 ms │ no change │
│ QQuery 97 │ 51.99 / 54.26 ±2.26 / 58.27 ms │ 55.20 / 57.14 ±1.20 / 58.97 ms │ 1.05x slower │
│ QQuery 98 │ 42.48 / 43.33 ±0.56 / 43.95 ms │ 44.36 / 45.90 ±1.77 / 49.17 ms │ 1.06x slower │
│ QQuery 99 │ 69.64 / 69.94 ±0.27 / 70.35 ms │ 71.73 / 73.06 ±1.04 / 74.80 ms │ no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 9547.35ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 9659.59ms │
│ Average Time (HEAD) │ 96.44ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 97.57ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 9 │
│ Queries with No Change │ 88 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 50.0s |
| Peak memory | 2.1 GiB |
| Avg memory | 1.4 GiB |
| CPU user | 209.2s |
| CPU sys | 5.7s |
| Peak spill | 0 B |
**tpcds — branch**
| Metric | Value |
|--------|-------|
| Wall time | 50.0s |
| Peak memory | 2.4 GiB |
| Avg memory | 1.7 GiB |
| CPU user | 209.0s |
| CPU sys | 5.7s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5608128222
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825052)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 792.69 ms │ 766.51 ms │ no change │
│ QQuery 1 │ 193.92 ms │ 194.83 ms │ no change │
│ QQuery 2 │ 464.31 ms │ 463.26 ms │ no change │
│ QQuery 3 │ 314.93 ms │ 314.80 ms │ no change │
│ QQuery 4 │ 1163.54 ms │ 1213.13 ms │ no change │
│ QQuery 5 │ 11344.25 ms │ 11137.76 ms │ no change │
│ QQuery 6 │ 2.71 ms │ 2.64 ms │ no change │
│ QQuery 7 │ 753.12 ms │ 713.00 ms │ +1.06x faster │
│ QQuery 8 │ 412.67 ms │ 424.94 ms │ no change │
│ QQuery 9 │ 3098.78 ms │ 2949.63 ms │ no change │
│ QQuery 10 │ 654.52 ms │ 641.13 ms │ no change │
│ QQuery 11 │ 2232.95 ms │ 2153.66 ms │ no change │
│ QQuery 12 │ 187.86 ms │ 190.86 ms │ no change │
│ QQuery 13 │ 546.80 ms │ 540.99 ms │ no change │
│ QQuery 14 │ 2814.59 ms │ 2723.45 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24977.62ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24430.58ms │
│ Average Time (HEAD) │ 1665.17ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1628.71ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 14 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 792.69 / 811.93 ±11.43 / 824.17 ms │ 766.51 / 814.39 ±28.51 / 842.72 ms │ no change │
│ QQuery 1 │ 193.92 / 194.57 ±0.65 / 195.82 ms │ 194.83 / 198.11 ±6.31 / 210.73 ms │ no change │
│ QQuery 2 │ 464.31 / 468.45 ±2.40 / 471.34 ms │ 463.26 / 465.03 ±1.84 / 468.11 ms │ no change │
│ QQuery 3 │ 314.93 / 316.11 ±0.97 / 317.62 ms │ 314.80 / 317.33 ±2.17 / 321.31 ms │ no change │
│ QQuery 4 │ 1163.54 / 1172.63 ±9.40 / 1190.52 ms │ 1213.13 / 1243.39 ±27.72 / 1294.47 ms │ 1.06x slower │
│ QQuery 5 │ 11344.25 / 11679.78 ±272.58 / 12080.60 ms │ 11137.76 / 11861.56 ±691.33 / 13042.99 ms │ no change │
│ QQuery 6 │ 2.71 / 3.00 ±0.34 / 3.64 ms │ 2.64 / 2.94 ±0.38 / 3.70 ms │ no change │
│ QQuery 7 │ 753.12 / 782.00 ±29.40 / 831.23 ms │ 713.00 / 794.68 ±56.94 / 869.02 ms │ no change │
│ QQuery 8 │ 412.67 / 432.39 ±17.91 / 455.30 ms │ 424.94 / 462.29 ±34.00 / 505.57 ms │ 1.07x slower │
│ QQuery 9 │ 3098.78 / 3197.84 ±89.66 / 3314.28 ms │ 2949.63 / 3100.09 ±90.37 / 3213.25 ms │ no change │
│ QQuery 10 │ 654.52 / 708.41 ±83.69 / 875.21 ms │ 641.13 / 683.30 ±43.03 / 758.94 ms │ no change │
│ QQuery 11 │ 2232.95 / 2273.54 ±25.35 / 2303.03 ms │ 2153.66 / 2236.02 ±75.83 / 2343.42 ms │ no change │
│ QQuery 12 │ 187.86 / 221.62 ±60.91 / 343.28 ms │ 190.86 / 215.93 ±45.83 / 307.45 ms │ no change │
│ QQuery 13 │ 546.80 / 577.67 ±29.61 / 633.57 ms │ 540.99 / 557.03 ±10.11 / 568.53 ms │ no change │
│ QQuery 14 │ 2814.59 / 2879.68 ±79.18 / 3029.40 ms │ 2723.45 / 2830.59 ±58.59 / 2894.79 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 25719.64ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 25782.69ms │
│ Average Time (HEAD) │ 1714.64ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1718.85ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.9 GiB |
| Avg memory | 5.3 GiB |
| CPU user | 1144.6s |
| CPU sys | 60.3s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.6 GiB |
| Avg memory | 5.3 GiB |
| CPU user | 1158.1s |
| CPU sys | 60.8s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5608122044
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825245)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 0 │ 772.66 ms │ 761.43 ms │ no change │
│ QQuery 1 │ 194.21 ms │ 194.11 ms │ no change │
│ QQuery 2 │ 464.01 ms │ 462.17 ms │ no change │
│ QQuery 3 │ 313.45 ms │ 313.62 ms │ no change │
│ QQuery 4 │ 1123.85 ms │ 1133.63 ms │ no change │
│ QQuery 5 │ 10532.80 ms │ 10623.00 ms │ no change │
│ QQuery 6 │ 2.66 ms │ 2.59 ms │ no change │
│ QQuery 7 │ 678.99 ms │ 667.71 ms │ no change │
│ QQuery 8 │ 416.82 ms │ 408.84 ms │ no change │
│ QQuery 9 │ 2997.11 ms │ 2961.76 ms │ no change │
│ QQuery 10 │ 629.78 ms │ 652.32 ms │ no change │
│ QQuery 11 │ 2168.68 ms │ 2199.99 ms │ no change │
│ QQuery 12 │ 186.88 ms │ 187.81 ms │ no change │
│ QQuery 13 │ 533.38 ms │ 546.84 ms │ no change │
│ QQuery 14 │ 2742.76 ms │ 2752.59 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23758.05ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23868.40ms │
│ Average Time (HEAD) │ 1583.87ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1591.23ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 15 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 772.66 / 780.63 ±7.53 / 794.42 ms │ 761.43 / 792.18 ±19.04 / 819.25 ms │ no change │
│ QQuery 1 │ 194.21 / 195.32 ±1.35 / 197.93 ms │ 194.11 / 201.68 ±9.22 / 216.79 ms │ no change │
│ QQuery 2 │ 464.01 / 466.25 ±2.06 / 469.75 ms │ 462.17 / 464.80 ±2.23 / 468.35 ms │ no change │
│ QQuery 3 │ 313.45 / 314.90 ±1.13 / 316.48 ms │ 313.62 / 316.10 ±2.27 / 320.14 ms │ no change │
│ QQuery 4 │ 1123.85 / 1148.61 ±16.33 / 1169.98 ms │ 1133.63 / 1163.01 ±20.91 / 1195.09 ms │ no change │
│ QQuery 5 │ 10532.80 / 11019.97 ±254.96 / 11235.71 ms │ 10623.00 / 10978.80 ±192.18 / 11164.27 ms │ no change │
│ QQuery 6 │ 2.66 / 2.97 ±0.33 / 3.56 ms │ 2.59 / 2.87 ±0.26 / 3.31 ms │ no change │
│ QQuery 7 │ 678.99 / 745.04 ±70.99 / 879.65 ms │ 667.71 / 717.12 ±51.42 / 811.60 ms │ no change │
│ QQuery 8 │ 416.82 / 427.36 ±16.59 / 460.45 ms │ 408.84 / 424.04 ±15.50 / 452.38 ms │ no change │
│ QQuery 9 │ 2997.11 / 3050.11 ±33.15 / 3097.43 ms │ 2961.76 / 3019.20 ±35.39 / 3060.94 ms │ no change │
│ QQuery 10 │ 629.78 / 686.55 ±91.43 / 868.44 ms │ 652.32 / 657.64 ±3.86 / 664.01 ms │ no change │
│ QQuery 11 │ 2168.68 / 2248.44 ±67.27 / 2364.90 ms │ 2199.99 / 2266.30 ±34.98 / 2300.98 ms │ no change │
│ QQuery 12 │ 186.88 / 224.09 ±64.74 / 353.20 ms │ 187.81 / 195.29 ±9.85 / 214.53 ms │ +1.15x faster │
│ QQuery 13 │ 533.38 / 554.14 ±13.28 / 575.06 ms │ 546.84 / 555.77 ±7.42 / 568.45 ms │ no change │
│ QQuery 14 │ 2742.76 / 2776.45 ±32.30 / 2824.68 ms │ 2752.59 / 2797.36 ±38.65 / 2862.40 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24640.85ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24552.17ms │
│ Average Time (HEAD) │ 1642.72ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1636.81ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 14 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 12.8 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1115.0s |
| CPU sys | 54.5s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 11.5 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1119.5s |
| CPU sys | 56.1s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767820250
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark external_aggr
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Q1(64.0 MB) │ 54.43 ms │ 54.11 ms │ no change │
│ Q1(32.0 MB) │ 50.20 ms │ 52.01 ms │ no change │
│ Q1(16.0 MB) │ 53.07 ms │ 52.76 ms │ no change │
│ Q2(512.0 MB) │ 277.40 ms │ 287.81 ms │ no change │
│ Q2(256.0 MB) │ 273.07 ms │ 254.85 ms │ +1.07x faster │
│ Q2(128.0 MB) │ 244.13 ms │ 241.41 ms │ no change │
│ Q2(64.0 MB) │ 242.31 ms │ 239.02 ms │ no change │
│ Q2(32.0 MB) │ 302.84 ms │ 300.01 ms │ no change │
└──────────────┴───────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1497.44ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 1481.99ms │
│ Average Time (HEAD) │ 187.18ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 185.25ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 7 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Q1(64.0 MB) │ 54.43 / 57.70 ±2.58 / 62.35 ms │ 54.11 / 56.99 ±3.89 / 64.66 ms │ no change │
│ Q1(32.0 MB) │ 50.20 / 52.79 ±1.44 / 54.34 ms │ 52.01 / 53.47 ±1.70 / 56.58 ms │ no change │
│ Q1(16.0 MB) │ 53.07 / 54.85 ±1.13 / 56.13 ms │ 52.76 / 54.40 ±1.30 / 55.75 ms │ no change │
│ Q2(512.0 MB) │ 277.40 / 287.96 ±7.88 / 300.11 ms │ 287.81 / 301.49 ±9.34 / 312.61 ms │ no change │
│ Q2(256.0 MB) │ 273.07 / 282.81 ±8.68 / 298.92 ms │ 254.85 / 278.94 ±13.32 / 294.19 ms │ no change │
│ Q2(128.0 MB) │ 244.13 / 256.20 ±7.41 / 262.67 ms │ 241.41 / 242.80 ±1.19 / 244.23 ms │ +1.06x faster │
│ Q2(64.0 MB) │ 242.31 / 252.69 ±12.01 / 275.73 ms │ 239.02 / 262.20 ±31.12 / 320.10 ms │ no change │
│ Q2(32.0 MB) │ 302.84 / 308.14 ±3.33 / 311.53 ms │ 300.01 / 304.20 ±3.39 / 309.50 ms │ no change │
└──────────────┴────────────────────────────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1553.15ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 1554.49ms │
│ Average Time (HEAD) │ 194.14ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 194.31ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 7 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
```
</p>
</details>
<details><summary>Memory Pool Peaks</summary>
Peak `MemoryPool` reservation per query — what DataFusion's accounting believes it reserved. Recorded only when the benchmark runs with `DATAFUSION_RUNTIME_MEMORY_LIMIT` set.
Base: `0576a0b (merge-base)` | Changed: `claude/datafusion-issue-25116-75677f`
**`external_aggr`**
| Query | Base | Changed | Change |
| --- | --- | --- | --- |
| 1(64.0 MB) | 36.8 MiB | 36.8 MiB | +0.0% |
| 1(32.0 MB) | 19.5 MiB | 18.6 MiB | -4.5% |
| 1(16.0 MB) | 10.2 MiB | 10.2 MiB | -0.6% |
| 2(512.0 MB) | 136.6 MiB | 138.5 MiB | +1.3% |
| 2(256.0 MB) | 97.9 MiB | 97.8 MiB | -0.1% |
| 2(128.0 MB) | 49.2 MiB | 49.1 MiB | -0.1% |
| 2(64.0 MB) | 32.5 MiB | 34.3 MiB | +5.5% |
| 2(32.0 MB) | 18.0 MiB | 18.4 MiB | +2.1% |
**Pool accounting vs. process RSS**
Max pool peak is the largest reservation any single query in the run reached; peak RSS covers the whole invocation, including data loading and allocator retention, and the two high-water marks need not coincide in time. The gap is therefore an upper bound on what the pool did not account for, not a measurement of it.
| Benchmark | Side | Max pool peak | Peak RSS | Gap | RSS / pool |
| --- | --- | --- | --- | --- | --- |
| external_aggr | base (`0576a0b (merge-base)`) | 136.6 MiB | 445.1 MiB | 308.5 MiB | 3.3× |
| external_aggr | changed (`claude/datafusion-issue-25116-75677f`) | 138.5 MiB | 441.5 MiB | 303.0 MiB | 3.2× |
</details>
<details><summary>Resource Usage</summary>
**external_aggr — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 90.0s |
| Peak memory | 445.1 MiB |
| Avg memory | 35.6 MiB |
| CPU user | 25.6s |
| CPU sys | 3.8s |
| Peak spill | 0 B |
**external_aggr — branch**
| Metric | Value |
|--------|-------|
| Wall time | 85.0s |
| Peak memory | 441.5 MiB |
| Avg memory | 35.8 MiB |
| CPU user | 25.4s |
| CPU sys | 3.9s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767720955
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676204)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767676204-2564-x9qgw 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 0576a0b400437ade5a6f3b102465d954a539f263 (0576a0b400437ade5a6f3b102465d954a539f263) to 0576a0b400437ade5a6f3b102465d954a539f263 [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..0576a0b400437ade5a6f3b102465d954a539f263)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
changed:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767675750
run benchmark clickbench_extended
baseline:
ref: 0576a0b400437ade5a6f3b102465d954a539f263
changed:
ref: 0576a0b400437ade5a6f3b102465d954a539f263
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5776111362
yep, thank you for the great feedback. should i go ahead and merge it?
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on code in PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#discussion_r4072830828
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -235,14 +235,23 @@ impl GroupsAccumulatorAdapter {
assert_eq!(values[0].len(), group_indices.len());
+ // groups_with_rows holds a list of group indexes that have any rows
+ // that need to be accumulated, stored in order of first appearance in
+ // this batch
+ let mut groups_with_rows = vec![];
Review Comment:
Filed into https://github.com/apache/datafusion/issues/25604
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767675578
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767847535
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.24 ms │ 1.38 ms │ 1.11x slower │
│ QQuery 1 │ 11.79 ms │ 12.20 ms │ no change │
│ QQuery 2 │ 36.64 ms │ 37.45 ms │ no change │
│ QQuery 3 │ 30.89 ms │ 32.75 ms │ 1.06x slower │
│ QQuery 4 │ 230.95 ms │ 276.31 ms │ 1.20x slower │
│ QQuery 5 │ 272.42 ms │ 268.51 ms │ no change │
│ QQuery 6 │ 1.37 ms │ 1.36 ms │ no change │
│ QQuery 7 │ 13.68 ms │ 13.52 ms │ no change │
│ QQuery 8 │ 336.82 ms │ 328.50 ms │ no change │
│ QQuery 9 │ 476.40 ms │ 460.12 ms │ no change │
│ QQuery 10 │ 68.89 ms │ 63.29 ms │ +1.09x faster │
│ QQuery 11 │ 75.88 ms │ 73.91 ms │ no change │
│ QQuery 12 │ 265.43 ms │ 258.49 ms │ no change │
│ QQuery 13 │ 356.44 ms │ 365.91 ms │ no change │
│ QQuery 14 │ 275.21 ms │ 273.51 ms │ no change │
│ QQuery 15 │ 279.39 ms │ 290.99 ms │ no change │
│ QQuery 16 │ 610.40 ms │ 604.88 ms │ no change │
│ QQuery 17 │ 618.59 ms │ 625.98 ms │ no change │
│ QQuery 18 │ 1260.96 ms │ 1260.64 ms │ no change │
│ QQuery 19 │ 27.05 ms │ 27.66 ms │ no change │
│ QQuery 20 │ 512.71 ms │ 521.45 ms │ no change │
│ QQuery 21 │ 508.87 ms │ 515.42 ms │ no change │
│ QQuery 22 │ 1005.98 ms │ 1021.58 ms │ no change │
│ QQuery 23 │ 3191.74 ms │ 3254.98 ms │ no change │
│ QQuery 24 │ 40.49 ms │ 40.38 ms │ no change │
│ QQuery 25 │ 105.32 ms │ 104.66 ms │ no change │
│ QQuery 26 │ 41.18 ms │ 41.51 ms │ no change │
│ QQuery 27 │ 511.30 ms │ 505.94 ms │ no change │
│ QQuery 28 │ 2929.78 ms │ 2894.46 ms │ no change │
│ QQuery 29 │ 41.24 ms │ 41.58 ms │ no change │
│ QQuery 30 │ 299.92 ms │ 312.01 ms │ no change │
│ QQuery 31 │ 278.63 ms │ 295.19 ms │ 1.06x slower │
│ QQuery 32 │ 1145.39 ms │ 1082.28 ms │ +1.06x faster │
│ QQuery 33 │ 1560.38 ms │ 1526.48 ms │ no change │
│ QQuery 34 │ 1595.04 ms │ 1550.48 ms │ no change │
│ QQuery 35 │ 375.69 ms │ 293.57 ms │ +1.28x faster │
│ QQuery 36 │ 64.72 ms │ 67.13 ms │ no change │
│ QQuery 37 │ 34.98 ms │ 35.23 ms │ no change │
│ QQuery 38 │ 45.03 ms │ 42.13 ms │ +1.07x faster │
│ QQuery 39 │ 146.52 ms │ 134.34 ms │ +1.09x faster │
│ QQuery 40 │ 15.50 ms │ 13.99 ms │ +1.11x faster │
│ QQuery 41 │ 14.52 ms │ 13.51 ms │ +1.07x faster │
│ QQuery 42 │ 13.34 ms │ 13.39 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19728.70ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19599.04ms │
│ Average Time (HEAD) │ 458.81ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 455.79ms │
│ Queries Faster │ 7 │
│ Queries Slower │ 4 │
│ Queries with No Change │ 32 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.24 / 4.00 ±5.43 / 14.86 ms │ 1.38 / 4.43 ±5.87 / 16.18 ms │ 1.11x slower │
│ QQuery 1 │ 11.79 / 12.07 ±0.43 / 12.92 ms │ 12.20 / 12.47 ±0.23 / 12.83 ms │ no change │
│ QQuery 2 │ 36.64 / 36.98 ±0.28 / 37.42 ms │ 37.45 / 37.57 ±0.13 / 37.81 ms │ no change │
│ QQuery 3 │ 30.89 / 31.58 ±0.81 / 33.11 ms │ 32.75 / 33.01 ±0.38 / 33.77 ms │ no change │
│ QQuery 4 │ 230.95 / 235.95 ±2.72 / 238.45 ms │ 276.31 / 284.93 ±8.11 / 294.81 ms │ 1.21x slower │
│ QQuery 5 │ 272.42 / 287.67 ±12.36 / 300.05 ms │ 268.51 / 281.24 ±9.33 / 292.08 ms │ no change │
│ QQuery 6 │ 1.37 / 1.52 ±0.22 / 1.96 ms │ 1.36 / 1.52 ±0.23 / 1.98 ms │ no change │
│ QQuery 7 │ 13.68 / 13.75 ±0.06 / 13.85 ms │ 13.52 / 13.84 ±0.28 / 14.34 ms │ no change │
│ QQuery 8 │ 336.82 / 341.88 ±4.61 / 348.80 ms │ 328.50 / 336.28 ±11.83 / 359.80 ms │ no change │
│ QQuery 9 │ 476.40 / 500.38 ±26.77 / 543.86 ms │ 460.12 / 468.43 ±6.74 / 477.55 ms │ +1.07x faster │
│ QQuery 10 │ 68.89 / 75.10 ±9.66 / 94.03 ms │ 63.29 / 64.00 ±0.67 / 65.14 ms │ +1.17x faster │
│ QQuery 11 │ 75.88 / 79.17 ±2.03 / 82.13 ms │ 73.91 / 74.67 ±0.45 / 75.14 ms │ +1.06x faster │
│ QQuery 12 │ 265.43 / 273.14 ±6.65 / 284.93 ms │ 258.49 / 262.61 ±3.23 / 267.48 ms │ no change │
│ QQuery 13 │ 356.44 / 372.34 ±11.73 / 392.23 ms │ 365.91 / 378.70 ±10.01 / 391.50 ms │ no change │
│ QQuery 14 │ 275.21 / 281.98 ±5.73 / 290.50 ms │ 273.51 / 282.82 ±5.29 / 287.78 ms │ no change │
│ QQuery 15 │ 279.39 / 288.85 ±10.69 / 303.15 ms │ 290.99 / 304.20 ±7.48 / 313.41 ms │ 1.05x slower │
│ QQuery 16 │ 610.40 / 620.81 ±6.45 / 627.19 ms │ 604.88 / 633.95 ±32.40 / 696.45 ms │ no change │
│ QQuery 17 │ 618.59 / 654.89 ±35.16 / 707.58 ms │ 625.98 / 671.12 ±36.23 / 716.49 ms │ no change │
│ QQuery 18 │ 1260.96 / 1300.00 ±39.15 / 1373.60 ms │ 1260.64 / 1294.23 ±26.26 / 1329.07 ms │ no change │
│ QQuery 19 │ 27.05 / 27.52 ±0.30 / 27.99 ms │ 27.66 / 28.11 ±0.42 / 28.81 ms │ no change │
│ QQuery 20 │ 512.71 / 524.04 ±11.45 / 545.86 ms │ 521.45 / 539.94 ±13.39 / 560.89 ms │ no change │
│ QQuery 21 │ 508.87 / 512.67 ±4.91 / 521.31 ms │ 515.42 / 540.29 ±19.24 / 573.62 ms │ 1.05x slower │
│ QQuery 22 │ 1005.98 / 1047.43 ±33.30 / 1089.18 ms │ 1021.58 / 1068.18 ±50.05 / 1161.36 ms │ no change │
│ QQuery 23 │ 3191.74 / 3390.92 ±151.04 / 3655.38 ms │ 3254.98 / 3492.11 ±227.14 / 3864.73 ms │ no change │
│ QQuery 24 │ 40.49 / 91.01 ±41.23 / 127.49 ms │ 40.38 / 40.98 ±0.62 / 42.17 ms │ +2.22x faster │
│ QQuery 25 │ 105.32 / 147.81 ±80.56 / 308.88 ms │ 104.66 / 197.46 ±177.25 / 551.88 ms │ 1.34x slower │
│ QQuery 26 │ 41.18 / 42.76 ±1.19 / 44.30 ms │ 41.51 / 41.85 ±0.44 / 42.68 ms │ no change │
│ QQuery 27 │ 511.30 / 520.74 ±7.06 / 530.93 ms │ 505.94 / 529.73 ±14.22 / 550.42 ms │ no change │
│ QQuery 28 │ 2929.78 / 2946.17 ±18.78 / 2980.98 ms │ 2894.46 / 2987.12 ±53.37 / 3056.90 ms │ no change │
│ QQuery 29 │ 41.24 / 41.69 ±0.48 / 42.58 ms │ 41.58 / 73.62 ±61.80 / 197.19 ms │ 1.77x slower │
│ QQuery 30 │ 299.92 / 316.17 ±13.46 / 339.52 ms │ 312.01 / 323.95 ±7.40 / 333.53 ms │ no change │
│ QQuery 31 │ 278.63 / 284.46 ±3.56 / 289.16 ms │ 295.19 / 327.90 ±22.96 / 355.62 ms │ 1.15x slower │
│ QQuery 32 │ 1145.39 / 1354.27 ±186.63 / 1678.58 ms │ 1082.28 / 1180.80 ±107.74 / 1384.65 ms │ +1.15x faster │
│ QQuery 33 │ 1560.38 / 1604.25 ±43.32 / 1672.24 ms │ 1526.48 / 1629.40 ±96.22 / 1807.33 ms │ no change │
│ QQuery 34 │ 1595.04 / 1873.00 ±184.11 / 2168.82 ms │ 1550.48 / 1623.88 ±66.33 / 1742.11 ms │ +1.15x faster │
│ QQuery 35 │ 375.69 / 483.94 ±200.57 / 884.78 ms │ 293.57 / 354.94 ±74.66 / 487.20 ms │ +1.36x faster │
│ QQuery 36 │ 64.72 / 72.50 ±6.89 / 82.72 ms │ 67.13 / 71.27 ±3.56 / 77.32 ms │ no change │
│ QQuery 37 │ 34.98 / 47.83 ±21.72 / 91.10 ms │ 35.23 / 37.81 ±4.99 / 47.79 ms │ +1.27x faster │
│ QQuery 38 │ 45.03 / 48.31 ±3.51 / 54.33 ms │ 42.13 / 51.50 ±9.22 / 65.17 ms │ 1.07x slower │
│ QQuery 39 │ 146.52 / 155.24 ±8.37 / 170.61 ms │ 134.34 / 146.24 ±8.49 / 157.35 ms │ +1.06x faster │
│ QQuery 40 │ 15.50 / 16.44 ±0.80 / 17.73 ms │ 13.99 / 15.21 ±1.70 / 18.51 ms │ +1.08x faster │
│ QQuery 41 │ 14.52 / 14.82 ±0.22 / 15.17 ms │ 13.51 / 13.91 ±0.46 / 14.78 ms │ +1.07x faster │
│ QQuery 42 │ 13.34 / 16.09 ±4.58 / 25.22 ms │ 13.39 / 19.39 ±7.39 / 31.87 ms │ 1.21x slower │
└───────────┴────────────────────────────────────────┴────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 20992.16ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 20775.62ms │
│ Average Time (HEAD) │ 488.19ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 483.15ms │
│ Queries Faster │ 11 │
│ Queries Slower │ 9 │
│ Queries with No Change │ 23 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 110.0s |
| Peak memory | 18.7 GiB |
| Avg memory | 5.8 GiB |
| CPU user | 1020.4s |
| CPU sys | 103.7s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 18.0 GiB |
| Avg memory | 6.4 GiB |
| CPU user | 1012.0s |
| CPU sys | 101.7s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767896170
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676730)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 0576a0b400437ade5a6f3b102465d954a539f263 (0576a0b400437ade5a6f3b102465d954a539f263) to 0576a0b400437ade5a6f3b102465d954a539f263 [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..0576a0b400437ade5a6f3b102465d954a539f263)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
changed:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 829.83 ms │ 929.52 ms │ 1.12x slower │
│ QQuery 1 │ 172.96 ms │ 173.57 ms │ no change │
│ QQuery 2 │ 444.81 ms │ 450.87 ms │ no change │
│ QQuery 3 │ 327.42 ms │ 330.34 ms │ no change │
│ QQuery 4 │ 1229.64 ms │ 1258.88 ms │ no change │
│ QQuery 5 │ 11249.16 ms │ 10859.19 ms │ no change │
│ QQuery 6 │ 3.03 ms │ 3.16 ms │ no change │
│ QQuery 7 │ 890.17 ms │ 902.05 ms │ no change │
│ QQuery 8 │ 473.75 ms │ 457.58 ms │ no change │
│ QQuery 9 │ 2966.02 ms │ 2944.41 ms │ no change │
│ QQuery 10 │ 655.01 ms │ 658.59 ms │ no change │
│ QQuery 11 │ 2206.28 ms │ 2197.62 ms │ no change │
│ QQuery 12 │ 200.44 ms │ 199.87 ms │ no change │
│ QQuery 13 │ 557.44 ms │ 560.23 ms │ no change │
│ QQuery 14 │ 579.53 ms │ 629.58 ms │ 1.09x slower │
└───────────┴─────────────┴──────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22785.49ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22555.47ms │
│ Average Time (HEAD) │ 1519.03ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1503.70ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 829.83 / 877.96 ±31.72 / 910.07 ms │ 929.52 / 960.90 ±26.94 / 993.45 ms │ 1.09x slower │
│ QQuery 1 │ 172.96 / 174.89 ±2.61 / 180.02 ms │ 173.57 / 177.81 ±4.85 / 185.61 ms │ no change │
│ QQuery 2 │ 444.81 / 450.07 ±5.20 / 459.79 ms │ 450.87 / 453.49 ±1.71 / 456.17 ms │ no change │
│ QQuery 3 │ 327.42 / 331.58 ±2.65 / 334.80 ms │ 330.34 / 333.94 ±2.13 / 336.93 ms │ no change │
│ QQuery 4 │ 1229.64 / 1254.41 ±17.89 / 1277.40 ms │ 1258.88 / 1280.91 ±17.42 / 1298.05 ms │ no change │
│ QQuery 5 │ 11249.16 / 11670.02 ±282.51 / 12051.44 ms │ 10859.19 / 11452.58 ±396.37 / 11911.24 ms │ no change │
│ QQuery 6 │ 3.03 / 3.20 ±0.19 / 3.57 ms │ 3.16 / 3.39 ±0.25 / 3.78 ms │ 1.06x slower │
│ QQuery 7 │ 890.17 / 914.38 ±19.69 / 940.57 ms │ 902.05 / 939.34 ±27.64 / 977.67 ms │ no change │
│ QQuery 8 │ 473.75 / 489.50 ±17.62 / 516.39 ms │ 457.58 / 467.68 ±9.18 / 482.04 ms │ no change │
│ QQuery 9 │ 2966.02 / 3012.26 ±46.68 / 3089.74 ms │ 2944.41 / 3024.72 ±84.71 / 3174.06 ms │ no change │
│ QQuery 10 │ 655.01 / 667.27 ±13.05 / 692.12 ms │ 658.59 / 700.26 ±57.09 / 812.90 ms │ no change │
│ QQuery 11 │ 2206.28 / 2287.04 ±59.69 / 2392.15 ms │ 2197.62 / 2271.94 ±43.19 / 2317.85 ms │ no change │
│ QQuery 12 │ 200.44 / 221.90 ±38.07 / 297.92 ms │ 199.87 / 230.77 ±51.18 / 332.80 ms │ no change │
│ QQuery 13 │ 557.44 / 575.69 ±18.49 / 609.82 ms │ 560.23 / 569.03 ±6.06 / 579.23 ms │ no change │
│ QQuery 14 │ 579.53 / 628.54 ±31.65 / 672.82 ms │ 629.58 / 680.33 ±36.97 / 741.65 ms │ 1.08x slower │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23558.72ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23547.11ms │
│ Average Time (HEAD) │ 1570.58ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1569.81ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 12 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 12.0 GiB |
| Avg memory | 5.0 GiB |
| CPU user | 1076.9s |
| CPU sys | 53.1s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 10.9 GiB |
| Avg memory | 5.0 GiB |
| CPU user | 1084.4s |
| CPU sys | 56.0s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5608105014
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825682)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 759.12 ms │ 770.99 ms │ no change │
│ QQuery 1 │ 194.40 ms │ 189.79 ms │ no change │
│ QQuery 2 │ 461.71 ms │ 510.47 ms │ 1.11x slower │
│ QQuery 3 │ 313.32 ms │ 315.77 ms │ no change │
│ QQuery 4 │ 1131.71 ms │ 1132.83 ms │ no change │
│ QQuery 5 │ 10617.39 ms │ 10005.41 ms │ +1.06x faster │
│ QQuery 6 │ 2.60 ms │ 2.59 ms │ no change │
│ QQuery 7 │ 695.37 ms │ 673.45 ms │ no change │
│ QQuery 8 │ 411.47 ms │ 415.34 ms │ no change │
│ QQuery 9 │ 2979.69 ms │ 2932.11 ms │ no change │
│ QQuery 10 │ 629.65 ms │ 626.06 ms │ no change │
│ QQuery 11 │ 2240.90 ms │ 2219.33 ms │ no change │
│ QQuery 12 │ 186.87 ms │ 185.86 ms │ no change │
│ QQuery 13 │ 533.66 ms │ 530.46 ms │ no change │
│ QQuery 14 │ 2674.22 ms │ 2240.20 ms │ +1.19x faster │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23832.07ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22750.66ms │
│ Average Time (HEAD) │ 1588.80ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1516.71ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 12 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 759.12 / 775.79 ±8.85 / 783.68 ms │ 770.99 / 778.00 ±4.39 / 783.14 ms │ no change │
│ QQuery 1 │ 194.40 / 194.90 ±0.29 / 195.24 ms │ 189.79 / 193.94 ±7.74 / 209.41 ms │ no change │
│ QQuery 2 │ 461.71 / 466.12 ±3.08 / 470.92 ms │ 510.47 / 514.87 ±3.30 / 520.03 ms │ 1.10x slower │
│ QQuery 3 │ 313.32 / 314.36 ±0.83 / 315.52 ms │ 315.77 / 317.66 ±2.03 / 321.36 ms │ no change │
│ QQuery 4 │ 1131.71 / 1160.09 ±15.96 / 1180.82 ms │ 1132.83 / 1151.24 ±13.10 / 1173.00 ms │ no change │
│ QQuery 5 │ 10617.39 / 11275.35 ±415.36 / 11839.92 ms │ 10005.41 / 10620.55 ±342.33 / 11029.05 ms │ +1.06x faster │
│ QQuery 6 │ 2.60 / 2.86 ±0.32 / 3.47 ms │ 2.59 / 2.79 ±0.28 / 3.35 ms │ no change │
│ QQuery 7 │ 695.37 / 745.57 ±37.28 / 803.94 ms │ 673.45 / 719.48 ±38.23 / 780.05 ms │ no change │
│ QQuery 8 │ 411.47 / 427.73 ±20.17 / 465.06 ms │ 415.34 / 427.30 ±9.27 / 435.61 ms │ no change │
│ QQuery 9 │ 2979.69 / 3046.91 ±56.50 / 3135.52 ms │ 2932.11 / 3015.93 ±54.76 / 3099.08 ms │ no change │
│ QQuery 10 │ 629.65 / 655.11 ±20.91 / 683.39 ms │ 626.06 / 688.56 ±96.04 / 879.16 ms │ 1.05x slower │
│ QQuery 11 │ 2240.90 / 2257.85 ±19.26 / 2286.40 ms │ 2219.33 / 2248.61 ±25.34 / 2289.91 ms │ no change │
│ QQuery 12 │ 186.87 / 195.43 ±13.16 / 221.30 ms │ 185.86 / 221.96 ±66.63 / 355.14 ms │ 1.14x slower │
│ QQuery 13 │ 533.66 / 544.86 ±6.29 / 553.01 ms │ 530.46 / 548.51 ±13.06 / 569.71 ms │ no change │
│ QQuery 14 │ 2674.22 / 2733.34 ±50.46 / 2803.73 ms │ 2240.20 / 2299.75 ±36.28 / 2346.56 ms │ +1.19x faster │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24796.29ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23749.15ms │
│ Average Time (HEAD) │ 1653.09ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1583.28ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 10 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 11.2 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1115.7s |
| CPU sys | 56.0s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 11.3 GiB |
| Avg memory | 5.0 GiB |
| CPU user | 1067.0s |
| CPU sys | 53.7s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605613135
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 783.31 ms │ 757.62 ms │ no change │
│ QQuery 1 │ 194.61 ms │ 189.27 ms │ no change │
│ QQuery 2 │ 459.13 ms │ 512.92 ms │ 1.12x slower │
│ QQuery 3 │ 314.12 ms │ 314.80 ms │ no change │
│ QQuery 4 │ 1137.90 ms │ 1117.75 ms │ no change │
│ QQuery 5 │ 11099.83 ms │ 10615.10 ms │ no change │
│ QQuery 6 │ 2.77 ms │ 2.58 ms │ +1.07x faster │
│ QQuery 7 │ 733.70 ms │ 667.76 ms │ +1.10x faster │
│ QQuery 8 │ 414.07 ms │ 418.55 ms │ no change │
│ QQuery 9 │ 2967.17 ms │ 2988.42 ms │ no change │
│ QQuery 10 │ 624.24 ms │ 643.00 ms │ no change │
│ QQuery 11 │ 2276.36 ms │ 2299.17 ms │ no change │
│ QQuery 12 │ 202.63 ms │ 188.90 ms │ +1.07x faster │
│ QQuery 13 │ 573.06 ms │ 546.76 ms │ no change │
│ QQuery 14 │ 2637.70 ms │ 2331.57 ms │ +1.13x faster │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24420.57ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23594.17ms │
│ Average Time (HEAD) │ 1628.04ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1572.94ms │
│ Queries Faster │ 4 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 10 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 783.31 / 895.13 ±71.85 / 984.09 ms │ 757.62 / 817.73 ±60.21 / 928.58 ms │ +1.09x faster │
│ QQuery 1 │ 194.61 / 195.22 ±0.96 / 197.13 ms │ 189.27 / 190.03 ±0.48 / 190.64 ms │ no change │
│ QQuery 2 │ 459.13 / 462.90 ±2.40 / 466.61 ms │ 512.92 / 520.64 ±7.37 / 530.67 ms │ 1.12x slower │
│ QQuery 3 │ 314.12 / 315.27 ±1.31 / 317.73 ms │ 314.80 / 317.38 ±2.21 / 320.69 ms │ no change │
│ QQuery 4 │ 1137.90 / 1179.10 ±25.13 / 1203.05 ms │ 1117.75 / 1129.00 ±12.77 / 1150.78 ms │ no change │
│ QQuery 5 │ 11099.83 / 11626.79 ±314.08 / 11937.69 ms │ 10615.10 / 10977.58 ±253.20 / 11344.85 ms │ +1.06x faster │
│ QQuery 6 │ 2.77 / 3.01 ±0.42 / 3.85 ms │ 2.58 / 2.84 ±0.34 / 3.50 ms │ +1.06x faster │
│ QQuery 7 │ 733.70 / 785.66 ±33.53 / 836.33 ms │ 667.76 / 696.42 ±19.53 / 714.35 ms │ +1.13x faster │
│ QQuery 8 │ 414.07 / 423.98 ±9.03 / 438.09 ms │ 418.55 / 429.50 ±11.68 / 446.73 ms │ no change │
│ QQuery 9 │ 2967.17 / 3027.53 ±68.42 / 3154.92 ms │ 2988.42 / 3043.21 ±52.91 / 3128.10 ms │ no change │
│ QQuery 10 │ 624.24 / 643.83 ±14.02 / 664.45 ms │ 643.00 / 652.30 ±6.77 / 661.53 ms │ no change │
│ QQuery 11 │ 2276.36 / 2332.74 ±89.15 / 2509.43 ms │ 2299.17 / 2346.14 ±39.51 / 2388.29 ms │ no change │
│ QQuery 12 │ 202.63 / 215.81 ±20.73 / 257.12 ms │ 188.90 / 224.98 ±65.80 / 356.47 ms │ no change │
│ QQuery 13 │ 573.06 / 575.59 ±2.96 / 581.15 ms │ 546.76 / 551.04 ±7.48 / 565.91 ms │ no change │
│ QQuery 14 │ 2637.70 / 2946.31 ±184.16 / 3158.77 ms │ 2331.57 / 2469.08 ±102.59 / 2628.38 ms │ +1.19x faster │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 25628.86ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24367.88ms │
│ Average Time (HEAD) │ 1708.59ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1624.53ms │
│ Queries Faster │ 5 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 9 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.7 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1140.7s |
| CPU sys | 60.0s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 11.4 GiB |
| Avg memory | 4.9 GiB |
| CPU user | 1087.0s |
| CPU sys | 57.8s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607779662
run benchmark clickbench_extended
baseline:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
changed:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611689371
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481966)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 773.03 ms │ 769.66 ms │ no change │
│ QQuery 1 │ 195.48 ms │ 195.05 ms │ no change │
│ QQuery 2 │ 458.60 ms │ 470.31 ms │ no change │
│ QQuery 3 │ 313.67 ms │ 317.55 ms │ no change │
│ QQuery 4 │ 1137.15 ms │ 1162.00 ms │ no change │
│ QQuery 5 │ 11088.98 ms │ 10395.01 ms │ +1.07x faster │
│ QQuery 6 │ 2.53 ms │ 3.12 ms │ 1.23x slower │
│ QQuery 7 │ 663.69 ms │ 785.96 ms │ 1.18x slower │
│ QQuery 8 │ 407.26 ms │ 454.54 ms │ 1.12x slower │
│ QQuery 9 │ 3114.61 ms │ 3061.22 ms │ no change │
│ QQuery 10 │ 630.99 ms │ 638.00 ms │ no change │
│ QQuery 11 │ 2106.57 ms │ 2202.45 ms │ no change │
│ QQuery 12 │ 196.62 ms │ 187.84 ms │ no change │
│ QQuery 13 │ 539.64 ms │ 539.10 ms │ no change │
│ QQuery 14 │ 2712.72 ms │ 2396.93 ms │ +1.13x faster │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24341.53ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23578.75ms │
│ Average Time (HEAD) │ 1622.77ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1571.92ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 10 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 773.03 / 800.28 ±32.82 / 864.32 ms │ 769.66 / 778.31 ±5.63 / 784.71 ms │ no change │
│ QQuery 1 │ 195.48 / 199.86 ±5.71 / 210.63 ms │ 195.05 / 195.92 ±0.88 / 197.58 ms │ no change │
│ QQuery 2 │ 458.60 / 460.81 ±1.33 / 462.51 ms │ 470.31 / 475.67 ±4.95 / 483.51 ms │ no change │
│ QQuery 3 │ 313.67 / 315.09 ±1.15 / 316.91 ms │ 317.55 / 319.84 ±1.37 / 321.68 ms │ no change │
│ QQuery 4 │ 1137.15 / 1155.66 ±11.23 / 1165.01 ms │ 1162.00 / 1289.12 ±74.60 / 1359.79 ms │ 1.12x slower │
│ QQuery 5 │ 11088.98 / 11779.06 ±363.95 / 12067.96 ms │ 10395.01 / 10904.60 ±485.22 / 11519.62 ms │ +1.08x faster │
│ QQuery 6 │ 2.53 / 2.82 ±0.39 / 3.59 ms │ 3.12 / 3.39 ±0.27 / 3.88 ms │ 1.20x slower │
│ QQuery 7 │ 663.69 / 742.92 ±70.67 / 856.38 ms │ 785.96 / 857.51 ±64.66 / 975.79 ms │ 1.15x slower │
│ QQuery 8 │ 407.26 / 427.80 ±34.69 / 497.03 ms │ 454.54 / 463.97 ±6.18 / 471.74 ms │ 1.08x slower │
│ QQuery 9 │ 3114.61 / 3220.73 ±104.56 / 3404.92 ms │ 3061.22 / 3104.45 ±27.15 / 3141.62 ms │ no change │
│ QQuery 10 │ 630.99 / 716.94 ±159.88 / 1036.61 ms │ 638.00 / 686.27 ±56.58 / 791.54 ms │ no change │
│ QQuery 11 │ 2106.57 / 2248.66 ±94.42 / 2391.66 ms │ 2202.45 / 2301.67 ±93.44 / 2475.15 ms │ no change │
│ QQuery 12 │ 196.62 / 212.59 ±24.83 / 261.85 ms │ 187.84 / 201.54 ±10.66 / 212.27 ms │ +1.05x faster │
│ QQuery 13 │ 539.64 / 559.29 ±10.22 / 567.16 ms │ 539.10 / 553.86 ±11.22 / 565.88 ms │ no change │
│ QQuery 14 │ 2712.72 / 2812.29 ±77.72 / 2925.45 ms │ 2396.93 / 2505.13 ±82.23 / 2650.62 ms │ +1.12x faster │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 25654.80ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24641.27ms │
│ Average Time (HEAD) │ 1710.32ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1642.75ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 4 │
│ Queries with No Change │ 8 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.9 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1132.3s |
| CPU sys | 59.9s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 11.2 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1102.8s |
| CPU sys | 61.5s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611689710
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 787.17 ms │ 796.85 ms │ no change │
│ QQuery 1 │ 193.98 ms │ 193.51 ms │ no change │
│ QQuery 2 │ 459.41 ms │ 466.08 ms │ no change │
│ QQuery 3 │ 313.64 ms │ 314.88 ms │ no change │
│ QQuery 4 │ 1143.64 ms │ 1150.20 ms │ no change │
│ QQuery 5 │ 11022.31 ms │ 10422.94 ms │ +1.06x faster │
│ QQuery 6 │ 2.63 ms │ 2.70 ms │ no change │
│ QQuery 7 │ 688.63 ms │ 678.65 ms │ no change │
│ QQuery 8 │ 409.92 ms │ 423.42 ms │ no change │
│ QQuery 9 │ 2920.16 ms │ 3062.95 ms │ no change │
│ QQuery 10 │ 630.08 ms │ 640.32 ms │ no change │
│ QQuery 11 │ 2242.31 ms │ 2158.98 ms │ no change │
│ QQuery 12 │ 189.12 ms │ 187.35 ms │ no change │
│ QQuery 13 │ 540.84 ms │ 545.00 ms │ no change │
│ QQuery 14 │ 2721.65 ms │ 2387.31 ms │ +1.14x faster │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24265.50ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23431.12ms │
│ Average Time (HEAD) │ 1617.70ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1562.07ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 787.17 / 799.02 ±7.30 / 809.99 ms │ 796.85 / 847.69 ±75.35 / 994.85 ms │ 1.06x slower │
│ QQuery 1 │ 193.98 / 194.80 ±0.47 / 195.31 ms │ 193.51 / 194.76 ±1.63 / 197.92 ms │ no change │
│ QQuery 2 │ 459.41 / 462.47 ±3.04 / 468.17 ms │ 466.08 / 468.50 ±3.44 / 475.17 ms │ no change │
│ QQuery 3 │ 313.64 / 314.70 ±1.25 / 317.09 ms │ 314.88 / 316.00 ±1.22 / 318.32 ms │ no change │
│ QQuery 4 │ 1143.64 / 1204.29 ±74.38 / 1345.89 ms │ 1150.20 / 1174.45 ±32.27 / 1238.24 ms │ no change │
│ QQuery 5 │ 11022.31 / 11399.61 ±344.33 / 11868.41 ms │ 10422.94 / 10966.35 ±323.63 / 11353.25 ms │ no change │
│ QQuery 6 │ 2.63 / 2.92 ±0.28 / 3.43 ms │ 2.70 / 2.92 ±0.32 / 3.53 ms │ no change │
│ QQuery 7 │ 688.63 / 723.95 ±21.49 / 744.43 ms │ 678.65 / 749.65 ±50.29 / 835.11 ms │ no change │
│ QQuery 8 │ 409.92 / 441.95 ±56.83 / 555.43 ms │ 423.42 / 440.77 ±15.32 / 463.73 ms │ no change │
│ QQuery 9 │ 2920.16 / 3183.28 ±164.13 / 3320.62 ms │ 3062.95 / 3194.88 ±71.25 / 3275.07 ms │ no change │
│ QQuery 10 │ 630.08 / 661.21 ±28.10 / 712.70 ms │ 640.32 / 671.15 ±41.96 / 753.53 ms │ no change │
│ QQuery 11 │ 2242.31 / 2331.88 ±72.93 / 2443.72 ms │ 2158.98 / 2280.39 ±61.62 / 2327.54 ms │ no change │
│ QQuery 12 │ 189.12 / 221.00 ±57.94 / 336.79 ms │ 187.35 / 223.89 ±64.37 / 352.49 ms │ no change │
│ QQuery 13 │ 540.84 / 559.25 ±13.06 / 577.69 ms │ 545.00 / 551.58 ±5.12 / 557.15 ms │ no change │
│ QQuery 14 │ 2721.65 / 2766.93 ±30.63 / 2809.44 ms │ 2387.31 / 2527.24 ±105.99 / 2689.10 ms │ +1.09x faster │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 25267.25ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24610.23ms │
│ Average Time (HEAD) │ 1684.48ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1640.68ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.2 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1136.1s |
| CPU sys | 58.6s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 12.5 GiB |
| Avg memory | 5.3 GiB |
| CPU user | 1097.7s |
| CPU sys | 59.9s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607862328
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825682)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5607825682-2284-fdg2d 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607863573
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825438)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5607825438-2285-bv8vz 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613974069
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943491)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5613943491-2298-8jtrf 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613943332
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5608109020
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825438)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 828.76 ms │ 848.55 ms │ no change │
│ QQuery 1 │ 194.21 ms │ 189.88 ms │ no change │
│ QQuery 2 │ 460.32 ms │ 512.20 ms │ 1.11x slower │
│ QQuery 3 │ 314.38 ms │ 315.77 ms │ no change │
│ QQuery 4 │ 1160.79 ms │ 1175.01 ms │ no change │
│ QQuery 5 │ 10688.97 ms │ 10289.83 ms │ no change │
│ QQuery 6 │ 2.68 ms │ 2.60 ms │ no change │
│ QQuery 7 │ 707.43 ms │ 705.64 ms │ no change │
│ QQuery 8 │ 421.43 ms │ 424.47 ms │ no change │
│ QQuery 9 │ 3009.65 ms │ 2914.19 ms │ no change │
│ QQuery 10 │ 629.31 ms │ 633.55 ms │ no change │
│ QQuery 11 │ 2224.68 ms │ 2176.31 ms │ no change │
│ QQuery 12 │ 188.92 ms │ 189.43 ms │ no change │
│ QQuery 13 │ 543.91 ms │ 536.92 ms │ no change │
│ QQuery 14 │ 2715.71 ms │ 2338.85 ms │ +1.16x faster │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24091.15ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23253.21ms │
│ Average Time (HEAD) │ 1606.08ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1550.21ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 828.76 / 854.33 ±17.98 / 880.80 ms │ 848.55 / 869.13 ±17.69 / 890.08 ms │ no change │
│ QQuery 1 │ 194.21 / 195.04 ±0.53 / 195.76 ms │ 189.88 / 191.62 ±2.82 / 197.25 ms │ no change │
│ QQuery 2 │ 460.32 / 462.72 ±2.05 / 466.47 ms │ 512.20 / 516.42 ±5.10 / 526.05 ms │ 1.12x slower │
│ QQuery 3 │ 314.38 / 316.69 ±1.63 / 318.41 ms │ 315.77 / 317.02 ±1.03 / 318.36 ms │ no change │
│ QQuery 4 │ 1160.79 / 1195.61 ±34.13 / 1255.84 ms │ 1175.01 / 1186.92 ±9.90 / 1198.98 ms │ no change │
│ QQuery 5 │ 10688.97 / 11473.67 ±491.03 / 12171.82 ms │ 10289.83 / 10645.94 ±414.85 / 11458.54 ms │ +1.08x faster │
│ QQuery 6 │ 2.68 / 2.98 ±0.25 / 3.34 ms │ 2.60 / 2.84 ±0.35 / 3.52 ms │ no change │
│ QQuery 7 │ 707.43 / 752.62 ±38.95 / 819.82 ms │ 705.64 / 736.56 ±31.89 / 794.10 ms │ no change │
│ QQuery 8 │ 421.43 / 438.32 ±23.75 / 484.52 ms │ 424.47 / 438.82 ±16.05 / 467.80 ms │ no change │
│ QQuery 9 │ 3009.65 / 3093.28 ±50.22 / 3139.30 ms │ 2914.19 / 3048.63 ±71.60 / 3122.91 ms │ no change │
│ QQuery 10 │ 629.31 / 659.39 ±34.09 / 725.67 ms │ 633.55 / 676.93 ±50.99 / 774.37 ms │ no change │
│ QQuery 11 │ 2224.68 / 2305.35 ±44.75 / 2355.97 ms │ 2176.31 / 2271.38 ±62.96 / 2343.86 ms │ no change │
│ QQuery 12 │ 188.92 / 232.57 ±83.23 / 399.00 ms │ 189.43 / 219.86 ±55.04 / 329.74 ms │ +1.06x faster │
│ QQuery 13 │ 543.91 / 559.34 ±8.80 / 568.30 ms │ 536.92 / 549.45 ±11.96 / 569.83 ms │ no change │
│ QQuery 14 │ 2715.71 / 2847.71 ±81.43 / 2966.56 ms │ 2338.85 / 2370.06 ±16.30 / 2385.62 ms │ +1.20x faster │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 25389.62ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24041.58ms │
│ Average Time (HEAD) │ 1692.64ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1602.77ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 11 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 12.2 GiB |
| Avg memory | 5.0 GiB |
| CPU user | 1145.3s |
| CPU sys | 58.7s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 10.3 GiB |
| Avg memory | 4.7 GiB |
| CPU user | 1082.0s |
| CPU sys | 55.3s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5766505272
Checking it out
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767894690
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767675986)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 840.53 ms │ 856.00 ms │ no change │
│ QQuery 1 │ 173.45 ms │ 174.06 ms │ no change │
│ QQuery 2 │ 446.30 ms │ 436.89 ms │ no change │
│ QQuery 3 │ 327.39 ms │ 328.13 ms │ no change │
│ QQuery 4 │ 1191.24 ms │ 1191.13 ms │ no change │
│ QQuery 5 │ 11311.75 ms │ 10507.10 ms │ +1.08x faster │
│ QQuery 6 │ 2.68 ms │ 2.99 ms │ 1.12x slower │
│ QQuery 7 │ 799.66 ms │ 834.32 ms │ no change │
│ QQuery 8 │ 454.81 ms │ 458.87 ms │ no change │
│ QQuery 9 │ 2860.68 ms │ 3155.54 ms │ 1.10x slower │
│ QQuery 10 │ 636.57 ms │ 644.98 ms │ no change │
│ QQuery 11 │ 2169.89 ms │ 2108.18 ms │ no change │
│ QQuery 12 │ 192.44 ms │ 197.31 ms │ no change │
│ QQuery 13 │ 543.36 ms │ 546.79 ms │ no change │
│ QQuery 14 │ 583.90 ms │ 600.59 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22534.65ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22042.90ms │
│ Average Time (HEAD) │ 1502.31ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1469.53ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 12 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 840.53 / 876.88 ±28.31 / 911.51 ms │ 856.00 / 887.02 ±28.51 / 930.73 ms │ no change │
│ QQuery 1 │ 173.45 / 175.49 ±2.87 / 181.18 ms │ 174.06 / 174.37 ±0.34 / 175.00 ms │ no change │
│ QQuery 2 │ 446.30 / 450.66 ±3.44 / 456.37 ms │ 436.89 / 441.28 ±2.44 / 444.15 ms │ no change │
│ QQuery 3 │ 327.39 / 330.38 ±2.49 / 334.58 ms │ 328.13 / 330.38 ±1.99 / 333.58 ms │ no change │
│ QQuery 4 │ 1191.24 / 1218.00 ±18.90 / 1248.82 ms │ 1191.13 / 1216.27 ±22.61 / 1256.54 ms │ no change │
│ QQuery 5 │ 11311.75 / 11551.29 ±242.26 / 11868.77 ms │ 10507.10 / 10930.52 ±398.78 / 11647.82 ms │ +1.06x faster │
│ QQuery 6 │ 2.68 / 2.99 ±0.37 / 3.72 ms │ 2.99 / 3.26 ±0.23 / 3.57 ms │ 1.09x slower │
│ QQuery 7 │ 799.66 / 852.10 ±38.26 / 890.16 ms │ 834.32 / 888.47 ±48.73 / 967.91 ms │ no change │
│ QQuery 8 │ 454.81 / 483.32 ±37.23 / 555.21 ms │ 458.87 / 491.83 ±62.34 / 616.44 ms │ no change │
│ QQuery 9 │ 2860.68 / 2976.84 ±65.30 / 3054.61 ms │ 3155.54 / 3257.67 ±66.68 / 3339.94 ms │ 1.09x slower │
│ QQuery 10 │ 636.57 / 671.20 ±31.10 / 718.32 ms │ 644.98 / 665.77 ±26.17 / 716.51 ms │ no change │
│ QQuery 11 │ 2169.89 / 2214.23 ±31.12 / 2259.15 ms │ 2108.18 / 2179.06 ±42.03 / 2224.95 ms │ no change │
│ QQuery 12 │ 192.44 / 221.81 ±52.20 / 326.08 ms │ 197.31 / 238.53 ±78.60 / 395.69 ms │ 1.08x slower │
│ QQuery 13 │ 543.36 / 574.80 ±28.59 / 609.80 ms │ 546.79 / 566.52 ±14.54 / 587.96 ms │ no change │
│ QQuery 14 │ 583.90 / 634.40 ±35.66 / 694.00 ms │ 600.59 / 640.74 ±26.12 / 683.12 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23234.39ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22911.71ms │
│ Average Time (HEAD) │ 1548.96ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1527.45ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 11 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 10.7 GiB |
| Avg memory | 4.9 GiB |
| CPU user | 1076.6s |
| CPU sys | 50.8s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 115.0s |
| Peak memory | 12.2 GiB |
| Avg memory | 5.0 GiB |
| CPU user | 1043.6s |
| CPU sys | 52.4s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767893412
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676454)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 862.31 ms │ 845.14 ms │ no change │
│ QQuery 1 │ 172.78 ms │ 173.93 ms │ no change │
│ QQuery 2 │ 445.46 ms │ 438.46 ms │ no change │
│ QQuery 3 │ 327.40 ms │ 327.45 ms │ no change │
│ QQuery 4 │ 1186.30 ms │ 1158.49 ms │ no change │
│ QQuery 5 │ 11146.15 ms │ 10510.01 ms │ +1.06x faster │
│ QQuery 6 │ 2.92 ms │ 2.91 ms │ no change │
│ QQuery 7 │ 818.11 ms │ 801.38 ms │ no change │
│ QQuery 8 │ 451.84 ms │ 460.43 ms │ no change │
│ QQuery 9 │ 2867.51 ms │ 3200.34 ms │ 1.12x slower │
│ QQuery 10 │ 659.23 ms │ 684.75 ms │ no change │
│ QQuery 11 │ 2102.05 ms │ 2150.18 ms │ no change │
│ QQuery 12 │ 194.53 ms │ 193.67 ms │ no change │
│ QQuery 13 │ 548.24 ms │ 547.06 ms │ no change │
│ QQuery 14 │ 600.61 ms │ 581.15 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22385.47ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22075.36ms │
│ Average Time (HEAD) │ 1492.36ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1471.69ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 862.31 / 896.73 ±21.57 / 930.43 ms │ 845.14 / 872.33 ±29.18 / 919.59 ms │ no change │
│ QQuery 1 │ 172.78 / 173.30 ±0.38 / 173.75 ms │ 173.93 / 174.32 ±0.41 / 175.03 ms │ no change │
│ QQuery 2 │ 445.46 / 450.58 ±4.44 / 458.06 ms │ 438.46 / 442.84 ±3.69 / 447.46 ms │ no change │
│ QQuery 3 │ 327.40 / 328.24 ±0.80 / 329.42 ms │ 327.45 / 327.96 ±0.46 / 328.52 ms │ no change │
│ QQuery 4 │ 1186.30 / 1207.70 ±15.16 / 1229.13 ms │ 1158.49 / 1192.03 ±22.79 / 1226.69 ms │ no change │
│ QQuery 5 │ 11146.15 / 11610.59 ±382.92 / 12237.78 ms │ 10510.01 / 10905.50 ±304.20 / 11342.72 ms │ +1.06x faster │
│ QQuery 6 │ 2.92 / 3.13 ±0.24 / 3.61 ms │ 2.91 / 3.19 ±0.31 / 3.72 ms │ no change │
│ QQuery 7 │ 818.11 / 852.78 ±27.02 / 885.62 ms │ 801.38 / 840.21 ±21.26 / 862.30 ms │ no change │
│ QQuery 8 │ 451.84 / 469.87 ±21.11 / 510.54 ms │ 460.43 / 479.94 ±20.49 / 512.65 ms │ no change │
│ QQuery 9 │ 2867.51 / 2994.93 ±87.25 / 3134.45 ms │ 3200.34 / 3272.16 ±59.39 / 3338.53 ms │ 1.09x slower │
│ QQuery 10 │ 659.23 / 686.26 ±22.34 / 719.06 ms │ 684.75 / 717.90 ±57.40 / 832.52 ms │ no change │
│ QQuery 11 │ 2102.05 / 2154.98 ±66.76 / 2283.31 ms │ 2150.18 / 2231.53 ±42.19 / 2271.37 ms │ no change │
│ QQuery 12 │ 194.53 / 211.83 ±25.69 / 262.93 ms │ 193.67 / 210.36 ±27.40 / 264.86 ms │ no change │
│ QQuery 13 │ 548.24 / 560.21 ±9.57 / 575.03 ms │ 547.06 / 569.71 ±24.75 / 615.94 ms │ no change │
│ QQuery 14 │ 600.61 / 659.31 ±45.02 / 721.05 ms │ 581.15 / 635.91 ±37.69 / 686.04 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23260.43ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22875.87ms │
│ Average Time (HEAD) │ 1550.70ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1525.06ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 10.8 GiB |
| Avg memory | 4.8 GiB |
| CPU user | 1072.4s |
| CPU sys | 52.0s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 115.0s |
| Peak memory | 11.8 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1038.2s |
| CPU sys | 51.0s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767890673
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767675750)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 0576a0b400437ade5a6f3b102465d954a539f263 (0576a0b400437ade5a6f3b102465d954a539f263) to 0576a0b400437ade5a6f3b102465d954a539f263 [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..0576a0b400437ade5a6f3b102465d954a539f263)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
changed:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 0 │ 752.51 ms │ 763.60 ms │ no change │
│ QQuery 1 │ 172.57 ms │ 172.53 ms │ no change │
│ QQuery 2 │ 447.44 ms │ 446.58 ms │ no change │
│ QQuery 3 │ 324.74 ms │ 324.28 ms │ no change │
│ QQuery 4 │ 1130.43 ms │ 1121.47 ms │ no change │
│ QQuery 5 │ 10719.67 ms │ 11032.29 ms │ no change │
│ QQuery 6 │ 2.65 ms │ 2.78 ms │ no change │
│ QQuery 7 │ 744.69 ms │ 732.23 ms │ no change │
│ QQuery 8 │ 436.73 ms │ 430.39 ms │ no change │
│ QQuery 9 │ 2853.69 ms │ 2735.46 ms │ no change │
│ QQuery 10 │ 625.10 ms │ 628.12 ms │ no change │
│ QQuery 11 │ 2133.56 ms │ 2103.45 ms │ no change │
│ QQuery 12 │ 186.45 ms │ 188.10 ms │ no change │
│ QQuery 13 │ 531.20 ms │ 537.32 ms │ no change │
│ QQuery 14 │ 554.93 ms │ 544.82 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 21616.37ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 21763.43ms │
│ Average Time (HEAD) │ 1441.09ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1450.90ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 15 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 752.51 / 766.43 ±7.51 / 774.04 ms │ 763.60 / 771.17 ±6.70 / 781.46 ms │ no change │
│ QQuery 1 │ 172.57 / 173.07 ±0.61 / 174.24 ms │ 172.53 / 172.89 ±0.25 / 173.25 ms │ no change │
│ QQuery 2 │ 447.44 / 449.83 ±2.02 / 452.69 ms │ 446.58 / 449.20 ±2.70 / 453.93 ms │ no change │
│ QQuery 3 │ 324.74 / 325.85 ±1.10 / 327.22 ms │ 324.28 / 326.28 ±1.50 / 328.09 ms │ no change │
│ QQuery 4 │ 1130.43 / 1137.17 ±7.41 / 1150.12 ms │ 1121.47 / 1146.90 ±18.84 / 1165.28 ms │ no change │
│ QQuery 5 │ 10719.67 / 10948.93 ±172.23 / 11239.96 ms │ 11032.29 / 11131.54 ±96.05 / 11296.63 ms │ no change │
│ QQuery 6 │ 2.65 / 2.93 ±0.32 / 3.53 ms │ 2.78 / 3.17 ±0.39 / 3.85 ms │ 1.08x slower │
│ QQuery 7 │ 744.69 / 783.19 ±28.47 / 821.72 ms │ 732.23 / 817.59 ±61.83 / 924.68 ms │ no change │
│ QQuery 8 │ 436.73 / 462.40 ±46.40 / 555.13 ms │ 430.39 / 449.44 ±25.24 / 498.82 ms │ no change │
│ QQuery 9 │ 2853.69 / 2925.68 ±56.13 / 3017.53 ms │ 2735.46 / 2885.50 ±84.22 / 2969.93 ms │ no change │
│ QQuery 10 │ 625.10 / 652.37 ±25.95 / 701.82 ms │ 628.12 / 681.63 ±92.04 / 865.23 ms │ no change │
│ QQuery 11 │ 2133.56 / 2162.62 ±27.03 / 2205.15 ms │ 2103.45 / 2145.46 ±33.90 / 2206.94 ms │ no change │
│ QQuery 12 │ 186.45 / 207.61 ±37.94 / 283.42 ms │ 188.10 / 225.45 ±43.25 / 284.57 ms │ 1.09x slower │
│ QQuery 13 │ 531.20 / 545.02 ±10.45 / 562.92 ms │ 537.32 / 546.24 ±5.32 / 552.87 ms │ no change │
│ QQuery 14 │ 554.93 / 612.73 ±52.24 / 678.82 ms │ 544.82 / 602.40 ±54.05 / 678.94 ms │ no change │
└───────────┴───────────────────────────────────────────┴──────────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22155.82ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22354.86ms │
│ Average Time (HEAD) │ 1477.05ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1490.32ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 115.0s |
| Peak memory | 11.4 GiB |
| Avg memory | 4.9 GiB |
| CPU user | 1040.9s |
| CPU sys | 48.6s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 115.0s |
| Peak memory | 12.2 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1036.5s |
| CPU sys | 46.5s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767906140
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767677168)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767677168-2571-tbn6t 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpcds
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607865008
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607825052)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5607825052-2283-lkpjl 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605748134
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark external_aggr
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Q1(64.0 MB) │ 53.09 ms │ 53.48 ms │ no change │
│ Q1(32.0 MB) │ 51.13 ms │ 51.30 ms │ no change │
│ Q1(16.0 MB) │ 48.00 ms │ 49.64 ms │ no change │
│ Q2(512.0 MB) │ 299.35 ms │ 291.22 ms │ no change │
│ Q2(256.0 MB) │ 270.96 ms │ 265.51 ms │ no change │
│ Q2(128.0 MB) │ 246.36 ms │ 242.21 ms │ no change │
│ Q2(64.0 MB) │ 243.25 ms │ 241.67 ms │ no change │
│ Q2(32.0 MB) │ 300.81 ms │ 300.62 ms │ no change │
└──────────────┴───────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1512.94ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 1495.65ms │
│ Average Time (HEAD) │ 189.12ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 186.96ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 8 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Q1(64.0 MB) │ 53.09 / 57.40 ±3.76 / 63.77 ms │ 53.48 / 58.25 ±3.69 / 64.58 ms │ no change │
│ Q1(32.0 MB) │ 51.13 / 53.16 ±1.57 / 55.95 ms │ 51.30 / 52.85 ±0.98 / 53.85 ms │ no change │
│ Q1(16.0 MB) │ 48.00 / 49.73 ±1.07 / 50.87 ms │ 49.64 / 52.04 ±2.91 / 57.64 ms │ no change │
│ Q2(512.0 MB) │ 299.35 / 304.95 ±6.76 / 318.02 ms │ 291.22 / 297.49 ±5.45 / 304.01 ms │ no change │
│ Q2(256.0 MB) │ 270.96 / 289.38 ±9.29 / 296.09 ms │ 265.51 / 281.30 ±20.32 / 321.49 ms │ no change │
│ Q2(128.0 MB) │ 246.36 / 249.86 ±2.45 / 253.60 ms │ 242.21 / 250.48 ±7.73 / 264.98 ms │ no change │
│ Q2(64.0 MB) │ 243.25 / 246.55 ±4.42 / 255.15 ms │ 241.67 / 243.38 ±1.43 / 245.94 ms │ no change │
│ Q2(32.0 MB) │ 300.81 / 306.89 ±3.82 / 311.43 ms │ 300.62 / 306.03 ±3.62 / 311.71 ms │ no change │
└──────────────┴───────────────────────────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1557.90ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 1541.82ms │
│ Average Time (HEAD) │ 194.74ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 192.73ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 8 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
```
</p>
</details>
<details><summary>Memory Pool Peaks</summary>
Peak `MemoryPool` reservation per query — what DataFusion's accounting believes it reserved. Recorded only when the benchmark runs with `DATAFUSION_RUNTIME_MEMORY_LIMIT` set.
Base: `4048898 (merge-base)` | Changed: `claude/datafusion-issue-25116-75677f`
**`external_aggr`**
| Query | Base | Changed | Change |
| --- | --- | --- | --- |
| 1(64.0 MB) | 35.0 MiB | 35.0 MiB | +0.0% |
| 1(32.0 MB) | 17.8 MiB | 18.6 MiB | +4.9% |
| 1(16.0 MB) | 11.4 MiB | 11.3 MiB | -0.9% |
| 2(512.0 MB) | 139.7 MiB | 135.5 MiB | -3.0% |
| 2(256.0 MB) | 98.5 MiB | 97.6 MiB | -0.9% |
| 2(128.0 MB) | 49.2 MiB | 49.0 MiB | -0.4% |
| 2(64.0 MB) | 30.5 MiB | 30.5 MiB | -0.0% |
| 2(32.0 MB) | 30.0 MiB | 30.0 MiB | +0.0% |
**Pool accounting vs. process RSS**
Max pool peak is the largest reservation any single query in the run reached; peak RSS covers the whole invocation, including data loading and allocator retention, and the two high-water marks need not coincide in time. The gap is therefore an upper bound on what the pool did not account for, not a measurement of it.
| Benchmark | Side | Max pool peak | Peak RSS | Gap | RSS / pool |
| --- | --- | --- | --- | --- | --- |
| external_aggr | base (`4048898 (merge-base)`) | 139.7 MiB | 476.3 MiB | 336.6 MiB | 3.4× |
| external_aggr | changed (`claude/datafusion-issue-25116-75677f`) | 135.5 MiB | 457.6 MiB | 322.1 MiB | 3.4× |
</details>
<details><summary>Resource Usage</summary>
**external_aggr — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 575.1s |
| Peak memory | 476.3 MiB |
| Avg memory | 8.6 MiB |
| CPU user | 25.6s |
| CPU sys | 4.1s |
| Peak spill | 0 B |
**external_aggr — branch**
| Metric | Value |
|--------|-------|
| Wall time | 555.1s |
| Peak memory | 457.6 MiB |
| Avg memory | 8.5 MiB |
| CPU user | 25.8s |
| CPU sys | 3.8s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613972913
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943885)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5613943885-2300-lkqzd 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613973873
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943661)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5613943661-2301-g5bs5 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611506063
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481966)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611481966-2290-98529 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611481966
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5614116107
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943885)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.26 ms │ 1.53 ms │ 1.21x slower │
│ QQuery 1 │ 12.14 ms │ 12.77 ms │ 1.05x slower │
│ QQuery 2 │ 37.24 ms │ 36.96 ms │ no change │
│ QQuery 3 │ 31.54 ms │ 30.49 ms │ no change │
│ QQuery 4 │ 245.52 ms │ 225.96 ms │ +1.09x faster │
│ QQuery 5 │ 291.87 ms │ 281.86 ms │ no change │
│ QQuery 6 │ 1.30 ms │ 1.27 ms │ no change │
│ QQuery 7 │ 13.84 ms │ 13.36 ms │ no change │
│ QQuery 8 │ 331.74 ms │ 328.63 ms │ no change │
│ QQuery 9 │ 461.96 ms │ 461.18 ms │ no change │
│ QQuery 10 │ 74.40 ms │ 72.70 ms │ no change │
│ QQuery 11 │ 86.26 ms │ 83.73 ms │ no change │
│ QQuery 12 │ 287.78 ms │ 276.14 ms │ no change │
│ QQuery 13 │ 372.92 ms │ 375.12 ms │ no change │
│ QQuery 14 │ 284.35 ms │ 286.56 ms │ no change │
│ QQuery 15 │ 270.84 ms │ 293.90 ms │ 1.09x slower │
│ QQuery 16 │ 621.68 ms │ 663.87 ms │ 1.07x slower │
│ QQuery 17 │ 638.66 ms │ 648.08 ms │ no change │
│ QQuery 18 │ 1299.86 ms │ 1335.16 ms │ no change │
│ QQuery 19 │ 27.13 ms │ 27.58 ms │ no change │
│ QQuery 20 │ 517.08 ms │ 523.87 ms │ no change │
│ QQuery 21 │ 522.02 ms │ 517.01 ms │ no change │
│ QQuery 22 │ 989.84 ms │ 997.98 ms │ no change │
│ QQuery 23 │ 3129.49 ms │ 3136.88 ms │ no change │
│ QQuery 24 │ 41.32 ms │ 40.89 ms │ no change │
│ QQuery 25 │ 110.80 ms │ 111.97 ms │ no change │
│ QQuery 26 │ 41.23 ms │ 41.23 ms │ no change │
│ QQuery 27 │ 512.86 ms │ 508.72 ms │ no change │
│ QQuery 28 │ 2952.68 ms │ 2954.78 ms │ no change │
│ QQuery 29 │ 41.50 ms │ 41.56 ms │ no change │
│ QQuery 30 │ 326.86 ms │ 330.86 ms │ no change │
│ QQuery 31 │ 302.61 ms │ 297.08 ms │ no change │
│ QQuery 32 │ 939.57 ms │ 978.94 ms │ no change │
│ QQuery 33 │ 1484.06 ms │ 1542.41 ms │ no change │
│ QQuery 34 │ 1576.47 ms │ 1545.00 ms │ no change │
│ QQuery 35 │ 286.56 ms │ 287.09 ms │ no change │
│ QQuery 36 │ 67.64 ms │ 66.63 ms │ no change │
│ QQuery 37 │ 35.82 ms │ 39.29 ms │ 1.10x slower │
│ QQuery 38 │ 40.61 ms │ 43.05 ms │ 1.06x slower │
│ QQuery 39 │ 141.99 ms │ 143.99 ms │ no change │
│ QQuery 40 │ 14.39 ms │ 14.33 ms │ no change │
│ QQuery 41 │ 14.01 ms │ 14.49 ms │ no change │
│ QQuery 42 │ 14.85 ms │ 13.38 ms │ +1.11x faster │
└───────────┴────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19496.57ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19648.26ms │
│ Average Time (HEAD) │ 453.41ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 456.94ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 6 │
│ Queries with No Change │ 35 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.26 / 4.22 ±5.76 / 15.74 ms │ 1.53 / 4.56 ±5.89 / 16.34 ms │ 1.08x slower │
│ QQuery 1 │ 12.14 / 12.67 ±0.32 / 13.10 ms │ 12.77 / 13.00 ±0.21 / 13.28 ms │ no change │
│ QQuery 2 │ 37.24 / 37.65 ±0.74 / 39.12 ms │ 36.96 / 37.26 ±0.38 / 37.94 ms │ no change │
│ QQuery 3 │ 31.54 / 32.16 ±0.74 / 33.38 ms │ 30.49 / 30.77 ±0.26 / 31.23 ms │ no change │
│ QQuery 4 │ 245.52 / 251.89 ±5.44 / 261.16 ms │ 225.96 / 241.04 ±15.06 / 261.41 ms │ no change │
│ QQuery 5 │ 291.87 / 302.64 ±8.78 / 315.13 ms │ 281.86 / 288.47 ±7.06 / 301.11 ms │ no change │
│ QQuery 6 │ 1.30 / 1.43 ±0.20 / 1.83 ms │ 1.27 / 1.42 ±0.23 / 1.87 ms │ no change │
│ QQuery 7 │ 13.84 / 13.99 ±0.18 / 14.33 ms │ 13.36 / 15.08 ±3.04 / 21.15 ms │ 1.08x slower │
│ QQuery 8 │ 331.74 / 353.76 ±13.29 / 366.96 ms │ 328.63 / 346.95 ±12.28 / 360.86 ms │ no change │
│ QQuery 9 │ 461.96 / 478.17 ±24.17 / 526.14 ms │ 461.18 / 474.08 ±11.80 / 494.80 ms │ no change │
│ QQuery 10 │ 74.40 / 78.91 ±4.88 / 86.09 ms │ 72.70 / 76.94 ±5.12 / 85.75 ms │ no change │
│ QQuery 11 │ 86.26 / 87.43 ±1.39 / 89.99 ms │ 83.73 / 86.68 ±2.20 / 89.70 ms │ no change │
│ QQuery 12 │ 287.78 / 299.88 ±8.83 / 309.56 ms │ 276.14 / 282.91 ±5.87 / 290.19 ms │ +1.06x faster │
│ QQuery 13 │ 372.92 / 387.68 ±10.32 / 403.66 ms │ 375.12 / 387.38 ±11.77 / 407.75 ms │ no change │
│ QQuery 14 │ 284.35 / 294.34 ±8.21 / 308.02 ms │ 286.56 / 298.52 ±8.96 / 308.34 ms │ no change │
│ QQuery 15 │ 270.84 / 298.12 ±14.62 / 312.52 ms │ 293.90 / 310.36 ±18.12 / 340.93 ms │ no change │
│ QQuery 16 │ 621.68 / 657.89 ±21.77 / 681.44 ms │ 663.87 / 691.14 ±26.65 / 730.61 ms │ 1.05x slower │
│ QQuery 17 │ 638.66 / 651.89 ±10.79 / 667.72 ms │ 648.08 / 664.37 ±26.63 / 717.18 ms │ no change │
│ QQuery 18 │ 1299.86 / 1316.56 ±19.00 / 1351.33 ms │ 1335.16 / 1367.25 ±26.34 / 1398.61 ms │ no change │
│ QQuery 19 │ 27.13 / 27.33 ±0.12 / 27.46 ms │ 27.58 / 32.39 ±7.43 / 47.10 ms │ 1.18x slower │
│ QQuery 20 │ 517.08 / 532.57 ±13.27 / 552.83 ms │ 523.87 / 530.41 ±6.07 / 540.02 ms │ no change │
│ QQuery 21 │ 522.02 / 535.80 ±15.47 / 560.53 ms │ 517.01 / 526.69 ±6.56 / 536.66 ms │ no change │
│ QQuery 22 │ 989.84 / 1013.79 ±18.69 / 1043.02 ms │ 997.98 / 1019.25 ±23.12 / 1059.09 ms │ no change │
│ QQuery 23 │ 3129.49 / 3162.56 ±27.65 / 3212.88 ms │ 3136.88 / 3182.55 ±40.98 / 3259.74 ms │ no change │
│ QQuery 24 │ 41.32 / 43.97 ±3.62 / 51.04 ms │ 40.89 / 42.07 ±1.43 / 44.75 ms │ no change │
│ QQuery 25 │ 110.80 / 115.16 ±3.19 / 119.70 ms │ 111.97 / 116.18 ±2.86 / 119.50 ms │ no change │
│ QQuery 26 │ 41.23 / 43.50 ±2.43 / 48.18 ms │ 41.23 / 42.87 ±1.97 / 46.50 ms │ no change │
│ QQuery 27 │ 512.86 / 521.34 ±4.98 / 526.20 ms │ 508.72 / 518.15 ±8.87 / 533.45 ms │ no change │
│ QQuery 28 │ 2952.68 / 2990.84 ±30.86 / 3036.62 ms │ 2954.78 / 2974.57 ±10.70 / 2985.69 ms │ no change │
│ QQuery 29 │ 41.50 / 46.43 ±8.44 / 63.23 ms │ 41.56 / 51.61 ±12.32 / 70.72 ms │ 1.11x slower │
│ QQuery 30 │ 326.86 / 332.72 ±4.43 / 339.94 ms │ 330.86 / 346.65 ±11.27 / 363.85 ms │ no change │
│ QQuery 31 │ 302.61 / 311.61 ±13.84 / 339.11 ms │ 297.08 / 310.18 ±9.07 / 324.34 ms │ no change │
│ QQuery 32 │ 939.57 / 960.28 ±12.09 / 975.18 ms │ 978.94 / 1017.30 ±43.89 / 1101.44 ms │ 1.06x slower │
│ QQuery 33 │ 1484.06 / 1504.39 ±19.37 / 1538.66 ms │ 1542.41 / 1597.19 ±36.57 / 1654.61 ms │ 1.06x slower │
│ QQuery 34 │ 1576.47 / 1618.46 ±37.67 / 1677.94 ms │ 1545.00 / 1582.07 ±36.55 / 1649.45 ms │ no change │
│ QQuery 35 │ 286.56 / 334.16 ±56.50 / 442.68 ms │ 287.09 / 306.01 ±20.60 / 333.32 ms │ +1.09x faster │
│ QQuery 36 │ 67.64 / 71.27 ±3.69 / 78.33 ms │ 66.63 / 79.00 ±12.79 / 103.27 ms │ 1.11x slower │
│ QQuery 37 │ 35.82 / 38.40 ±2.72 / 43.15 ms │ 39.29 / 41.41 ±2.60 / 46.42 ms │ 1.08x slower │
│ QQuery 38 │ 40.61 / 45.96 ±3.49 / 51.32 ms │ 43.05 / 47.92 ±5.54 / 57.80 ms │ no change │
│ QQuery 39 │ 141.99 / 157.41 ±16.15 / 186.90 ms │ 143.99 / 153.16 ±8.00 / 166.41 ms │ no change │
│ QQuery 40 │ 14.39 / 16.30 ±2.01 / 20.09 ms │ 14.33 / 17.78 ±4.19 / 25.34 ms │ 1.09x slower │
│ QQuery 41 │ 14.01 / 14.56 ±0.56 / 15.60 ms │ 14.49 / 14.84 ±0.28 / 15.21 ms │ no change │
│ QQuery 42 │ 14.85 / 15.49 ±0.88 / 17.21 ms │ 13.38 / 15.27 ±3.45 / 22.18 ms │ no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 20015.57ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 20183.73ms │
│ Average Time (HEAD) │ 465.48ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 469.39ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 10 │
│ Queries with No Change │ 31 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 11.8 GiB |
| Avg memory | 4.4 GiB |
| CPU user | 1031.7s |
| CPU sys | 70.2s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 11.7 GiB |
| Avg memory | 4.5 GiB |
| CPU user | 1036.1s |
| CPU sys | 72.3s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767718833
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767676929-2566-d8s6m 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767718235
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767675578)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767675578-2563-bq9d4 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607825682
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5614130693
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943491)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.35 ms │ 1.41 ms │ no change │
│ QQuery 1 │ 12.44 ms │ 12.36 ms │ no change │
│ QQuery 2 │ 37.73 ms │ 37.36 ms │ no change │
│ QQuery 3 │ 32.48 ms │ 32.98 ms │ no change │
│ QQuery 4 │ 269.85 ms │ 268.80 ms │ no change │
│ QQuery 5 │ 315.17 ms │ 304.53 ms │ no change │
│ QQuery 6 │ 1.49 ms │ 1.47 ms │ no change │
│ QQuery 7 │ 14.52 ms │ 14.45 ms │ no change │
│ QQuery 8 │ 395.17 ms │ 388.30 ms │ no change │
│ QQuery 9 │ 535.30 ms │ 522.68 ms │ no change │
│ QQuery 10 │ 76.50 ms │ 75.66 ms │ no change │
│ QQuery 11 │ 88.04 ms │ 88.13 ms │ no change │
│ QQuery 12 │ 302.07 ms │ 318.00 ms │ 1.05x slower │
│ QQuery 13 │ 410.64 ms │ 437.41 ms │ 1.07x slower │
│ QQuery 14 │ 327.29 ms │ 323.32 ms │ no change │
│ QQuery 15 │ 315.10 ms │ 328.88 ms │ no change │
│ QQuery 16 │ 683.36 ms │ 694.00 ms │ no change │
│ QQuery 17 │ 694.21 ms │ 704.43 ms │ no change │
│ QQuery 18 │ 1437.39 ms │ 1458.03 ms │ no change │
│ QQuery 19 │ 29.25 ms │ 29.09 ms │ no change │
│ QQuery 20 │ 541.13 ms │ 536.02 ms │ no change │
│ QQuery 21 │ 533.88 ms │ 535.17 ms │ no change │
│ QQuery 22 │ 1038.85 ms │ 1055.51 ms │ no change │
│ QQuery 23 │ 3374.60 ms │ 3327.61 ms │ no change │
│ QQuery 24 │ 42.93 ms │ 42.30 ms │ no change │
│ QQuery 25 │ 119.06 ms │ 116.56 ms │ no change │
│ QQuery 26 │ 43.41 ms │ 43.60 ms │ no change │
│ QQuery 27 │ 532.07 ms │ 538.81 ms │ no change │
│ QQuery 28 │ 3077.73 ms │ 3063.34 ms │ no change │
│ QQuery 29 │ 43.94 ms │ 43.55 ms │ no change │
│ QQuery 30 │ 341.95 ms │ 343.81 ms │ no change │
│ QQuery 31 │ 307.25 ms │ 304.90 ms │ no change │
│ QQuery 32 │ 1067.36 ms │ 1102.80 ms │ no change │
│ QQuery 33 │ 1715.53 ms │ 1720.57 ms │ no change │
│ QQuery 34 │ 1757.12 ms │ 1750.34 ms │ no change │
│ QQuery 35 │ 374.71 ms │ 350.95 ms │ +1.07x faster │
│ QQuery 36 │ 77.42 ms │ 75.07 ms │ no change │
│ QQuery 37 │ 39.85 ms │ 39.36 ms │ no change │
│ QQuery 38 │ 46.92 ms │ 45.05 ms │ no change │
│ QQuery 39 │ 167.42 ms │ 174.78 ms │ no change │
│ QQuery 40 │ 16.29 ms │ 15.75 ms │ no change │
│ QQuery 41 │ 15.53 ms │ 15.46 ms │ no change │
│ QQuery 42 │ 14.82 ms │ 14.76 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 21269.11ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 21297.38ms │
│ Average Time (HEAD) │ 494.63ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 495.29ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 40 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.35 / 4.37 ±5.79 / 15.96 ms │ 1.41 / 4.42 ±5.81 / 16.04 ms │ no change │
│ QQuery 1 │ 12.44 / 12.94 ±0.32 / 13.30 ms │ 12.36 / 12.81 ±0.29 / 13.16 ms │ no change │
│ QQuery 2 │ 37.73 / 37.96 ±0.28 / 38.49 ms │ 37.36 / 37.73 ±0.27 / 38.14 ms │ no change │
│ QQuery 3 │ 32.48 / 33.32 ±0.73 / 34.64 ms │ 32.98 / 33.17 ±0.35 / 33.87 ms │ no change │
│ QQuery 4 │ 269.85 / 278.70 ±5.13 / 283.47 ms │ 268.80 / 270.99 ±2.50 / 275.36 ms │ no change │
│ QQuery 5 │ 315.17 / 321.84 ±3.76 / 326.61 ms │ 304.53 / 314.15 ±6.76 / 322.54 ms │ no change │
│ QQuery 6 │ 1.49 / 1.62 ±0.19 / 2.00 ms │ 1.47 / 1.63 ±0.21 / 2.04 ms │ no change │
│ QQuery 7 │ 14.52 / 17.13 ±4.96 / 27.06 ms │ 14.45 / 14.60 ±0.16 / 14.83 ms │ +1.17x faster │
│ QQuery 8 │ 395.17 / 401.05 ±4.66 / 408.48 ms │ 388.30 / 394.25 ±6.25 / 406.31 ms │ no change │
│ QQuery 9 │ 535.30 / 540.98 ±4.23 / 545.95 ms │ 522.68 / 536.81 ±8.91 / 547.09 ms │ no change │
│ QQuery 10 │ 76.50 / 77.30 ±0.69 / 78.44 ms │ 75.66 / 76.99 ±0.99 / 78.17 ms │ no change │
│ QQuery 11 │ 88.04 / 92.47 ±5.75 / 103.57 ms │ 88.13 / 92.03 ±5.43 / 102.80 ms │ no change │
│ QQuery 12 │ 302.07 / 319.22 ±14.44 / 344.66 ms │ 318.00 / 323.88 ±5.66 / 334.24 ms │ no change │
│ QQuery 13 │ 410.64 / 442.74 ±21.47 / 476.03 ms │ 437.41 / 450.77 ±10.93 / 467.23 ms │ no change │
│ QQuery 14 │ 327.29 / 331.50 ±4.79 / 340.42 ms │ 323.32 / 337.84 ±11.39 / 350.34 ms │ no change │
│ QQuery 15 │ 315.10 / 333.43 ±14.79 / 359.05 ms │ 328.88 / 335.03 ±6.14 / 346.85 ms │ no change │
│ QQuery 16 │ 683.36 / 708.14 ±19.28 / 733.83 ms │ 694.00 / 715.57 ±15.29 / 733.23 ms │ no change │
│ QQuery 17 │ 694.21 / 710.94 ±16.69 / 741.38 ms │ 704.43 / 743.94 ±29.27 / 793.38 ms │ no change │
│ QQuery 18 │ 1437.39 / 1464.29 ±20.11 / 1490.54 ms │ 1458.03 / 1509.18 ±51.59 / 1593.84 ms │ no change │
│ QQuery 19 │ 29.25 / 33.08 ±7.34 / 47.77 ms │ 29.09 / 32.74 ±5.56 / 43.81 ms │ no change │
│ QQuery 20 │ 541.13 / 547.54 ±6.95 / 558.96 ms │ 536.02 / 553.11 ±19.60 / 591.36 ms │ no change │
│ QQuery 21 │ 533.88 / 540.75 ±6.24 / 550.65 ms │ 535.17 / 540.37 ±4.41 / 547.61 ms │ no change │
│ QQuery 22 │ 1038.85 / 1050.94 ±9.67 / 1064.30 ms │ 1055.51 / 1076.44 ±21.77 / 1118.24 ms │ no change │
│ QQuery 23 │ 3374.60 / 3407.94 ±30.18 / 3459.90 ms │ 3327.61 / 3334.41 ±5.78 / 3344.29 ms │ no change │
│ QQuery 24 │ 42.93 / 43.58 ±0.52 / 44.50 ms │ 42.30 / 47.34 ±8.95 / 65.20 ms │ 1.09x slower │
│ QQuery 25 │ 119.06 / 122.30 ±3.70 / 129.55 ms │ 116.56 / 128.95 ±17.07 / 162.23 ms │ 1.05x slower │
│ QQuery 26 │ 43.41 / 44.48 ±1.32 / 47.07 ms │ 43.60 / 44.92 ±0.73 / 45.79 ms │ no change │
│ QQuery 27 │ 532.07 / 538.76 ±4.08 / 544.44 ms │ 538.81 / 549.24 ±7.58 / 560.39 ms │ no change │
│ QQuery 28 │ 3077.73 / 3106.90 ±21.03 / 3139.71 ms │ 3063.34 / 3090.95 ±15.17 / 3106.19 ms │ no change │
│ QQuery 29 │ 43.94 / 59.85 ±21.84 / 101.02 ms │ 43.55 / 67.69 ±39.27 / 145.27 ms │ 1.13x slower │
│ QQuery 30 │ 341.95 / 357.20 ±20.12 / 395.70 ms │ 343.81 / 351.23 ±6.16 / 362.27 ms │ no change │
│ QQuery 31 │ 307.25 / 331.15 ±29.52 / 388.72 ms │ 304.90 / 319.16 ±13.43 / 336.31 ms │ no change │
│ QQuery 32 │ 1067.36 / 1116.98 ±42.00 / 1193.03 ms │ 1102.80 / 1129.27 ±29.79 / 1185.64 ms │ no change │
│ QQuery 33 │ 1715.53 / 1771.16 ±57.06 / 1864.82 ms │ 1720.57 / 1740.14 ±13.46 / 1755.61 ms │ no change │
│ QQuery 34 │ 1757.12 / 1787.50 ±25.20 / 1820.89 ms │ 1750.34 / 1786.94 ±30.50 / 1826.67 ms │ no change │
│ QQuery 35 │ 374.71 / 420.35 ±62.02 / 541.24 ms │ 350.95 / 400.55 ±69.44 / 536.00 ms │ no change │
│ QQuery 36 │ 77.42 / 87.02 ±7.53 / 98.86 ms │ 75.07 / 83.45 ±6.40 / 90.40 ms │ no change │
│ QQuery 37 │ 39.85 / 40.45 ±0.53 / 41.22 ms │ 39.36 / 44.98 ±9.39 / 63.73 ms │ 1.11x slower │
│ QQuery 38 │ 46.92 / 47.45 ±0.57 / 48.53 ms │ 45.05 / 48.89 ±5.68 / 59.89 ms │ no change │
│ QQuery 39 │ 167.42 / 177.63 ±6.27 / 184.62 ms │ 174.78 / 187.43 ±7.99 / 199.93 ms │ 1.06x slower │
│ QQuery 40 │ 16.29 / 18.37 ±3.67 / 25.70 ms │ 15.75 / 19.05 ±5.36 / 29.71 ms │ no change │
│ QQuery 41 │ 15.53 / 16.16 ±0.54 / 16.95 ms │ 15.46 / 15.72 ±0.22 / 16.12 ms │ no change │
│ QQuery 42 │ 14.82 / 17.43 ±4.41 / 26.21 ms │ 14.76 / 16.00 ±2.32 / 20.64 ms │ +1.09x faster │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 21814.92ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 21814.76ms │
│ Average Time (HEAD) │ 507.32ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 507.32ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 5 │
│ Queries with No Change │ 36 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 110.0s |
| Peak memory | 12.0 GiB |
| Avg memory | 4.9 GiB |
| CPU user | 1107.5s |
| CPU sys | 84.7s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 110.0s |
| Peak memory | 11.5 GiB |
| Avg memory | 4.6 GiB |
| CPU user | 1105.6s |
| CPU sys | 86.8s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605295544
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5605256201-2271-q8ffw 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611629076
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611482086)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611482086-2295-rks7g 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5614109713
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943332)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.22 ms │ 1.23 ms │ no change │
│ QQuery 1 │ 11.95 ms │ 11.72 ms │ no change │
│ QQuery 2 │ 36.46 ms │ 36.32 ms │ no change │
│ QQuery 3 │ 30.54 ms │ 30.57 ms │ no change │
│ QQuery 4 │ 221.66 ms │ 222.60 ms │ no change │
│ QQuery 5 │ 272.18 ms │ 273.69 ms │ no change │
│ QQuery 6 │ 1.26 ms │ 1.28 ms │ no change │
│ QQuery 7 │ 13.03 ms │ 13.02 ms │ no change │
│ QQuery 8 │ 330.80 ms │ 329.41 ms │ no change │
│ QQuery 9 │ 456.94 ms │ 438.79 ms │ no change │
│ QQuery 10 │ 70.71 ms │ 71.39 ms │ no change │
│ QQuery 11 │ 81.40 ms │ 82.35 ms │ no change │
│ QQuery 12 │ 266.89 ms │ 264.72 ms │ no change │
│ QQuery 13 │ 362.30 ms │ 364.47 ms │ no change │
│ QQuery 14 │ 285.68 ms │ 280.66 ms │ no change │
│ QQuery 15 │ 270.37 ms │ 266.66 ms │ no change │
│ QQuery 16 │ 610.88 ms │ 618.49 ms │ no change │
│ QQuery 17 │ 612.98 ms │ 615.45 ms │ no change │
│ QQuery 18 │ 1245.87 ms │ 1264.71 ms │ no change │
│ QQuery 19 │ 27.47 ms │ 27.57 ms │ no change │
│ QQuery 20 │ 515.08 ms │ 521.40 ms │ no change │
│ QQuery 21 │ 514.27 ms │ 521.28 ms │ no change │
│ QQuery 22 │ 981.27 ms │ 990.92 ms │ no change │
│ QQuery 23 │ 3053.71 ms │ 3107.38 ms │ no change │
│ QQuery 24 │ 41.38 ms │ 41.60 ms │ no change │
│ QQuery 25 │ 110.60 ms │ 111.51 ms │ no change │
│ QQuery 26 │ 41.32 ms │ 42.39 ms │ no change │
│ QQuery 27 │ 510.13 ms │ 520.14 ms │ no change │
│ QQuery 28 │ 2892.37 ms │ 2921.54 ms │ no change │
│ QQuery 29 │ 41.39 ms │ 41.38 ms │ no change │
│ QQuery 30 │ 298.04 ms │ 305.42 ms │ no change │
│ QQuery 31 │ 282.54 ms │ 285.67 ms │ no change │
│ QQuery 32 │ 914.16 ms │ 930.78 ms │ no change │
│ QQuery 33 │ 1438.68 ms │ 1461.47 ms │ no change │
│ QQuery 34 │ 1462.77 ms │ 1486.45 ms │ no change │
│ QQuery 35 │ 278.55 ms │ 286.14 ms │ no change │
│ QQuery 36 │ 68.56 ms │ 66.45 ms │ no change │
│ QQuery 37 │ 35.23 ms │ 35.24 ms │ no change │
│ QQuery 38 │ 41.71 ms │ 39.93 ms │ no change │
│ QQuery 39 │ 134.37 ms │ 142.29 ms │ 1.06x slower │
│ QQuery 40 │ 13.98 ms │ 14.31 ms │ no change │
│ QQuery 41 │ 13.70 ms │ 13.53 ms │ no change │
│ QQuery 42 │ 12.94 ms │ 13.54 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 18907.35ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19115.88ms │
│ Average Time (HEAD) │ 439.71ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 444.56ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 1 │
│ Queries with No Change │ 42 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.22 / 3.98 ±5.46 / 14.89 ms │ 1.23 / 3.94 ±5.37 / 14.69 ms │ no change │
│ QQuery 1 │ 11.95 / 12.06 ±0.10 / 12.24 ms │ 11.72 / 12.08 ±0.23 / 12.44 ms │ no change │
│ QQuery 2 │ 36.46 / 36.77 ±0.30 / 37.27 ms │ 36.32 / 36.70 ±0.24 / 37.04 ms │ no change │
│ QQuery 3 │ 30.54 / 31.24 ±0.72 / 32.32 ms │ 30.57 / 30.84 ±0.18 / 31.05 ms │ no change │
│ QQuery 4 │ 221.66 / 224.57 ±3.09 / 230.04 ms │ 222.60 / 225.98 ±2.68 / 230.81 ms │ no change │
│ QQuery 5 │ 272.18 / 276.80 ±3.86 / 281.47 ms │ 273.69 / 275.22 ±1.12 / 276.90 ms │ no change │
│ QQuery 6 │ 1.26 / 1.40 ±0.21 / 1.81 ms │ 1.28 / 1.43 ±0.23 / 1.89 ms │ no change │
│ QQuery 7 │ 13.03 / 13.19 ±0.17 / 13.46 ms │ 13.02 / 13.19 ±0.09 / 13.30 ms │ no change │
│ QQuery 8 │ 330.80 / 332.75 ±1.41 / 334.64 ms │ 329.41 / 331.48 ±1.98 / 334.94 ms │ no change │
│ QQuery 9 │ 456.94 / 462.69 ±4.91 / 469.79 ms │ 438.79 / 452.22 ±8.02 / 463.54 ms │ no change │
│ QQuery 10 │ 70.71 / 72.24 ±1.62 / 75.16 ms │ 71.39 / 74.54 ±5.08 / 84.62 ms │ no change │
│ QQuery 11 │ 81.40 / 82.61 ±0.94 / 84.14 ms │ 82.35 / 83.10 ±0.87 / 84.55 ms │ no change │
│ QQuery 12 │ 266.89 / 271.28 ±3.86 / 277.75 ms │ 264.72 / 273.44 ±5.45 / 279.52 ms │ no change │
│ QQuery 13 │ 362.30 / 376.49 ±10.14 / 392.36 ms │ 364.47 / 378.78 ±19.04 / 416.00 ms │ no change │
│ QQuery 14 │ 285.68 / 292.79 ±5.39 / 300.47 ms │ 280.66 / 286.41 ±4.33 / 294.01 ms │ no change │
│ QQuery 15 │ 270.37 / 279.70 ±9.91 / 298.54 ms │ 266.66 / 275.32 ±8.07 / 290.48 ms │ no change │
│ QQuery 16 │ 610.88 / 619.42 ±5.66 / 625.55 ms │ 618.49 / 624.84 ±5.04 / 630.01 ms │ no change │
│ QQuery 17 │ 612.98 / 621.96 ±8.39 / 634.96 ms │ 615.45 / 631.40 ±14.79 / 652.62 ms │ no change │
│ QQuery 18 │ 1245.87 / 1286.82 ±22.18 / 1305.21 ms │ 1264.71 / 1303.93 ±28.98 / 1353.26 ms │ no change │
│ QQuery 19 │ 27.47 / 34.58 ±13.12 / 60.78 ms │ 27.57 / 35.56 ±15.45 / 66.46 ms │ no change │
│ QQuery 20 │ 515.08 / 529.61 ±19.01 / 566.95 ms │ 521.40 / 529.06 ±10.51 / 549.59 ms │ no change │
│ QQuery 21 │ 514.27 / 520.09 ±6.10 / 531.89 ms │ 521.28 / 527.83 ±6.14 / 539.13 ms │ no change │
│ QQuery 22 │ 981.27 / 993.06 ±10.17 / 1008.33 ms │ 990.92 / 998.79 ±6.43 / 1010.30 ms │ no change │
│ QQuery 23 │ 3053.71 / 3101.44 ±32.73 / 3145.86 ms │ 3107.38 / 3152.13 ±50.01 / 3222.12 ms │ no change │
│ QQuery 24 │ 41.38 / 46.82 ±4.89 / 53.08 ms │ 41.60 / 45.48 ±4.54 / 51.49 ms │ no change │
│ QQuery 25 │ 110.60 / 112.63 ±1.67 / 114.96 ms │ 111.51 / 115.58 ±3.69 / 122.17 ms │ no change │
│ QQuery 26 │ 41.32 / 42.92 ±1.16 / 44.88 ms │ 42.39 / 43.04 ±0.82 / 44.65 ms │ no change │
│ QQuery 27 │ 510.13 / 521.62 ±10.98 / 540.48 ms │ 520.14 / 526.65 ±9.08 / 544.66 ms │ no change │
│ QQuery 28 │ 2892.37 / 2928.44 ±28.82 / 2972.09 ms │ 2921.54 / 2928.05 ±4.88 / 2933.44 ms │ no change │
│ QQuery 29 │ 41.39 / 47.45 ±10.76 / 68.95 ms │ 41.38 / 47.37 ±10.90 / 69.15 ms │ no change │
│ QQuery 30 │ 298.04 / 313.91 ±17.87 / 348.17 ms │ 305.42 / 315.40 ±9.03 / 331.31 ms │ no change │
│ QQuery 31 │ 282.54 / 292.21 ±6.79 / 300.48 ms │ 285.67 / 298.93 ±12.01 / 319.47 ms │ no change │
│ QQuery 32 │ 914.16 / 960.83 ±28.91 / 986.49 ms │ 930.78 / 938.45 ±7.07 / 950.51 ms │ no change │
│ QQuery 33 │ 1438.68 / 1474.17 ±22.55 / 1499.67 ms │ 1461.47 / 1475.37 ±8.86 / 1488.97 ms │ no change │
│ QQuery 34 │ 1462.77 / 1475.26 ±11.98 / 1492.30 ms │ 1486.45 / 1508.63 ±20.06 / 1537.56 ms │ no change │
│ QQuery 35 │ 278.55 / 326.35 ±87.53 / 500.99 ms │ 286.14 / 339.36 ±58.50 / 415.92 ms │ no change │
│ QQuery 36 │ 68.56 / 72.99 ±3.48 / 78.30 ms │ 66.45 / 70.55 ±5.68 / 81.74 ms │ no change │
│ QQuery 37 │ 35.23 / 39.64 ±4.04 / 45.75 ms │ 35.24 / 37.13 ±2.06 / 40.67 ms │ +1.07x faster │
│ QQuery 38 │ 41.71 / 42.78 ±0.97 / 44.51 ms │ 39.93 / 45.73 ±8.10 / 61.63 ms │ 1.07x slower │
│ QQuery 39 │ 134.37 / 158.20 ±19.07 / 188.19 ms │ 142.29 / 150.94 ±5.20 / 157.76 ms │ no change │
│ QQuery 40 │ 13.98 / 15.23 ±2.07 / 19.37 ms │ 14.31 / 14.51 ±0.31 / 15.12 ms │ no change │
│ QQuery 41 │ 13.70 / 13.88 ±0.24 / 14.35 ms │ 13.53 / 13.91 ±0.26 / 14.27 ms │ no change │
│ QQuery 42 │ 12.94 / 13.21 ±0.21 / 13.50 ms │ 13.54 / 22.95 ±10.21 / 40.74 ms │ 1.74x slower │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19376.11ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19496.26ms │
│ Average Time (HEAD) │ 450.61ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 453.40ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 40 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 10.9 GiB |
| Avg memory | 4.4 GiB |
| CPU user | 992.1s |
| CPU sys | 70.0s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 11.1 GiB |
| Avg memory | 4.4 GiB |
| CPU user | 994.3s |
| CPU sys | 72.9s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5800393815
🚀
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb merged PR #25123:
URL: https://github.com/apache/datafusion/pull/25123
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5707373229
@alamb would you have time to review this PR by chance? I'm happy to split out the benchmarks if that helps.
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767719484
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767676929-2567-qfbt8 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark external_aggr
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#discussion_r4066394780
##########
benchmarks/queries/clickbench/README.md:
##########
@@ -301,6 +300,46 @@ LIMIT 10;
+### Q15: Correlation of ad slot dimensions per user
+
+**Question**: "How correlated are the width and height of the screens each user
+browses from?"
+
+**Important Query Properties**: `covar_samp` over a very high cardinality
+`GROUP BY` (about 17.6M distinct `UserID`s). `covar_samp` has no native
Review Comment:
I think the detail about the "no native accumulator" is an implementation detail that is not an important query property. I recommend removing everything after "`covar_samp` has no native accumulator" and then just mention Q16 has the low cardinality version of this
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -235,14 +235,23 @@ impl GroupsAccumulatorAdapter {
assert_eq!(values[0].len(), group_indices.len());
+ // groups_with_rows holds a list of group indexes that have any rows
+ // that need to be accumulated, stored in order of first appearance in
+ // this batch
+ let mut groups_with_rows = vec![];
Review Comment:
You could potentially avoid this allocation on each call to `invoke_per_accumulator` by storing it as a field on self
So something like
```rust
let groups_with_rows = &mut self.scratch;
groups_with_rows.clear();
```
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -380,6 +379,16 @@ impl GroupsAccumulatorAdapter {
})();
self.adjust_allocation(sizes_pre, sizes_post);
+
+ // An error leaves the scratch indexes of the groups after the failed
+ // one in place. The next batch finds the groups it reaches by an empty
+ // `indices`, so clear them. `clear` keeps the capacity, thus the
Review Comment:
I don't think we need to explain how clear works for Vecs here
Also it seems like if one batch returns an error, then the accumulator shouldn't be invoked again (I would expect the query to error) -- so this code is probably dead
However, that being said, I don't see any reason it is bad, but I do think it is unecessary
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -1134,6 +1143,87 @@ mod tests {
Ok(())
}
+ /// Counts the rows it is handed. `MaxAccumulator` cannot see a row that
+ /// reaches it twice, so the tests need an accumulator that can.
+ #[derive(Debug, Default)]
+ struct RowCountAccumulator {
+ rows: i64,
+ }
+
+ impl Accumulator for RowCountAccumulator {
+ fn update_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+ self.rows += values[0].len() as i64;
+ Ok(())
+ }
+ fn merge_batch(&mut self, values: &[ArrayRef]) -> Result<()> {
+ self.update_batch(values)
+ }
+ fn evaluate(&mut self) -> Result<ScalarValue> {
+ Ok(ScalarValue::Int64(Some(self.rows)))
+ }
+ fn state(&mut self) -> Result<Vec<ScalarValue>> {
+ Ok(vec![ScalarValue::Int64(Some(self.rows))])
+ }
+ fn size(&self) -> usize {
+ size_of::<Self>()
+ }
+ }
+
+ fn row_count_adapter() -> GroupsAccumulatorAdapter {
+ GroupsAccumulatorAdapter::new(|| {
+ Ok(Box::new(RowCountAccumulator::default()) as Box<dyn Accumulator>)
+ })
+ }
+
+ fn max_adapter() -> GroupsAccumulatorAdapter {
+ GroupsAccumulatorAdapter::new(|| {
+ Ok(Box::new(MaxAccumulator::try_new(&DataType::Int64)?)
+ as Box<dyn Accumulator>)
+ })
+ }
+
+ /// Every row must reach its own group's accumulator, whatever order the
Review Comment:
do we really need new test coverage? I think this code is pretty well covered by slt tests
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -253,27 +262,17 @@ impl GroupsAccumulatorAdapter {
}
self.add_allocation(indices_allocation_delta);
- // groups_with_rows holds a list of group indexes that have
- // any rows that need to be accumulated, stored in order of
- // group_index
-
- let mut groups_with_rows = vec![];
-
// batch_indices holds indices into values, each group is contiguous
- let mut batch_indices = vec![];
+ let mut batch_indices = Vec::with_capacity(group_indices.len());
Review Comment:
same thing here -- could reuse the allocations across calls
```rust
struct Scratch {
batch_indices: Vec<usize>,
offsets: Vec<usize>,
groups_with_rows: Vec<usize>
}
```
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -1159,6 +1249,100 @@ mod tests {
Ok(())
}
+ /// A failed update must not leave rows behind for the next batch: each
Review Comment:
see above -- I am not sure this is an important property (though it is not incorrect)
##########
datafusion/functions-aggregate/benches/groups_accumulator_adapter.rs:
##########
@@ -0,0 +1,163 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+//! What it costs to run an [`Accumulator`] through [`GroupsAccumulatorAdapter`],
+//! over a sweep of `GROUP BY` cardinalities.
+//!
+//! Two accumulators, because the adapter's cost and the aggregate's cost move
+//! in opposite directions as the group count grows:
+//!
+//! * `routing` wraps an accumulator that only counts the rows it is handed, so
+//! what the benchmark measures is the adapter routing the batch and nothing
+//! else. It is the upper bound on what a change to the routing can do.
+//! * `covar_samp` wraps a real aggregate that has no native `GroupsAccumulator`,
+//! so it shows how much of that upper bound a query actually sees.
+
+use std::hint::black_box;
Review Comment:
I don't think this is necessary as we have end to end coverage, but it isn't incorrect
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201
run benchmark clickbench_extended clickbench_partitioned external_aggr tpch
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605294029
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5605256201-2274-xpstt 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpch
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5608117770
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5607779662)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 40488988ad596c9b093ad60e1453430d803ce33c (40488988ad596c9b093ad60e1453430d803ce33c) to 40488988ad596c9b093ad60e1453430d803ce33c [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..40488988ad596c9b093ad60e1453430d803ce33c)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
changed:
ref: "40488988ad596c9b093ad60e1453430d803ce33c"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 776.91 ms │ 769.18 ms │ no change │
│ QQuery 1 │ 194.18 ms │ 194.26 ms │ no change │
│ QQuery 2 │ 460.69 ms │ 458.55 ms │ no change │
│ QQuery 3 │ 313.45 ms │ 315.06 ms │ no change │
│ QQuery 4 │ 1149.26 ms │ 1166.28 ms │ no change │
│ QQuery 5 │ 10570.54 ms │ 11107.29 ms │ 1.05x slower │
│ QQuery 6 │ 2.59 ms │ 2.71 ms │ no change │
│ QQuery 7 │ 670.39 ms │ 680.42 ms │ no change │
│ QQuery 8 │ 407.79 ms │ 409.74 ms │ no change │
│ QQuery 9 │ 2891.29 ms │ 2921.26 ms │ no change │
│ QQuery 10 │ 633.84 ms │ 626.75 ms │ no change │
│ QQuery 11 │ 2252.96 ms │ 2275.01 ms │ no change │
│ QQuery 12 │ 186.45 ms │ 186.61 ms │ no change │
│ QQuery 13 │ 537.83 ms │ 549.49 ms │ no change │
│ QQuery 14 │ 2541.42 ms │ 2677.53 ms │ 1.05x slower │
└───────────┴─────────────┴──────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 23589.58ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24340.13ms │
│ Average Time (HEAD) │ 1572.64ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1622.68ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 776.91 / 792.98 ±10.49 / 807.93 ms │ 769.18 / 794.49 ±16.95 / 819.53 ms │ no change │
│ QQuery 1 │ 194.18 / 196.70 ±3.04 / 202.62 ms │ 194.26 / 194.71 ±0.39 / 195.29 ms │ no change │
│ QQuery 2 │ 460.69 / 462.69 ±1.67 / 464.88 ms │ 458.55 / 461.28 ±1.47 / 462.99 ms │ no change │
│ QQuery 3 │ 313.45 / 315.31 ±2.09 / 319.21 ms │ 315.06 / 316.86 ±1.77 / 319.98 ms │ no change │
│ QQuery 4 │ 1149.26 / 1161.07 ±9.91 / 1174.46 ms │ 1166.28 / 1168.00 ±2.26 / 1172.35 ms │ no change │
│ QQuery 5 │ 10570.54 / 11335.98 ±484.72 / 12028.31 ms │ 11107.29 / 11610.43 ±295.36 / 11932.32 ms │ no change │
│ QQuery 6 │ 2.59 / 2.88 ±0.34 / 3.54 ms │ 2.71 / 9.63 ±13.21 / 36.04 ms │ 3.35x slower │
│ QQuery 7 │ 670.39 / 721.09 ±62.97 / 843.35 ms │ 680.42 / 732.70 ±56.79 / 838.39 ms │ no change │
│ QQuery 8 │ 407.79 / 411.99 ±3.84 / 418.15 ms │ 409.74 / 424.56 ±18.80 / 461.04 ms │ no change │
│ QQuery 9 │ 2891.29 / 2956.11 ±50.43 / 3014.63 ms │ 2921.26 / 3038.60 ±79.75 / 3138.05 ms │ no change │
│ QQuery 10 │ 633.84 / 646.90 ±11.55 / 666.21 ms │ 626.75 / 658.14 ±30.09 / 696.91 ms │ no change │
│ QQuery 11 │ 2252.96 / 2268.78 ±16.95 / 2301.31 ms │ 2275.01 / 2289.40 ±14.87 / 2307.97 ms │ no change │
│ QQuery 12 │ 186.45 / 204.79 ±21.08 / 240.24 ms │ 186.61 / 223.37 ±68.90 / 361.10 ms │ 1.09x slower │
│ QQuery 13 │ 537.83 / 544.57 ±5.28 / 551.87 ms │ 549.49 / 559.81 ±6.84 / 570.40 ms │ no change │
│ QQuery 14 │ 2541.42 / 2663.05 ±77.20 / 2751.87 ms │ 2677.53 / 2739.62 ±35.03 / 2786.53 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24684.89ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 25221.60ms │
│ Average Time (HEAD) │ 1645.66ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1681.44ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 11.9 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1120.7s |
| CPU sys | 54.2s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.7 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1127.6s |
| CPU sys | 57.2s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611619061
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpch
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1 │ 39.08 ms │ 39.14 ms │ no change │
│ QQuery 2 │ 19.31 ms │ 19.22 ms │ no change │
│ QQuery 3 │ 30.06 ms │ 29.50 ms │ no change │
│ QQuery 4 │ 18.29 ms │ 18.38 ms │ no change │
│ QQuery 5 │ 38.18 ms │ 37.42 ms │ no change │
│ QQuery 6 │ 16.70 ms │ 16.53 ms │ no change │
│ QQuery 7 │ 43.47 ms │ 43.05 ms │ no change │
│ QQuery 8 │ 42.63 ms │ 42.61 ms │ no change │
│ QQuery 9 │ 52.58 ms │ 52.65 ms │ no change │
│ QQuery 10 │ 44.03 ms │ 44.40 ms │ no change │
│ QQuery 11 │ 14.46 ms │ 14.26 ms │ no change │
│ QQuery 12 │ 25.10 ms │ 24.67 ms │ no change │
│ QQuery 13 │ 43.47 ms │ 43.40 ms │ no change │
│ QQuery 14 │ 25.58 ms │ 25.73 ms │ no change │
│ QQuery 15 │ 32.17 ms │ 32.13 ms │ no change │
│ QQuery 16 │ 14.49 ms │ 14.32 ms │ no change │
│ QQuery 17 │ 78.08 ms │ 78.69 ms │ no change │
│ QQuery 18 │ 63.48 ms │ 64.12 ms │ no change │
│ QQuery 19 │ 34.24 ms │ 34.47 ms │ no change │
│ QQuery 20 │ 32.45 ms │ 33.46 ms │ no change │
│ QQuery 21 │ 59.64 ms │ 60.46 ms │ no change │
│ QQuery 22 │ 15.03 ms │ 15.26 ms │ no change │
└───────────┴──────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 782.51ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 783.87ms │
│ Average Time (HEAD) │ 35.57ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 35.63ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 22 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴──────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1 │ 39.08 / 39.99 ±1.23 / 42.43 ms │ 39.14 / 39.90 ±1.06 / 41.95 ms │ no change │
│ QQuery 2 │ 19.31 / 19.65 ±0.29 / 20.06 ms │ 19.22 / 20.14 ±1.01 / 21.64 ms │ no change │
│ QQuery 3 │ 30.06 / 31.56 ±1.20 / 33.14 ms │ 29.50 / 30.48 ±1.41 / 33.29 ms │ no change │
│ QQuery 4 │ 18.29 / 19.23 ±0.77 / 20.24 ms │ 18.38 / 18.49 ±0.10 / 18.61 ms │ no change │
│ QQuery 5 │ 38.18 / 38.37 ±0.13 / 38.55 ms │ 37.42 / 38.26 ±1.16 / 40.54 ms │ no change │
│ QQuery 6 │ 16.70 / 17.04 ±0.51 / 18.04 ms │ 16.53 / 16.74 ±0.23 / 17.15 ms │ no change │
│ QQuery 7 │ 43.47 / 45.44 ±1.81 / 47.76 ms │ 43.05 / 45.51 ±3.88 / 53.20 ms │ no change │
│ QQuery 8 │ 42.63 / 44.23 ±1.81 / 47.69 ms │ 42.61 / 43.41 ±0.98 / 45.35 ms │ no change │
│ QQuery 9 │ 52.58 / 53.40 ±0.79 / 54.60 ms │ 52.65 / 53.37 ±0.41 / 53.86 ms │ no change │
│ QQuery 10 │ 44.03 / 45.06 ±0.94 / 46.67 ms │ 44.40 / 46.59 ±2.13 / 50.15 ms │ no change │
│ QQuery 11 │ 14.46 / 14.63 ±0.19 / 15.01 ms │ 14.26 / 14.41 ±0.19 / 14.78 ms │ no change │
│ QQuery 12 │ 25.10 / 25.42 ±0.27 / 25.84 ms │ 24.67 / 25.18 ±0.41 / 25.72 ms │ no change │
│ QQuery 13 │ 43.47 / 44.80 ±1.53 / 47.72 ms │ 43.40 / 45.00 ±2.52 / 50.01 ms │ no change │
│ QQuery 14 │ 25.58 / 25.70 ±0.12 / 25.91 ms │ 25.73 / 25.92 ±0.17 / 26.13 ms │ no change │
│ QQuery 15 │ 32.17 / 32.62 ±0.53 / 33.62 ms │ 32.13 / 32.48 ±0.24 / 32.82 ms │ no change │
│ QQuery 16 │ 14.49 / 14.70 ±0.17 / 14.92 ms │ 14.32 / 14.50 ±0.16 / 14.72 ms │ no change │
│ QQuery 17 │ 78.08 / 81.08 ±2.58 / 85.00 ms │ 78.69 / 79.50 ±0.87 / 80.60 ms │ no change │
│ QQuery 18 │ 63.48 / 65.37 ±1.28 / 66.63 ms │ 64.12 / 66.42 ±2.19 / 70.09 ms │ no change │
│ QQuery 19 │ 34.24 / 35.43 ±0.98 / 36.83 ms │ 34.47 / 35.00 ±0.54 / 36.02 ms │ no change │
│ QQuery 20 │ 32.45 / 33.34 ±0.72 / 34.49 ms │ 33.46 / 33.86 ±0.55 / 34.96 ms │ no change │
│ QQuery 21 │ 59.64 / 60.45 ±0.59 / 61.18 ms │ 60.46 / 60.87 ±0.42 / 61.41 ms │ no change │
│ QQuery 22 │ 15.03 / 16.04 ±1.78 / 19.59 ms │ 15.26 / 16.58 ±2.34 / 21.25 ms │ no change │
└───────────┴────────────────────────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 803.53ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 802.62ms │
│ Average Time (HEAD) │ 36.52ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 36.48ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 22 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴──────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 1.0 GiB |
| Avg memory | 625.7 MiB |
| CPU user | 22.7s |
| CPU sys | 1.9s |
| Peak spill | 0 B |
**tpch — branch**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 1.2 GiB |
| Avg memory | 650.3 MiB |
| CPU user | 22.7s |
| CPU sys | 1.9s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613943885
run benchmark clickbench_partitioned
baseline:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
changed:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767892014
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676204)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing 0576a0b400437ade5a6f3b102465d954a539f263 (0576a0b400437ade5a6f3b102465d954a539f263) to 0576a0b400437ade5a6f3b102465d954a539f263 [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..0576a0b400437ade5a6f3b102465d954a539f263)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
changed:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 0 │ 839.23 ms │ 825.40 ms │ no change │
│ QQuery 1 │ 173.44 ms │ 172.76 ms │ no change │
│ QQuery 2 │ 445.72 ms │ 449.20 ms │ no change │
│ QQuery 3 │ 327.79 ms │ 327.43 ms │ no change │
│ QQuery 4 │ 1166.60 ms │ 1149.46 ms │ no change │
│ QQuery 5 │ 10916.98 ms │ 11028.78 ms │ no change │
│ QQuery 6 │ 2.86 ms │ 2.72 ms │ no change │
│ QQuery 7 │ 771.33 ms │ 781.14 ms │ no change │
│ QQuery 8 │ 449.84 ms │ 445.93 ms │ no change │
│ QQuery 9 │ 2912.26 ms │ 2894.25 ms │ no change │
│ QQuery 10 │ 636.16 ms │ 633.68 ms │ no change │
│ QQuery 11 │ 2146.23 ms │ 2192.22 ms │ no change │
│ QQuery 12 │ 191.84 ms │ 191.18 ms │ no change │
│ QQuery 13 │ 531.15 ms │ 546.81 ms │ no change │
│ QQuery 14 │ 580.16 ms │ 575.50 ms │ no change │
└───────────┴─────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22091.59ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22216.43ms │
│ Average Time (HEAD) │ 1472.77ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1481.10ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 15 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 839.23 / 855.72 ±10.38 / 869.53 ms │ 825.40 / 842.00 ±9.18 / 853.27 ms │ no change │
│ QQuery 1 │ 173.44 / 175.50 ±3.80 / 183.10 ms │ 172.76 / 177.34 ±8.06 / 193.42 ms │ no change │
│ QQuery 2 │ 445.72 / 449.36 ±2.72 / 454.07 ms │ 449.20 / 451.41 ±2.29 / 454.85 ms │ no change │
│ QQuery 3 │ 327.79 / 329.32 ±1.06 / 331.04 ms │ 327.43 / 328.55 ±1.51 / 331.48 ms │ no change │
│ QQuery 4 │ 1166.60 / 1172.92 ±4.85 / 1179.61 ms │ 1149.46 / 1165.46 ±11.39 / 1184.28 ms │ no change │
│ QQuery 5 │ 10916.98 / 11416.08 ±295.43 / 11671.79 ms │ 11028.78 / 11196.25 ±134.63 / 11401.46 ms │ no change │
│ QQuery 6 │ 2.86 / 3.04 ±0.26 / 3.55 ms │ 2.72 / 2.96 ±0.35 / 3.65 ms │ no change │
│ QQuery 7 │ 771.33 / 813.33 ±29.14 / 861.72 ms │ 781.14 / 818.09 ±31.18 / 874.27 ms │ no change │
│ QQuery 8 │ 449.84 / 460.19 ±12.28 / 475.56 ms │ 445.93 / 462.58 ±26.70 / 515.59 ms │ no change │
│ QQuery 9 │ 2912.26 / 3031.15 ±126.24 / 3187.06 ms │ 2894.25 / 2999.86 ±53.14 / 3037.01 ms │ no change │
│ QQuery 10 │ 636.16 / 643.69 ±7.00 / 654.92 ms │ 633.68 / 649.68 ±14.66 / 671.55 ms │ no change │
│ QQuery 11 │ 2146.23 / 2220.31 ±46.83 / 2274.33 ms │ 2192.22 / 2202.90 ±9.27 / 2218.85 ms │ no change │
│ QQuery 12 │ 191.84 / 227.08 ±65.57 / 358.17 ms │ 191.18 / 209.64 ±25.73 / 258.81 ms │ +1.08x faster │
│ QQuery 13 │ 531.15 / 549.80 ±13.75 / 568.09 ms │ 546.81 / 558.23 ±8.83 / 572.01 ms │ no change │
│ QQuery 14 │ 580.16 / 615.69 ±47.12 / 705.17 ms │ 575.50 / 626.73 ±50.86 / 716.18 ms │ no change │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 22963.17ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 22691.69ms │
│ Average Time (HEAD) │ 1530.88ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1512.78ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 14 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 120.0s |
| Peak memory | 11.8 GiB |
| Avg memory | 4.9 GiB |
| CPU user | 1065.6s |
| CPU sys | 50.6s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 115.0s |
| Peak memory | 11.9 GiB |
| Avg memory | 5.2 GiB |
| CPU user | 1051.3s |
| CPU sys | 50.2s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767718456
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767676929-2568-hs4ng 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpch
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929
run benchmark clickbench_partitioned external_aggr tpch
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767677168
run benchmark tpcds
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#discussion_r4071529092
##########
datafusion/functions-aggregate-common/src/aggregate/groups_accumulator.rs:
##########
@@ -253,27 +262,17 @@ impl GroupsAccumulatorAdapter {
}
self.add_allocation(indices_allocation_delta);
- // groups_with_rows holds a list of group indexes that have
- // any rows that need to be accumulated, stored in order of
- // group_index
-
- let mut groups_with_rows = vec![];
-
// batch_indices holds indices into values, each group is contiguous
- let mut batch_indices = vec![];
+ let mut batch_indices = Vec::with_capacity(group_indices.len());
Review Comment:
we can file a follow on issue for this idea perahsp
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5776381020
> yep, thank you for the great feedback. should i go ahead and merge it?
it is fine with me -- we could wait until tomorrow too to give anyone else who has ideas time to weigh in
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767791554
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676929)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpch
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 43.09 ms │ 43.86 ms │ no change │
│ QQuery 2 │ 20.05 ms │ 20.51 ms │ no change │
│ QQuery 3 │ 30.35 ms │ 31.34 ms │ no change │
│ QQuery 4 │ 18.62 ms │ 19.22 ms │ no change │
│ QQuery 5 │ 38.78 ms │ 40.05 ms │ no change │
│ QQuery 6 │ 17.75 ms │ 17.88 ms │ no change │
│ QQuery 7 │ 46.18 ms │ 47.43 ms │ no change │
│ QQuery 8 │ 44.52 ms │ 44.15 ms │ no change │
│ QQuery 9 │ 55.41 ms │ 53.92 ms │ no change │
│ QQuery 10 │ 46.56 ms │ 44.71 ms │ no change │
│ QQuery 11 │ 15.66 ms │ 15.15 ms │ no change │
│ QQuery 12 │ 24.22 ms │ 22.53 ms │ +1.07x faster │
│ QQuery 13 │ 48.65 ms │ 46.38 ms │ no change │
│ QQuery 14 │ 27.04 ms │ 26.50 ms │ no change │
│ QQuery 15 │ 34.12 ms │ 33.56 ms │ no change │
│ QQuery 16 │ 15.56 ms │ 15.08 ms │ no change │
│ QQuery 17 │ 86.25 ms │ 86.77 ms │ no change │
│ QQuery 18 │ 68.39 ms │ 69.00 ms │ no change │
│ QQuery 19 │ 35.03 ms │ 34.86 ms │ no change │
│ QQuery 20 │ 35.83 ms │ 35.38 ms │ no change │
│ QQuery 21 │ 64.65 ms │ 63.22 ms │ no change │
│ QQuery 22 │ 15.72 ms │ 15.13 ms │ no change │
└───────────┴──────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 832.41ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 826.63ms │
│ Average Time (HEAD) │ 37.84ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 37.57ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 21 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴──────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 43.09 / 44.02 ±0.98 / 45.60 ms │ 43.86 / 45.18 ±1.30 / 46.78 ms │ no change │
│ QQuery 2 │ 20.05 / 20.22 ±0.16 / 20.50 ms │ 20.51 / 21.06 ±0.85 / 22.75 ms │ no change │
│ QQuery 3 │ 30.35 / 30.63 ±0.17 / 30.81 ms │ 31.34 / 32.13 ±1.14 / 34.39 ms │ no change │
│ QQuery 4 │ 18.62 / 19.38 ±0.91 / 20.61 ms │ 19.22 / 20.00 ±1.24 / 22.47 ms │ no change │
│ QQuery 5 │ 38.78 / 39.67 ±1.14 / 41.92 ms │ 40.05 / 40.39 ±0.38 / 41.12 ms │ no change │
│ QQuery 6 │ 17.75 / 18.83 ±1.17 / 20.27 ms │ 17.88 / 18.01 ±0.15 / 18.22 ms │ no change │
│ QQuery 7 │ 46.18 / 47.82 ±1.01 / 49.25 ms │ 47.43 / 49.08 ±1.60 / 51.90 ms │ no change │
│ QQuery 8 │ 44.52 / 45.44 ±1.27 / 47.95 ms │ 44.15 / 44.42 ±0.47 / 45.37 ms │ no change │
│ QQuery 9 │ 55.41 / 56.68 ±1.00 / 57.85 ms │ 53.92 / 55.51 ±1.14 / 57.42 ms │ no change │
│ QQuery 10 │ 46.56 / 48.11 ±1.22 / 49.73 ms │ 44.71 / 45.11 ±0.40 / 45.79 ms │ +1.07x faster │
│ QQuery 11 │ 15.66 / 17.28 ±2.60 / 22.47 ms │ 15.15 / 15.41 ±0.25 / 15.85 ms │ +1.12x faster │
│ QQuery 12 │ 24.22 / 27.00 ±3.78 / 34.34 ms │ 22.53 / 23.20 ±0.66 / 24.40 ms │ +1.16x faster │
│ QQuery 13 │ 48.65 / 49.10 ±0.46 / 49.91 ms │ 46.38 / 46.69 ±0.40 / 47.47 ms │ no change │
│ QQuery 14 │ 27.04 / 27.97 ±1.19 / 30.32 ms │ 26.50 / 26.59 ±0.12 / 26.82 ms │ no change │
│ QQuery 15 │ 34.12 / 35.00 ±0.95 / 36.77 ms │ 33.56 / 34.18 ±0.51 / 34.80 ms │ no change │
│ QQuery 16 │ 15.56 / 15.69 ±0.09 / 15.81 ms │ 15.08 / 15.38 ±0.23 / 15.75 ms │ no change │
│ QQuery 17 │ 86.25 / 87.81 ±0.87 / 88.70 ms │ 86.77 / 88.64 ±1.25 / 90.10 ms │ no change │
│ QQuery 18 │ 68.39 / 70.33 ±1.13 / 71.87 ms │ 69.00 / 70.82 ±1.66 / 73.82 ms │ no change │
│ QQuery 19 │ 35.03 / 35.85 ±0.92 / 37.19 ms │ 34.86 / 35.94 ±0.88 / 37.37 ms │ no change │
│ QQuery 20 │ 35.83 / 36.27 ±0.28 / 36.62 ms │ 35.38 / 35.93 ±0.31 / 36.34 ms │ no change │
│ QQuery 21 │ 64.65 / 66.74 ±1.58 / 69.56 ms │ 63.22 / 65.88 ±2.03 / 68.67 ms │ no change │
│ QQuery 22 │ 15.72 / 16.01 ±0.16 / 16.18 ms │ 15.13 / 15.46 ±0.22 / 15.80 ms │ no change │
└───────────┴────────────────────────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 855.84ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 844.99ms │
│ Average Time (HEAD) │ 38.90ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 38.41ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 19 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴──────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 1.0 GiB |
| Avg memory | 604.2 MiB |
| CPU user | 23.0s |
| CPU sys | 2.0s |
| Peak spill | 0 B |
**tpch — branch**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 1.0 GiB |
| Avg memory | 623.5 MiB |
| CPU user | 23.2s |
| CPU sys | 2.0s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "codecov-commenter (via GitHub)" <gi...@apache.org>.
codecov-commenter commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5609755131
## [Codecov](https://app.codecov.io/gh/apache/datafusion/pull/25123?dropdown=coverage&src=pr&el=h1&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache) Report
:x: Patch coverage is `93.67089%` with `5 lines` in your changes missing coverage. Please review.
:white_check_mark: Project coverage is 81.80%. Comparing base ([`4048898`](https://app.codecov.io/gh/apache/datafusion/commit/40488988ad596c9b093ad60e1453430d803ce33c?dropdown=coverage&el=desc&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache)) to head ([`3238cd6`](https://app.codecov.io/gh/apache/datafusion/commit/3238cd640c6028263dfe8689deae10080e7be8b9?dropdown=coverage&el=desc&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache)).
| [Files with missing lines](https://app.codecov.io/gh/apache/datafusion/pull/25123?dropdown=coverage&src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache) | Patch % | Lines |
|---|---|---|
| [...gregate-common/src/aggregate/groups\_accumulator.rs](https://app.codecov.io/gh/apache/datafusion/pull/25123?src=pr&el=tree&filepath=datafusion%2Ffunctions-aggregate-common%2Fsrc%2Faggregate%2Fgroups_accumulator.rs&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache#diff-ZGF0YWZ1c2lvbi9mdW5jdGlvbnMtYWdncmVnYXRlLWNvbW1vbi9zcmMvYWdncmVnYXRlL2dyb3Vwc19hY2N1bXVsYXRvci5ycw==) | 93.67% | [1 Missing and 4 partials :warning: ](https://app.codecov.io/gh/apache/datafusion/pull/25123?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache) |
<details><summary>Additional details and impacted files</summary>
```diff
@@ Coverage Diff @@
## main #25123 +/- ##
=======================================
Coverage 81.80% 81.80%
=======================================
Files 1130 1130
Lines 417754 417816 +62
Branches 417754 417816 +62
=======================================
+ Hits 341754 341806 +52
- Misses 55875 55879 +4
- Partials 20125 20131 +6
```
</details>
[:umbrella: View full report in Codecov by Harness](https://app.codecov.io/gh/apache/datafusion/pull/25123?dropdown=coverage&src=pr&el=continue&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache).
:loudspeaker: Have feedback on the report? [Share it here](https://about.codecov.io/codecov-pr-comment-feedback/?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=apache).
<details><summary> :rocket: New features to boost your workflow: </summary>
- :snowflake: [Test Analytics](https://docs.codecov.com/docs/test-analytics): Detect flaky tests, report on failures, and find test suite problems.
- :package: [JS Bundle Analysis](https://docs.codecov.com/docs/javascript-bundle-analysis): Save yourself from yourself by tracking and limiting bundle sizes in JS merges.
</details>
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611505342
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611481841-2294-wrl9m 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpch
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611499243
run benchmark tpcds
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611747953
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark external_aggr
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Q1(64.0 MB) │ 53.26 ms │ 51.06 ms │ no change │
│ Q1(32.0 MB) │ 49.58 ms │ 49.48 ms │ no change │
│ Q1(16.0 MB) │ 46.83 ms │ 47.42 ms │ no change │
│ Q2(512.0 MB) │ 274.24 ms │ 280.56 ms │ no change │
│ Q2(256.0 MB) │ 265.19 ms │ 270.83 ms │ no change │
│ Q2(128.0 MB) │ 240.07 ms │ 240.52 ms │ no change │
│ Q2(64.0 MB) │ 235.07 ms │ 238.59 ms │ no change │
│ Q2(32.0 MB) │ 297.03 ms │ 294.92 ms │ no change │
└──────────────┴───────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1461.26ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 1473.37ms │
│ Average Time (HEAD) │ 182.66ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 184.17ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 8 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Q1(64.0 MB) │ 53.26 / 56.11 ±3.50 / 62.96 ms │ 51.06 / 54.59 ±4.14 / 62.31 ms │ no change │
│ Q1(32.0 MB) │ 49.58 / 51.76 ±1.65 / 54.51 ms │ 49.48 / 51.41 ±1.41 / 53.23 ms │ no change │
│ Q1(16.0 MB) │ 46.83 / 49.53 ±1.49 / 51.28 ms │ 47.42 / 49.33 ±2.08 / 53.19 ms │ no change │
│ Q2(512.0 MB) │ 274.24 / 284.81 ±9.16 / 300.93 ms │ 280.56 / 290.35 ±9.04 / 301.56 ms │ no change │
│ Q2(256.0 MB) │ 265.19 / 286.75 ±22.11 / 329.24 ms │ 270.83 / 294.69 ±16.78 / 323.11 ms │ no change │
│ Q2(128.0 MB) │ 240.07 / 249.05 ±9.45 / 260.92 ms │ 240.52 / 257.76 ±22.69 / 301.66 ms │ no change │
│ Q2(64.0 MB) │ 235.07 / 237.89 ±1.60 / 239.67 ms │ 238.59 / 241.87 ±3.39 / 247.43 ms │ no change │
│ Q2(32.0 MB) │ 297.03 / 299.87 ±2.46 / 303.58 ms │ 294.92 / 303.14 ±4.56 / 307.52 ms │ no change │
└──────────────┴────────────────────────────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1515.78ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 1543.15ms │
│ Average Time (HEAD) │ 189.47ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 192.89ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 8 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
```
</p>
</details>
<details><summary>Memory Pool Peaks</summary>
Peak `MemoryPool` reservation per query — what DataFusion's accounting believes it reserved. Recorded only when the benchmark runs with `DATAFUSION_RUNTIME_MEMORY_LIMIT` set.
Base: `4048898 (merge-base)` | Changed: `claude/datafusion-issue-25116-75677f`
**`external_aggr`**
| Query | Base | Changed | Change |
| --- | --- | --- | --- |
| 1(64.0 MB) | 36.8 MiB | 38.5 MiB | +4.8% |
| 1(32.0 MB) | 17.8 MiB | 17.3 MiB | -2.4% |
| 1(16.0 MB) | 11.4 MiB | 11.4 MiB | -0.1% |
| 2(512.0 MB) | 136.9 MiB | 136.6 MiB | -0.2% |
| 2(256.0 MB) | 98.3 MiB | 98.2 MiB | -0.1% |
| 2(128.0 MB) | 50.1 MiB | 49.5 MiB | -1.2% |
| 2(64.0 MB) | 29.2 MiB | 29.2 MiB | +0.0% |
| 2(32.0 MB) | 30.0 MiB | 30.0 MiB | +0.0% |
**Pool accounting vs. process RSS**
Max pool peak is the largest reservation any single query in the run reached; peak RSS covers the whole invocation, including data loading and allocator retention, and the two high-water marks need not coincide in time. The gap is therefore an upper bound on what the pool did not account for, not a measurement of it.
| Benchmark | Side | Max pool peak | Peak RSS | Gap | RSS / pool |
| --- | --- | --- | --- | --- | --- |
| external_aggr | base (`4048898 (merge-base)`) | 136.9 MiB | 432.8 MiB | 295.9 MiB | 3.2× |
| external_aggr | changed (`claude/datafusion-issue-25116-75677f`) | 136.6 MiB | 427.4 MiB | 290.8 MiB | 3.1× |
</details>
<details><summary>Resource Usage</summary>
**external_aggr — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 530.1s |
| Peak memory | 432.8 MiB |
| Avg memory | 8.1 MiB |
| CPU user | 25.6s |
| CPU sys | 3.8s |
| Peak spill | 0 B |
**external_aggr — branch**
| Metric | Value |
|--------|-------|
| Wall time | 530.1s |
| Peak memory | 427.4 MiB |
| Avg memory | 8.6 MiB |
| CPU user | 25.2s |
| CPU sys | 3.8s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611784446
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611482086)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 757.06 ms │ 876.55 ms │ 1.16x slower │
│ QQuery 1 │ 194.07 ms │ 194.39 ms │ no change │
│ QQuery 2 │ 460.41 ms │ 463.42 ms │ no change │
│ QQuery 3 │ 313.83 ms │ 315.91 ms │ no change │
│ QQuery 4 │ 1139.91 ms │ 1214.99 ms │ 1.07x slower │
│ QQuery 5 │ 10830.56 ms │ 10577.75 ms │ no change │
│ QQuery 6 │ 2.58 ms │ 2.57 ms │ no change │
│ QQuery 7 │ 693.11 ms │ 685.55 ms │ no change │
│ QQuery 8 │ 412.47 ms │ 408.28 ms │ no change │
│ QQuery 9 │ 2999.10 ms │ 2923.03 ms │ no change │
│ QQuery 10 │ 649.48 ms │ 633.83 ms │ no change │
│ QQuery 11 │ 2098.13 ms │ 2125.61 ms │ no change │
│ QQuery 12 │ 189.97 ms │ 184.37 ms │ no change │
│ QQuery 13 │ 553.75 ms │ 543.97 ms │ no change │
│ QQuery 14 │ 2825.36 ms │ 2356.74 ms │ +1.20x faster │
└───────────┴─────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 24119.79ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 23506.95ms │
│ Average Time (HEAD) │ 1607.99ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1567.13ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 12 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_extended.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 757.06 / 772.43 ±10.28 / 787.14 ms │ 876.55 / 900.21 ±17.87 / 923.36 ms │ 1.17x slower │
│ QQuery 1 │ 194.07 / 196.92 ±4.07 / 204.99 ms │ 194.39 / 195.57 ±1.62 / 198.78 ms │ no change │
│ QQuery 2 │ 460.41 / 462.93 ±1.79 / 465.03 ms │ 463.42 / 467.55 ±2.30 / 470.00 ms │ no change │
│ QQuery 3 │ 313.83 / 315.11 ±1.22 / 317.06 ms │ 315.91 / 316.51 ±0.42 / 317.14 ms │ no change │
│ QQuery 4 │ 1139.91 / 1156.87 ±17.43 / 1185.34 ms │ 1214.99 / 1230.90 ±17.79 / 1263.08 ms │ 1.06x slower │
│ QQuery 5 │ 10830.56 / 11203.19 ±264.89 / 11631.23 ms │ 10577.75 / 10834.78 ±212.37 / 11193.06 ms │ no change │
│ QQuery 6 │ 2.58 / 2.79 ±0.27 / 3.33 ms │ 2.57 / 2.81 ±0.41 / 3.62 ms │ no change │
│ QQuery 7 │ 693.11 / 732.58 ±28.36 / 768.91 ms │ 685.55 / 735.12 ±43.66 / 784.98 ms │ no change │
│ QQuery 8 │ 412.47 / 434.08 ±29.94 / 493.28 ms │ 408.28 / 441.11 ±64.90 / 570.91 ms │ no change │
│ QQuery 9 │ 2999.10 / 3056.54 ±49.80 / 3138.52 ms │ 2923.03 / 3007.69 ±50.91 / 3057.11 ms │ no change │
│ QQuery 10 │ 649.48 / 730.80 ±143.87 / 1018.10 ms │ 633.83 / 660.58 ±34.71 / 728.23 ms │ +1.11x faster │
│ QQuery 11 │ 2098.13 / 2229.50 ±90.19 / 2370.89 ms │ 2125.61 / 2212.14 ±51.51 / 2276.54 ms │ no change │
│ QQuery 12 │ 189.97 / 225.24 ±53.83 / 331.25 ms │ 184.37 / 222.04 ±70.71 / 363.41 ms │ no change │
│ QQuery 13 │ 553.75 / 571.81 ±15.63 / 600.52 ms │ 543.97 / 552.18 ±8.68 / 565.86 ms │ no change │
│ QQuery 14 │ 2825.36 / 2949.25 ±81.16 / 3046.54 ms │ 2356.74 / 2416.33 ±47.18 / 2471.39 ms │ +1.22x faster │
└───────────┴───────────────────────────────────────────┴───────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 25040.05ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 24195.52ms │
│ Average Time (HEAD) │ 1669.34ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 1613.03ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 11 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_extended — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 130.0s |
| Peak memory | 11.4 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1134.5s |
| CPU sys | 59.0s |
| Peak spill | 0 B |
**clickbench_extended — branch**
| Metric | Value |
|--------|-------|
| Wall time | 125.0s |
| Peak memory | 12.4 GiB |
| Avg memory | 5.1 GiB |
| CPU user | 1081.5s |
| CPU sys | 55.4s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613969431
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5613943332)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5613943332-2299-2s44w 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605488049
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpch
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1 │ 38.39 ms │ 37.98 ms │ no change │
│ QQuery 2 │ 19.05 ms │ 18.65 ms │ no change │
│ QQuery 3 │ 28.74 ms │ 28.36 ms │ no change │
│ QQuery 4 │ 17.98 ms │ 17.74 ms │ no change │
│ QQuery 5 │ 36.00 ms │ 35.60 ms │ no change │
│ QQuery 6 │ 15.76 ms │ 16.11 ms │ no change │
│ QQuery 7 │ 40.61 ms │ 40.62 ms │ no change │
│ QQuery 8 │ 40.90 ms │ 40.60 ms │ no change │
│ QQuery 9 │ 48.23 ms │ 49.08 ms │ no change │
│ QQuery 10 │ 41.99 ms │ 42.05 ms │ no change │
│ QQuery 11 │ 13.51 ms │ 13.51 ms │ no change │
│ QQuery 12 │ 23.73 ms │ 23.45 ms │ no change │
│ QQuery 13 │ 39.43 ms │ 39.60 ms │ no change │
│ QQuery 14 │ 24.22 ms │ 24.50 ms │ no change │
│ QQuery 15 │ 30.43 ms │ 30.74 ms │ no change │
│ QQuery 16 │ 13.86 ms │ 13.83 ms │ no change │
│ QQuery 17 │ 70.83 ms │ 70.26 ms │ no change │
│ QQuery 18 │ 60.13 ms │ 58.33 ms │ no change │
│ QQuery 19 │ 32.62 ms │ 32.51 ms │ no change │
│ QQuery 20 │ 31.16 ms │ 30.99 ms │ no change │
│ QQuery 21 │ 55.51 ms │ 55.05 ms │ no change │
│ QQuery 22 │ 14.10 ms │ 14.21 ms │ no change │
└───────────┴──────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 737.19ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 733.75ms │
│ Average Time (HEAD) │ 33.51ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 33.35ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 22 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴──────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1 │ 38.39 / 40.29 ±2.06 / 43.43 ms │ 37.98 / 39.23 ±2.24 / 43.70 ms │ no change │
│ QQuery 2 │ 19.05 / 19.19 ±0.18 / 19.53 ms │ 18.65 / 18.86 ±0.13 / 19.03 ms │ no change │
│ QQuery 3 │ 28.74 / 29.03 ±0.25 / 29.43 ms │ 28.36 / 28.70 ±0.31 / 29.21 ms │ no change │
│ QQuery 4 │ 17.98 / 18.07 ±0.08 / 18.20 ms │ 17.74 / 17.89 ±0.12 / 18.10 ms │ no change │
│ QQuery 5 │ 36.00 / 36.54 ±0.46 / 37.22 ms │ 35.60 / 35.78 ±0.17 / 36.04 ms │ no change │
│ QQuery 6 │ 15.76 / 15.95 ±0.15 / 16.14 ms │ 16.11 / 16.63 ±0.96 / 18.55 ms │ no change │
│ QQuery 7 │ 40.61 / 42.70 ±1.53 / 45.23 ms │ 40.62 / 41.94 ±1.46 / 44.67 ms │ no change │
│ QQuery 8 │ 40.90 / 41.63 ±0.70 / 42.64 ms │ 40.60 / 42.49 ±2.37 / 47.18 ms │ no change │
│ QQuery 9 │ 48.23 / 49.70 ±1.10 / 51.01 ms │ 49.08 / 50.22 ±0.67 / 50.86 ms │ no change │
│ QQuery 10 │ 41.99 / 42.19 ±0.12 / 42.35 ms │ 42.05 / 42.21 ±0.22 / 42.64 ms │ no change │
│ QQuery 11 │ 13.51 / 13.79 ±0.31 / 14.32 ms │ 13.51 / 13.73 ±0.13 / 13.85 ms │ no change │
│ QQuery 12 │ 23.73 / 23.97 ±0.18 / 24.20 ms │ 23.45 / 24.23 ±0.46 / 24.89 ms │ no change │
│ QQuery 13 │ 39.43 / 39.82 ±0.22 / 40.10 ms │ 39.60 / 40.54 ±1.08 / 42.32 ms │ no change │
│ QQuery 14 │ 24.22 / 24.51 ±0.19 / 24.74 ms │ 24.50 / 24.60 ±0.10 / 24.72 ms │ no change │
│ QQuery 15 │ 30.43 / 30.81 ±0.27 / 31.15 ms │ 30.74 / 31.03 ±0.16 / 31.20 ms │ no change │
│ QQuery 16 │ 13.86 / 13.97 ±0.15 / 14.26 ms │ 13.83 / 13.98 ±0.18 / 14.32 ms │ no change │
│ QQuery 17 │ 70.83 / 71.78 ±1.09 / 73.63 ms │ 70.26 / 71.65 ±1.59 / 74.63 ms │ no change │
│ QQuery 18 │ 60.13 / 61.20 ±1.01 / 62.46 ms │ 58.33 / 59.84 ±1.09 / 61.46 ms │ no change │
│ QQuery 19 │ 32.62 / 33.53 ±1.13 / 35.70 ms │ 32.51 / 32.74 ±0.23 / 33.11 ms │ no change │
│ QQuery 20 │ 31.16 / 31.91 ±0.47 / 32.49 ms │ 30.99 / 31.52 ±0.36 / 32.11 ms │ no change │
│ QQuery 21 │ 55.51 / 56.73 ±1.33 / 59.21 ms │ 55.05 / 56.42 ±1.30 / 58.66 ms │ no change │
│ QQuery 22 │ 14.10 / 14.24 ±0.11 / 14.41 ms │ 14.21 / 14.33 ±0.10 / 14.51 ms │ no change │
└───────────┴────────────────────────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 751.55ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 748.58ms │
│ Average Time (HEAD) │ 34.16ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 34.03ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 22 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴──────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 1.3 GiB |
| Avg memory | 510.1 MiB |
| CPU user | 21.0s |
| CPU sys | 1.6s |
| Peak spill | 0 B |
**tpch — branch**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 1.2 GiB |
| Avg memory | 489.7 MiB |
| CPU user | 20.9s |
| CPU sys | 1.7s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605546665
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 0 │ 1.20 ms │ 1.21 ms │ no change │
│ QQuery 1 │ 11.68 ms │ 11.76 ms │ no change │
│ QQuery 2 │ 35.92 ms │ 36.08 ms │ no change │
│ QQuery 3 │ 30.70 ms │ 30.55 ms │ no change │
│ QQuery 4 │ 223.75 ms │ 219.20 ms │ no change │
│ QQuery 5 │ 273.54 ms │ 270.89 ms │ no change │
│ QQuery 6 │ 1.24 ms │ 1.24 ms │ no change │
│ QQuery 7 │ 12.98 ms │ 12.90 ms │ no change │
│ QQuery 8 │ 323.33 ms │ 321.66 ms │ no change │
│ QQuery 9 │ 436.33 ms │ 452.30 ms │ no change │
│ QQuery 10 │ 70.33 ms │ 68.66 ms │ no change │
│ QQuery 11 │ 81.32 ms │ 80.12 ms │ no change │
│ QQuery 12 │ 264.58 ms │ 264.56 ms │ no change │
│ QQuery 13 │ 363.34 ms │ 360.63 ms │ no change │
│ QQuery 14 │ 279.69 ms │ 280.77 ms │ no change │
│ QQuery 15 │ 268.77 ms │ 263.95 ms │ no change │
│ QQuery 16 │ 607.40 ms │ 613.58 ms │ no change │
│ QQuery 17 │ 610.19 ms │ 614.54 ms │ no change │
│ QQuery 18 │ 1249.79 ms │ 1265.66 ms │ no change │
│ QQuery 19 │ 27.01 ms │ 27.16 ms │ no change │
│ QQuery 20 │ 514.29 ms │ 514.97 ms │ no change │
│ QQuery 21 │ 512.30 ms │ 511.36 ms │ no change │
│ QQuery 22 │ 982.52 ms │ 985.65 ms │ no change │
│ QQuery 23 │ 3073.48 ms │ 3008.62 ms │ no change │
│ QQuery 24 │ 41.95 ms │ 40.79 ms │ no change │
│ QQuery 25 │ 110.31 ms │ 110.85 ms │ no change │
│ QQuery 26 │ 41.04 ms │ 41.09 ms │ no change │
│ QQuery 27 │ 513.24 ms │ 513.91 ms │ no change │
│ QQuery 28 │ 2885.33 ms │ 2907.21 ms │ no change │
│ QQuery 29 │ 40.90 ms │ 41.10 ms │ no change │
│ QQuery 30 │ 299.05 ms │ 302.21 ms │ no change │
│ QQuery 31 │ 273.89 ms │ 280.04 ms │ no change │
│ QQuery 32 │ 919.74 ms │ 914.38 ms │ no change │
│ QQuery 33 │ 1448.94 ms │ 1449.75 ms │ no change │
│ QQuery 34 │ 1439.74 ms │ 1445.86 ms │ no change │
│ QQuery 35 │ 272.66 ms │ 271.80 ms │ no change │
│ QQuery 36 │ 66.34 ms │ 65.64 ms │ no change │
│ QQuery 37 │ 35.08 ms │ 35.80 ms │ no change │
│ QQuery 38 │ 41.09 ms │ 42.82 ms │ no change │
│ QQuery 39 │ 135.05 ms │ 136.92 ms │ no change │
│ QQuery 40 │ 13.99 ms │ 13.88 ms │ no change │
│ QQuery 41 │ 13.38 ms │ 13.43 ms │ no change │
│ QQuery 42 │ 12.83 ms │ 13.04 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 18860.20ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 18858.53ms │
│ Average Time (HEAD) │ 438.61ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 438.57ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 43 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.20 / 3.92 ±5.37 / 14.67 ms │ 1.21 / 3.92 ±5.36 / 14.64 ms │ no change │
│ QQuery 1 │ 11.68 / 11.89 ±0.13 / 12.07 ms │ 11.76 / 11.90 ±0.16 / 12.15 ms │ no change │
│ QQuery 2 │ 35.92 / 36.38 ±0.31 / 36.75 ms │ 36.08 / 36.20 ±0.11 / 36.38 ms │ no change │
│ QQuery 3 │ 30.70 / 31.23 ±0.76 / 32.74 ms │ 30.55 / 30.76 ±0.17 / 30.90 ms │ no change │
│ QQuery 4 │ 223.75 / 225.83 ±2.50 / 230.24 ms │ 219.20 / 222.28 ±2.47 / 224.92 ms │ no change │
│ QQuery 5 │ 273.54 / 275.84 ±1.38 / 277.22 ms │ 270.89 / 274.52 ±1.94 / 276.20 ms │ no change │
│ QQuery 6 │ 1.24 / 1.39 ±0.22 / 1.82 ms │ 1.24 / 1.39 ±0.22 / 1.82 ms │ no change │
│ QQuery 7 │ 12.98 / 13.20 ±0.15 / 13.44 ms │ 12.90 / 12.96 ±0.05 / 13.04 ms │ no change │
│ QQuery 8 │ 323.33 / 327.90 ±2.64 / 331.31 ms │ 321.66 / 328.05 ±4.64 / 336.04 ms │ no change │
│ QQuery 9 │ 436.33 / 446.60 ±6.42 / 456.32 ms │ 452.30 / 457.84 ±4.02 / 463.66 ms │ no change │
│ QQuery 10 │ 70.33 / 70.60 ±0.27 / 71.10 ms │ 68.66 / 71.49 ±4.42 / 80.27 ms │ no change │
│ QQuery 11 │ 81.32 / 83.75 ±3.78 / 91.26 ms │ 80.12 / 80.61 ±0.49 / 81.55 ms │ no change │
│ QQuery 12 │ 264.58 / 269.45 ±4.69 / 277.23 ms │ 264.56 / 270.88 ±4.62 / 277.10 ms │ no change │
│ QQuery 13 │ 363.34 / 376.70 ±17.85 / 411.08 ms │ 360.63 / 366.98 ±4.80 / 371.78 ms │ no change │
│ QQuery 14 │ 279.69 / 282.78 ±2.84 / 287.23 ms │ 280.77 / 289.16 ±6.76 / 299.94 ms │ no change │
│ QQuery 15 │ 268.77 / 278.23 ±6.15 / 285.01 ms │ 263.95 / 270.12 ±5.12 / 278.99 ms │ no change │
│ QQuery 16 │ 607.40 / 614.92 ±5.18 / 621.58 ms │ 613.58 / 627.29 ±9.07 / 640.74 ms │ no change │
│ QQuery 17 │ 610.19 / 620.10 ±6.77 / 630.95 ms │ 614.54 / 627.60 ±7.11 / 636.09 ms │ no change │
│ QQuery 18 │ 1249.79 / 1265.87 ±13.59 / 1283.47 ms │ 1265.66 / 1286.53 ±21.35 / 1326.73 ms │ no change │
│ QQuery 19 │ 27.01 / 28.83 ±3.38 / 35.59 ms │ 27.16 / 27.33 ±0.20 / 27.72 ms │ +1.05x faster │
│ QQuery 20 │ 514.29 / 534.27 ±23.57 / 578.97 ms │ 514.97 / 520.04 ±4.14 / 524.21 ms │ no change │
│ QQuery 21 │ 512.30 / 520.35 ±6.45 / 529.44 ms │ 511.36 / 518.86 ±3.99 / 522.35 ms │ no change │
│ QQuery 22 │ 982.52 / 990.81 ±10.18 / 1010.40 ms │ 985.65 / 994.69 ±7.00 / 1006.53 ms │ no change │
│ QQuery 23 │ 3073.48 / 3100.51 ±20.37 / 3132.57 ms │ 3008.62 / 3038.40 ±25.67 / 3078.24 ms │ no change │
│ QQuery 24 │ 41.95 / 44.47 ±2.23 / 47.50 ms │ 40.79 / 43.14 ±3.83 / 50.75 ms │ no change │
│ QQuery 25 │ 110.31 / 114.47 ±7.19 / 128.82 ms │ 110.85 / 115.19 ±5.11 / 124.73 ms │ no change │
│ QQuery 26 │ 41.04 / 47.74 ±12.53 / 72.78 ms │ 41.09 / 44.57 ±4.72 / 53.89 ms │ +1.07x faster │
│ QQuery 27 │ 513.24 / 531.44 ±15.99 / 558.31 ms │ 513.91 / 521.98 ±7.28 / 534.76 ms │ no change │
│ QQuery 28 │ 2885.33 / 2927.59 ±23.81 / 2951.81 ms │ 2907.21 / 2942.99 ±28.88 / 2985.56 ms │ no change │
│ QQuery 29 │ 40.90 / 44.50 ±6.52 / 57.52 ms │ 41.10 / 63.55 ±35.85 / 134.16 ms │ 1.43x slower │
│ QQuery 30 │ 299.05 / 309.47 ±8.64 / 324.47 ms │ 302.21 / 313.40 ±15.64 / 343.52 ms │ no change │
│ QQuery 31 │ 273.89 / 297.94 ±19.00 / 325.28 ms │ 280.04 / 295.66 ±13.39 / 318.01 ms │ no change │
│ QQuery 32 │ 919.74 / 947.12 ±15.67 / 968.18 ms │ 914.38 / 950.97 ±30.05 / 990.27 ms │ no change │
│ QQuery 33 │ 1448.94 / 1483.74 ±33.28 / 1546.00 ms │ 1449.75 / 1479.28 ±25.52 / 1519.70 ms │ no change │
│ QQuery 34 │ 1439.74 / 1495.03 ±33.86 / 1528.79 ms │ 1445.86 / 1502.97 ±50.94 / 1597.05 ms │ no change │
│ QQuery 35 │ 272.66 / 300.25 ±37.44 / 373.98 ms │ 271.80 / 298.23 ±26.73 / 348.57 ms │ no change │
│ QQuery 36 │ 66.34 / 67.15 ±0.94 / 68.41 ms │ 65.64 / 72.54 ±8.29 / 88.28 ms │ 1.08x slower │
│ QQuery 37 │ 35.08 / 35.45 ±0.45 / 36.30 ms │ 35.80 / 36.72 ±1.30 / 39.29 ms │ no change │
│ QQuery 38 │ 41.09 / 45.81 ±4.45 / 53.82 ms │ 42.82 / 46.85 ±5.50 / 57.71 ms │ no change │
│ QQuery 39 │ 135.05 / 151.92 ±10.88 / 168.86 ms │ 136.92 / 151.83 ±12.28 / 168.88 ms │ no change │
│ QQuery 40 │ 13.99 / 16.45 ±4.62 / 25.69 ms │ 13.88 / 16.97 ±4.96 / 26.86 ms │ no change │
│ QQuery 41 │ 13.38 / 16.88 ±4.09 / 21.95 ms │ 13.43 / 13.65 ±0.20 / 13.93 ms │ +1.24x faster │
│ QQuery 42 │ 12.83 / 13.33 ±0.65 / 14.57 ms │ 13.04 / 15.45 ±4.46 / 24.36 ms │ 1.16x slower │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19302.11ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19295.70ms │
│ Average Time (HEAD) │ 448.89ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 448.74ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 37 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 12.2 GiB |
| Avg memory | 4.6 GiB |
| CPU user | 987.0s |
| CPU sys | 68.2s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 12.0 GiB |
| Avg memory | 4.2 GiB |
| CPU user | 985.0s |
| CPU sys | 71.7s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607825052
run benchmark clickbench_extended
baseline:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
changed:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767719383
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767675986)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767675986-2565-sv9x9 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767676730
run benchmark clickbench_extended
baseline:
ref: 0576a0b400437ade5a6f3b102465d954a539f263
changed:
ref: 0576a0b400437ade5a6f3b102465d954a539f263
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5775921247
Ok, it looks like we have nicely [Occam'd ](https://en.wikipedia.org/wiki/Occam%27s_razor)this 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767719444
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767676454)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767676454-2570-zl5gz 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767718359
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767675750)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5767675750-2562-2mgm8 6.12.94+ #1 SMP Tue Aug 4 08:44:15 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 0576a0b400437ade5a6f3b102465d954a539f263 (0576a0b400437ade5a6f3b102465d954a539f263) to 0576a0b400437ade5a6f3b102465d954a539f263 [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..0576a0b400437ade5a6f3b102465d954a539f263)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_extended
baseline:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
changed:
ref: "0576a0b400437ade5a6f3b102465d954a539f263"
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767958137
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5767677168)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (917e7b25b3ddb0091d6b5eac1bd98212e2c685b0) to 0576a0b (merge-base) [diff](https://github.com/apache/datafusion/compare/0576a0b400437ade5a6f3b102465d954a539f263..917e7b25b3ddb0091d6b5eac1bd98212e2c685b0)
<details><summary>Run configuration</summary>
```yaml
run benchmark tpcds
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 5.61 ms │ 6.03 ms │ 1.07x slower │
│ QQuery 2 │ 81.96 ms │ 81.59 ms │ no change │
│ QQuery 3 │ 29.22 ms │ 29.02 ms │ no change │
│ QQuery 4 │ 461.07 ms │ 468.33 ms │ no change │
│ QQuery 5 │ 51.80 ms │ 52.70 ms │ no change │
│ QQuery 6 │ 36.51 ms │ 36.31 ms │ no change │
│ QQuery 7 │ 74.80 ms │ 74.29 ms │ no change │
│ QQuery 8 │ 36.55 ms │ 36.78 ms │ no change │
│ QQuery 9 │ 51.87 ms │ 54.97 ms │ 1.06x slower │
│ QQuery 10 │ 62.01 ms │ 62.18 ms │ no change │
│ QQuery 11 │ 287.62 ms │ 300.05 ms │ no change │
│ QQuery 12 │ 29.79 ms │ 29.66 ms │ no change │
│ QQuery 13 │ 119.07 ms │ 119.60 ms │ no change │
│ QQuery 14 │ 397.80 ms │ 399.50 ms │ no change │
│ QQuery 15 │ 54.86 ms │ 57.05 ms │ no change │
│ QQuery 16 │ 6.50 ms │ 6.74 ms │ no change │
│ QQuery 17 │ 79.22 ms │ 79.25 ms │ no change │
│ QQuery 18 │ 104.73 ms │ 106.48 ms │ no change │
│ QQuery 19 │ 42.12 ms │ 41.75 ms │ no change │
│ QQuery 20 │ 35.95 ms │ 36.24 ms │ no change │
│ QQuery 21 │ 17.21 ms │ 17.50 ms │ no change │
│ QQuery 22 │ 64.51 ms │ 64.68 ms │ no change │
│ QQuery 23 │ 317.91 ms │ 317.24 ms │ no change │
│ QQuery 24 │ 198.65 ms │ 201.39 ms │ no change │
│ QQuery 25 │ 108.55 ms │ 109.06 ms │ no change │
│ QQuery 26 │ 48.93 ms │ 48.52 ms │ no change │
│ QQuery 27 │ 6.30 ms │ 6.32 ms │ no change │
│ QQuery 28 │ 57.12 ms │ 57.22 ms │ no change │
│ QQuery 29 │ 94.93 ms │ 96.92 ms │ no change │
│ QQuery 30 │ 33.21 ms │ 33.24 ms │ no change │
│ QQuery 31 │ 109.29 ms │ 110.59 ms │ no change │
│ QQuery 32 │ 20.72 ms │ 21.58 ms │ no change │
│ QQuery 33 │ 37.88 ms │ 38.15 ms │ no change │
│ QQuery 34 │ 10.07 ms │ 10.27 ms │ no change │
│ QQuery 35 │ 72.40 ms │ 74.70 ms │ no change │
│ QQuery 36 │ 5.78 ms │ 5.92 ms │ no change │
│ QQuery 37 │ 7.02 ms │ 7.06 ms │ no change │
│ QQuery 38 │ 62.37 ms │ 63.60 ms │ no change │
│ QQuery 39 │ 73.79 ms │ 75.21 ms │ no change │
│ QQuery 40 │ 23.82 ms │ 24.75 ms │ no change │
│ QQuery 41 │ 11.40 ms │ 11.75 ms │ no change │
│ QQuery 42 │ 23.67 ms │ 24.37 ms │ no change │
│ QQuery 43 │ 5.12 ms │ 5.47 ms │ 1.07x slower │
│ QQuery 44 │ 9.50 ms │ 9.64 ms │ no change │
│ QQuery 45 │ 38.91 ms │ 40.89 ms │ 1.05x slower │
│ QQuery 46 │ 12.25 ms │ 12.41 ms │ no change │
│ QQuery 47 │ 230.99 ms │ 231.31 ms │ no change │
│ QQuery 48 │ 96.41 ms │ 95.42 ms │ no change │
│ QQuery 49 │ 71.65 ms │ 71.85 ms │ no change │
│ QQuery 50 │ 58.60 ms │ 58.73 ms │ no change │
│ QQuery 51 │ 92.18 ms │ 93.28 ms │ no change │
│ QQuery 52 │ 24.76 ms │ 24.73 ms │ no change │
│ QQuery 53 │ 29.47 ms │ 29.70 ms │ no change │
│ QQuery 54 │ 54.86 ms │ 54.79 ms │ no change │
│ QQuery 55 │ 24.00 ms │ 23.74 ms │ no change │
│ QQuery 56 │ 38.95 ms │ 38.60 ms │ no change │
│ QQuery 57 │ 172.89 ms │ 169.65 ms │ no change │
│ QQuery 58 │ 111.12 ms │ 108.97 ms │ no change │
│ QQuery 59 │ 118.32 ms │ 116.76 ms │ no change │
│ QQuery 60 │ 38.84 ms │ 39.14 ms │ no change │
│ QQuery 61 │ 12.03 ms │ 12.10 ms │ no change │
│ QQuery 62 │ 45.35 ms │ 44.66 ms │ no change │
│ QQuery 63 │ 29.71 ms │ 29.63 ms │ no change │
│ QQuery 64 │ 360.43 ms │ 359.79 ms │ no change │
│ QQuery 65 │ 129.23 ms │ 134.04 ms │ no change │
│ QQuery 66 │ 77.87 ms │ 78.41 ms │ no change │
│ QQuery 67 │ 253.74 ms │ 257.43 ms │ no change │
│ QQuery 68 │ 12.11 ms │ 11.93 ms │ no change │
│ QQuery 69 │ 56.74 ms │ 56.79 ms │ no change │
│ QQuery 70 │ 105.36 ms │ 104.76 ms │ no change │
│ QQuery 71 │ 35.25 ms │ 35.29 ms │ no change │
│ QQuery 72 │ 1816.39 ms │ 1715.67 ms │ +1.06x faster │
│ QQuery 73 │ 10.11 ms │ 10.28 ms │ no change │
│ QQuery 74 │ 167.58 ms │ 166.41 ms │ no change │
│ QQuery 75 │ 143.09 ms │ 144.67 ms │ no change │
│ QQuery 76 │ 35.34 ms │ 34.64 ms │ no change │
│ QQuery 77 │ 61.12 ms │ 61.35 ms │ no change │
│ QQuery 78 │ 164.82 ms │ 163.33 ms │ no change │
│ QQuery 79 │ 66.57 ms │ 67.63 ms │ no change │
│ QQuery 80 │ 96.20 ms │ 96.69 ms │ no change │
│ QQuery 81 │ 26.37 ms │ 26.18 ms │ no change │
│ QQuery 82 │ 16.94 ms │ 16.90 ms │ no change │
│ QQuery 83 │ 33.87 ms │ 34.38 ms │ no change │
│ QQuery 84 │ 29.87 ms │ 30.24 ms │ no change │
│ QQuery 85 │ 104.12 ms │ 104.52 ms │ no change │
│ QQuery 86 │ 26.00 ms │ 25.78 ms │ no change │
│ QQuery 87 │ 63.24 ms │ 63.37 ms │ no change │
│ QQuery 88 │ 61.87 ms │ 61.36 ms │ no change │
│ QQuery 89 │ 35.81 ms │ 35.34 ms │ no change │
│ QQuery 90 │ 16.81 ms │ 17.03 ms │ no change │
│ QQuery 91 │ 44.87 ms │ 44.91 ms │ no change │
│ QQuery 92 │ 29.86 ms │ 29.82 ms │ no change │
│ QQuery 93 │ 48.77 ms │ 48.92 ms │ no change │
│ QQuery 94 │ 38.98 ms │ 39.10 ms │ no change │
│ QQuery 95 │ 80.42 ms │ 81.29 ms │ no change │
│ QQuery 96 │ 24.07 ms │ 24.07 ms │ no change │
│ QQuery 97 │ 51.37 ms │ 52.32 ms │ no change │
│ QQuery 98 │ 43.18 ms │ 43.09 ms │ no change │
│ QQuery 99 │ 67.08 ms │ 66.19 ms │ no change │
└───────────┴────────────┴──────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 9205.93ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 9153.96ms │
│ Average Time (HEAD) │ 92.99ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 92.46ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 4 │
│ Queries with No Change │ 94 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 5.61 / 6.26 ±0.98 / 8.21 ms │ 6.03 / 6.61 ±1.04 / 8.68 ms │ 1.06x slower │
│ QQuery 2 │ 81.96 / 82.11 ±0.23 / 82.57 ms │ 81.59 / 81.97 ±0.39 / 82.71 ms │ no change │
│ QQuery 3 │ 29.22 / 29.42 ±0.14 / 29.62 ms │ 29.02 / 29.32 ±0.19 / 29.63 ms │ no change │
│ QQuery 4 │ 461.07 / 468.97 ±6.93 / 481.59 ms │ 468.33 / 479.62 ±9.03 / 491.91 ms │ no change │
│ QQuery 5 │ 51.80 / 52.92 ±1.05 / 54.88 ms │ 52.70 / 53.27 ±0.41 / 53.72 ms │ no change │
│ QQuery 6 │ 36.51 / 36.74 ±0.15 / 36.93 ms │ 36.31 / 36.98 ±0.46 / 37.64 ms │ no change │
│ QQuery 7 │ 74.80 / 75.40 ±0.57 / 76.39 ms │ 74.29 / 75.01 ±0.60 / 75.93 ms │ no change │
│ QQuery 8 │ 36.55 / 36.96 ±0.36 / 37.44 ms │ 36.78 / 36.99 ±0.17 / 37.31 ms │ no change │
│ QQuery 9 │ 51.87 / 56.69 ±5.41 / 66.92 ms │ 54.97 / 58.40 ±3.76 / 65.23 ms │ no change │
│ QQuery 10 │ 62.01 / 62.48 ±0.25 / 62.73 ms │ 62.18 / 63.12 ±0.64 / 63.97 ms │ no change │
│ QQuery 11 │ 287.62 / 289.66 ±1.96 / 293.36 ms │ 300.05 / 305.40 ±2.94 / 309.08 ms │ 1.05x slower │
│ QQuery 12 │ 29.79 / 31.31 ±2.41 / 36.12 ms │ 29.66 / 30.23 ±0.39 / 30.62 ms │ no change │
│ QQuery 13 │ 119.07 / 119.95 ±1.01 / 121.75 ms │ 119.60 / 120.36 ±0.76 / 121.77 ms │ no change │
│ QQuery 14 │ 397.80 / 401.06 ±2.39 / 404.25 ms │ 399.50 / 401.85 ±1.66 / 404.07 ms │ no change │
│ QQuery 15 │ 54.86 / 56.65 ±1.87 / 60.13 ms │ 57.05 / 59.38 ±2.59 / 64.36 ms │ no change │
│ QQuery 16 │ 6.50 / 6.77 ±0.18 / 7.04 ms │ 6.74 / 6.95 ±0.20 / 7.30 ms │ no change │
│ QQuery 17 │ 79.22 / 80.25 ±0.78 / 81.60 ms │ 79.25 / 80.43 ±0.97 / 82.03 ms │ no change │
│ QQuery 18 │ 104.73 / 106.63 ±2.08 / 110.59 ms │ 106.48 / 107.30 ±0.75 / 108.34 ms │ no change │
│ QQuery 19 │ 42.12 / 43.47 ±1.44 / 46.18 ms │ 41.75 / 42.62 ±0.49 / 43.02 ms │ no change │
│ QQuery 20 │ 35.95 / 36.32 ±0.27 / 36.69 ms │ 36.24 / 36.84 ±0.44 / 37.58 ms │ no change │
│ QQuery 21 │ 17.21 / 17.57 ±0.25 / 17.83 ms │ 17.50 / 18.49 ±1.56 / 21.59 ms │ 1.05x slower │
│ QQuery 22 │ 64.51 / 65.18 ±0.91 / 66.91 ms │ 64.68 / 66.25 ±0.99 / 67.67 ms │ no change │
│ QQuery 23 │ 317.91 / 320.88 ±2.03 / 323.84 ms │ 317.24 / 324.87 ±4.73 / 332.15 ms │ no change │
│ QQuery 24 │ 198.65 / 201.23 ±2.38 / 205.52 ms │ 201.39 / 204.45 ±2.37 / 208.19 ms │ no change │
│ QQuery 25 │ 108.55 / 109.78 ±1.53 / 112.71 ms │ 109.06 / 112.30 ±2.96 / 117.72 ms │ no change │
│ QQuery 26 │ 48.93 / 49.67 ±0.73 / 50.92 ms │ 48.52 / 48.76 ±0.25 / 49.20 ms │ no change │
│ QQuery 27 │ 6.30 / 6.55 ±0.20 / 6.88 ms │ 6.32 / 6.55 ±0.23 / 6.98 ms │ no change │
│ QQuery 28 │ 57.12 / 59.80 ±2.10 / 62.38 ms │ 57.22 / 60.19 ±2.47 / 63.11 ms │ no change │
│ QQuery 29 │ 94.93 / 97.48 ±3.64 / 104.66 ms │ 96.92 / 98.58 ±1.62 / 101.31 ms │ no change │
│ QQuery 30 │ 33.21 / 33.70 ±0.63 / 34.91 ms │ 33.24 / 33.60 ±0.21 / 33.88 ms │ no change │
│ QQuery 31 │ 109.29 / 111.10 ±2.63 / 116.31 ms │ 110.59 / 112.53 ±1.14 / 113.67 ms │ no change │
│ QQuery 32 │ 20.72 / 20.90 ±0.13 / 21.11 ms │ 21.58 / 22.50 ±1.33 / 25.11 ms │ 1.08x slower │
│ QQuery 33 │ 37.88 / 38.07 ±0.11 / 38.20 ms │ 38.15 / 38.76 ±0.50 / 39.38 ms │ no change │
│ QQuery 34 │ 10.07 / 10.68 ±0.74 / 12.15 ms │ 10.27 / 10.52 ±0.14 / 10.67 ms │ no change │
│ QQuery 35 │ 72.40 / 73.39 ±0.89 / 74.50 ms │ 74.70 / 75.11 ±0.46 / 76.00 ms │ no change │
│ QQuery 36 │ 5.78 / 5.91 ±0.19 / 6.28 ms │ 5.92 / 6.13 ±0.21 / 6.52 ms │ no change │
│ QQuery 37 │ 7.02 / 7.17 ±0.08 / 7.24 ms │ 7.06 / 7.17 ±0.07 / 7.29 ms │ no change │
│ QQuery 38 │ 62.37 / 63.79 ±1.87 / 67.48 ms │ 63.60 / 64.89 ±1.63 / 68.06 ms │ no change │
│ QQuery 39 │ 73.79 / 74.63 ±0.60 / 75.55 ms │ 75.21 / 75.78 ±0.55 / 76.58 ms │ no change │
│ QQuery 40 │ 23.82 / 24.20 ±0.41 / 24.92 ms │ 24.75 / 24.94 ±0.18 / 25.23 ms │ no change │
│ QQuery 41 │ 11.40 / 11.54 ±0.14 / 11.80 ms │ 11.75 / 11.84 ±0.07 / 11.93 ms │ no change │
│ QQuery 42 │ 23.67 / 24.52 ±0.59 / 25.42 ms │ 24.37 / 25.56 ±1.56 / 28.64 ms │ no change │
│ QQuery 43 │ 5.12 / 5.23 ±0.14 / 5.51 ms │ 5.47 / 5.66 ±0.16 / 5.85 ms │ 1.08x slower │
│ QQuery 44 │ 9.50 / 9.70 ±0.22 / 10.12 ms │ 9.64 / 9.76 ±0.10 / 9.88 ms │ no change │
│ QQuery 45 │ 38.91 / 40.19 ±0.98 / 41.53 ms │ 40.89 / 41.82 ±1.06 / 43.78 ms │ no change │
│ QQuery 46 │ 12.25 / 12.55 ±0.21 / 12.82 ms │ 12.41 / 12.66 ±0.31 / 13.26 ms │ no change │
│ QQuery 47 │ 230.99 / 233.94 ±2.02 / 236.17 ms │ 231.31 / 236.46 ±4.22 / 242.81 ms │ no change │
│ QQuery 48 │ 96.41 / 97.35 ±1.17 / 99.47 ms │ 95.42 / 96.29 ±0.90 / 97.95 ms │ no change │
│ QQuery 49 │ 71.65 / 72.38 ±0.64 / 73.49 ms │ 71.85 / 73.10 ±1.32 / 75.62 ms │ no change │
│ QQuery 50 │ 58.60 / 60.34 ±2.16 / 64.60 ms │ 58.73 / 59.13 ±0.39 / 59.87 ms │ no change │
│ QQuery 51 │ 92.18 / 94.47 ±1.70 / 97.03 ms │ 93.28 / 96.53 ±4.48 / 105.31 ms │ no change │
│ QQuery 52 │ 24.76 / 25.38 ±0.55 / 26.11 ms │ 24.73 / 24.90 ±0.14 / 25.15 ms │ no change │
│ QQuery 53 │ 29.47 / 29.89 ±0.31 / 30.25 ms │ 29.70 / 29.99 ±0.21 / 30.26 ms │ no change │
│ QQuery 54 │ 54.86 / 55.66 ±0.89 / 57.38 ms │ 54.79 / 55.35 ±0.51 / 56.11 ms │ no change │
│ QQuery 55 │ 24.00 / 24.52 ±0.34 / 24.89 ms │ 23.74 / 24.36 ±0.50 / 25.14 ms │ no change │
│ QQuery 56 │ 38.95 / 39.31 ±0.27 / 39.77 ms │ 38.60 / 39.27 ±0.58 / 40.14 ms │ no change │
│ QQuery 57 │ 172.89 / 174.69 ±1.93 / 178.30 ms │ 169.65 / 171.19 ±1.87 / 174.81 ms │ no change │
│ QQuery 58 │ 111.12 / 112.94 ±2.22 / 117.09 ms │ 108.97 / 111.89 ±3.75 / 119.22 ms │ no change │
│ QQuery 59 │ 118.32 / 118.94 ±0.49 / 119.79 ms │ 116.76 / 118.17 ±1.73 / 121.52 ms │ no change │
│ QQuery 60 │ 38.84 / 39.41 ±0.34 / 39.80 ms │ 39.14 / 40.12 ±1.10 / 42.26 ms │ no change │
│ QQuery 61 │ 12.03 / 12.15 ±0.09 / 12.31 ms │ 12.10 / 12.99 ±1.52 / 16.02 ms │ 1.07x slower │
│ QQuery 62 │ 45.35 / 45.69 ±0.20 / 45.91 ms │ 44.66 / 45.27 ±0.53 / 46.25 ms │ no change │
│ QQuery 63 │ 29.71 / 30.66 ±1.42 / 33.44 ms │ 29.63 / 29.98 ±0.26 / 30.41 ms │ no change │
│ QQuery 64 │ 360.43 / 364.54 ±2.17 / 366.70 ms │ 359.79 / 363.08 ±2.19 / 365.69 ms │ no change │
│ QQuery 65 │ 129.23 / 133.59 ±2.75 / 137.54 ms │ 134.04 / 135.04 ±1.62 / 138.27 ms │ no change │
│ QQuery 66 │ 77.87 / 81.26 ±5.08 / 91.36 ms │ 78.41 / 79.59 ±1.82 / 83.20 ms │ no change │
│ QQuery 67 │ 253.74 / 258.94 ±4.32 / 263.81 ms │ 257.43 / 259.08 ±1.77 / 262.20 ms │ no change │
│ QQuery 68 │ 12.11 / 12.23 ±0.10 / 12.41 ms │ 11.93 / 12.08 ±0.12 / 12.25 ms │ no change │
│ QQuery 69 │ 56.74 / 58.72 ±2.03 / 61.87 ms │ 56.79 / 58.95 ±3.79 / 66.51 ms │ no change │
│ QQuery 70 │ 105.36 / 105.96 ±0.53 / 106.66 ms │ 104.76 / 107.09 ±2.77 / 112.52 ms │ no change │
│ QQuery 71 │ 35.25 / 35.68 ±0.31 / 36.17 ms │ 35.29 / 35.81 ±0.56 / 36.88 ms │ no change │
│ QQuery 72 │ 1816.39 / 1909.08 ±80.80 / 2058.11 ms │ 1715.67 / 1811.44 ±62.75 / 1898.13 ms │ +1.05x faster │
│ QQuery 73 │ 10.11 / 10.49 ±0.23 / 10.75 ms │ 10.28 / 10.61 ±0.24 / 11.02 ms │ no change │
│ QQuery 74 │ 167.58 / 169.61 ±1.17 / 171.08 ms │ 166.41 / 167.71 ±1.74 / 171.12 ms │ no change │
│ QQuery 75 │ 143.09 / 144.86 ±1.81 / 148.18 ms │ 144.67 / 146.95 ±1.94 / 150.28 ms │ no change │
│ QQuery 76 │ 35.34 / 37.12 ±2.76 / 42.59 ms │ 34.64 / 35.21 ±0.41 / 35.79 ms │ +1.05x faster │
│ QQuery 77 │ 61.12 / 61.80 ±0.41 / 62.39 ms │ 61.35 / 64.48 ±5.39 / 75.20 ms │ no change │
│ QQuery 78 │ 164.82 / 168.66 ±2.98 / 173.76 ms │ 163.33 / 167.31 ±2.90 / 171.30 ms │ no change │
│ QQuery 79 │ 66.57 / 67.09 ±0.33 / 67.44 ms │ 67.63 / 69.54 ±2.67 / 74.83 ms │ no change │
│ QQuery 80 │ 96.20 / 98.97 ±3.48 / 105.74 ms │ 96.69 / 98.25 ±2.04 / 102.24 ms │ no change │
│ QQuery 81 │ 26.37 / 26.72 ±0.21 / 26.99 ms │ 26.18 / 26.60 ±0.27 / 27.02 ms │ no change │
│ QQuery 82 │ 16.94 / 17.17 ±0.16 / 17.37 ms │ 16.90 / 19.22 ±3.25 / 25.62 ms │ 1.12x slower │
│ QQuery 83 │ 33.87 / 34.34 ±0.26 / 34.66 ms │ 34.38 / 34.87 ±0.52 / 35.51 ms │ no change │
│ QQuery 84 │ 29.87 / 30.35 ±0.31 / 30.84 ms │ 30.24 / 31.51 ±1.25 / 33.85 ms │ no change │
│ QQuery 85 │ 104.12 / 108.02 ±3.87 / 115.39 ms │ 104.52 / 105.21 ±0.78 / 106.59 ms │ no change │
│ QQuery 86 │ 26.00 / 26.50 ±0.30 / 26.91 ms │ 25.78 / 26.10 ±0.19 / 26.38 ms │ no change │
│ QQuery 87 │ 63.24 / 63.96 ±0.60 / 64.74 ms │ 63.37 / 66.19 ±3.21 / 72.40 ms │ no change │
│ QQuery 88 │ 61.87 / 62.25 ±0.33 / 62.76 ms │ 61.36 / 62.24 ±0.69 / 63.47 ms │ no change │
│ QQuery 89 │ 35.81 / 37.30 ±2.45 / 42.19 ms │ 35.34 / 35.91 ±0.40 / 36.46 ms │ no change │
│ QQuery 90 │ 16.81 / 16.95 ±0.11 / 17.13 ms │ 17.03 / 17.16 ±0.10 / 17.33 ms │ no change │
│ QQuery 91 │ 44.87 / 45.31 ±0.54 / 46.33 ms │ 44.91 / 46.06 ±1.20 / 48.29 ms │ no change │
│ QQuery 92 │ 29.86 / 30.75 ±0.86 / 32.41 ms │ 29.82 / 30.77 ±0.71 / 31.90 ms │ no change │
│ QQuery 93 │ 48.77 / 50.29 ±1.48 / 53.11 ms │ 48.92 / 50.83 ±1.41 / 52.98 ms │ no change │
│ QQuery 94 │ 38.98 / 39.43 ±0.44 / 40.19 ms │ 39.10 / 39.52 ±0.40 / 40.20 ms │ no change │
│ QQuery 95 │ 80.42 / 84.35 ±4.10 / 92.30 ms │ 81.29 / 82.91 ±1.90 / 86.30 ms │ no change │
│ QQuery 96 │ 24.07 / 24.25 ±0.13 / 24.39 ms │ 24.07 / 24.23 ±0.14 / 24.42 ms │ no change │
│ QQuery 97 │ 51.37 / 53.91 ±2.19 / 56.66 ms │ 52.32 / 53.12 ±1.10 / 55.19 ms │ no change │
│ QQuery 98 │ 43.18 / 43.89 ±0.50 / 44.74 ms │ 43.09 / 43.96 ±0.66 / 45.07 ms │ no change │
│ QQuery 99 │ 67.08 / 69.34 ±4.01 / 77.34 ms │ 66.19 / 67.77 ±2.60 / 72.95 ms │ no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 9426.82ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 9383.73ms │
│ Average Time (HEAD) │ 95.22ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 94.79ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 7 │
│ Queries with No Change │ 90 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴───────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 50.0s |
| Peak memory | 2.0 GiB |
| Avg memory | 1.4 GiB |
| CPU user | 206.7s |
| CPU sys | 5.6s |
| Peak spill | 0 B |
**tpcds — branch**
| Metric | Value |
|--------|-------|
| Wall time | 50.0s |
| Peak memory | 2.2 GiB |
| Avg memory | 1.5 GiB |
| CPU user | 201.3s |
| CPU sys | 5.8s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611504278
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5611481841-2293-wk8br 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark external_aggr
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611659224
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing claude/datafusion-issue-25116-75677f (c41c95406817a4deb488c1d198bb5e78b6816c2d) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..c41c95406817a4deb488c1d198bb5e78b6816c2d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>
<details><summary>Details</summary>
<p>
```
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.22 ms │ 1.22 ms │ no change │
│ QQuery 1 │ 12.15 ms │ 11.66 ms │ no change │
│ QQuery 2 │ 36.25 ms │ 36.23 ms │ no change │
│ QQuery 3 │ 31.04 ms │ 30.65 ms │ no change │
│ QQuery 4 │ 229.95 ms │ 218.63 ms │ no change │
│ QQuery 5 │ 278.79 ms │ 272.57 ms │ no change │
│ QQuery 6 │ 1.27 ms │ 1.23 ms │ no change │
│ QQuery 7 │ 13.63 ms │ 13.01 ms │ no change │
│ QQuery 8 │ 336.20 ms │ 326.60 ms │ no change │
│ QQuery 9 │ 463.84 ms │ 452.80 ms │ no change │
│ QQuery 10 │ 71.12 ms │ 71.09 ms │ no change │
│ QQuery 11 │ 82.78 ms │ 83.11 ms │ no change │
│ QQuery 12 │ 276.66 ms │ 297.85 ms │ 1.08x slower │
│ QQuery 13 │ 372.42 ms │ 403.54 ms │ 1.08x slower │
│ QQuery 14 │ 290.66 ms │ 300.40 ms │ no change │
│ QQuery 15 │ 276.84 ms │ 325.62 ms │ 1.18x slower │
│ QQuery 16 │ 613.73 ms │ 708.62 ms │ 1.15x slower │
│ QQuery 17 │ 628.72 ms │ 698.16 ms │ 1.11x slower │
│ QQuery 18 │ 1297.06 ms │ 1436.57 ms │ 1.11x slower │
│ QQuery 19 │ 27.56 ms │ 29.37 ms │ 1.07x slower │
│ QQuery 20 │ 521.90 ms │ 534.02 ms │ no change │
│ QQuery 21 │ 511.40 ms │ 526.49 ms │ no change │
│ QQuery 22 │ 976.91 ms │ 1002.98 ms │ no change │
│ QQuery 23 │ 3060.41 ms │ 3100.91 ms │ no change │
│ QQuery 24 │ 40.77 ms │ 40.97 ms │ no change │
│ QQuery 25 │ 110.72 ms │ 111.52 ms │ no change │
│ QQuery 26 │ 41.64 ms │ 41.34 ms │ no change │
│ QQuery 27 │ 509.89 ms │ 516.75 ms │ no change │
│ QQuery 28 │ 2892.23 ms │ 2938.98 ms │ no change │
│ QQuery 29 │ 41.17 ms │ 41.82 ms │ no change │
│ QQuery 30 │ 302.27 ms │ 310.06 ms │ no change │
│ QQuery 31 │ 280.62 ms │ 288.45 ms │ no change │
│ QQuery 32 │ 953.09 ms │ 938.39 ms │ no change │
│ QQuery 33 │ 1444.36 ms │ 1486.30 ms │ no change │
│ QQuery 34 │ 1479.28 ms │ 1484.50 ms │ no change │
│ QQuery 35 │ 277.21 ms │ 293.77 ms │ 1.06x slower │
│ QQuery 36 │ 66.65 ms │ 69.86 ms │ no change │
│ QQuery 37 │ 35.36 ms │ 35.97 ms │ no change │
│ QQuery 38 │ 40.29 ms │ 44.09 ms │ 1.09x slower │
│ QQuery 39 │ 133.89 ms │ 148.40 ms │ 1.11x slower │
│ QQuery 40 │ 13.90 ms │ 14.83 ms │ 1.07x slower │
│ QQuery 41 │ 13.74 ms │ 14.65 ms │ 1.07x slower │
│ QQuery 42 │ 13.38 ms │ 14.25 ms │ 1.07x slower │
└───────────┴────────────┴──────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19102.96ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 19718.25ms │
│ Average Time (HEAD) │ 444.25ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 458.56ms │
│ Queries Faster │ 0 │
│ Queries Slower │ 13 │
│ Queries with No Change │ 30 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
Distribution per query (min / mean ±stddev / max):
Comparing HEAD and claude_datafusion-issue-25116-75677f
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ claude_datafusion-issue-25116-75677f ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.22 / 4.08 ±5.57 / 15.22 ms │ 1.22 / 3.90 ±5.30 / 14.51 ms │ no change │
│ QQuery 1 │ 12.15 / 12.52 ±0.23 / 12.85 ms │ 11.66 / 11.87 ±0.17 / 12.14 ms │ +1.05x faster │
│ QQuery 2 │ 36.25 / 36.72 ±0.26 / 36.98 ms │ 36.23 / 36.33 ±0.08 / 36.46 ms │ no change │
│ QQuery 3 │ 31.04 / 31.69 ±0.75 / 33.14 ms │ 30.65 / 30.94 ±0.28 / 31.31 ms │ no change │
│ QQuery 4 │ 229.95 / 233.08 ±2.28 / 236.18 ms │ 218.63 / 224.48 ±3.67 / 228.31 ms │ no change │
│ QQuery 5 │ 278.79 / 283.52 ±2.87 / 287.19 ms │ 272.57 / 275.40 ±3.41 / 281.34 ms │ no change │
│ QQuery 6 │ 1.27 / 1.41 ±0.22 / 1.85 ms │ 1.23 / 1.38 ±0.22 / 1.80 ms │ no change │
│ QQuery 7 │ 13.63 / 13.94 ±0.20 / 14.14 ms │ 13.01 / 13.20 ±0.17 / 13.41 ms │ +1.06x faster │
│ QQuery 8 │ 336.20 / 341.72 ±3.79 / 345.86 ms │ 326.60 / 331.69 ±4.13 / 339.22 ms │ no change │
│ QQuery 9 │ 463.84 / 472.40 ±7.29 / 485.53 ms │ 452.80 / 486.12 ±31.43 / 541.86 ms │ no change │
│ QQuery 10 │ 71.12 / 73.10 ±1.92 / 76.50 ms │ 71.09 / 71.85 ±0.48 / 72.36 ms │ no change │
│ QQuery 11 │ 82.78 / 84.20 ±1.97 / 88.09 ms │ 83.11 / 90.21 ±9.45 / 108.71 ms │ 1.07x slower │
│ QQuery 12 │ 276.66 / 282.94 ±5.78 / 292.42 ms │ 297.85 / 312.23 ±12.88 / 334.35 ms │ 1.10x slower │
│ QQuery 13 │ 372.42 / 383.05 ±8.14 / 396.72 ms │ 403.54 / 417.09 ±14.17 / 435.18 ms │ 1.09x slower │
│ QQuery 14 │ 290.66 / 295.19 ±4.61 / 302.52 ms │ 300.40 / 311.77 ±7.10 / 319.41 ms │ 1.06x slower │
│ QQuery 15 │ 276.84 / 287.64 ±8.51 / 298.39 ms │ 325.62 / 335.83 ±9.73 / 353.96 ms │ 1.17x slower │
│ QQuery 16 │ 613.73 / 630.42 ±8.96 / 640.71 ms │ 708.62 / 723.82 ±10.89 / 740.61 ms │ 1.15x slower │
│ QQuery 17 │ 628.72 / 638.30 ±9.24 / 655.14 ms │ 698.16 / 710.37 ±8.46 / 723.49 ms │ 1.11x slower │
│ QQuery 18 │ 1297.06 / 1314.45 ±11.18 / 1328.08 ms │ 1436.57 / 1484.07 ±31.34 / 1529.32 ms │ 1.13x slower │
│ QQuery 19 │ 27.56 / 33.49 ±7.05 / 44.79 ms │ 29.37 / 30.74 ±2.60 / 35.94 ms │ +1.09x faster │
│ QQuery 20 │ 521.90 / 529.20 ±6.93 / 540.29 ms │ 534.02 / 543.25 ±6.02 / 552.05 ms │ no change │
│ QQuery 21 │ 511.40 / 521.81 ±8.16 / 531.25 ms │ 526.49 / 546.04 ±13.47 / 566.02 ms │ no change │
│ QQuery 22 │ 976.91 / 987.99 ±7.31 / 999.47 ms │ 1002.98 / 1033.99 ±16.57 / 1050.38 ms │ no change │
│ QQuery 23 │ 3060.41 / 3110.77 ±38.13 / 3174.37 ms │ 3100.91 / 3148.65 ±57.51 / 3249.81 ms │ no change │
│ QQuery 24 │ 40.77 / 43.76 ±2.94 / 48.31 ms │ 40.97 / 45.16 ±6.93 / 58.97 ms │ no change │
│ QQuery 25 │ 110.72 / 118.19 ±12.09 / 142.23 ms │ 111.52 / 114.90 ±4.05 / 122.79 ms │ no change │
│ QQuery 26 │ 41.64 / 44.89 ±5.10 / 55.02 ms │ 41.34 / 42.96 ±1.82 / 46.16 ms │ no change │
│ QQuery 27 │ 509.89 / 518.85 ±5.86 / 525.31 ms │ 516.75 / 519.42 ±2.53 / 523.35 ms │ no change │
│ QQuery 28 │ 2892.23 / 2925.11 ±23.06 / 2954.83 ms │ 2938.98 / 2958.61 ±17.49 / 2983.96 ms │ no change │
│ QQuery 29 │ 41.17 / 47.03 ±9.96 / 66.86 ms │ 41.82 / 52.57 ±18.38 / 89.19 ms │ 1.12x slower │
│ QQuery 30 │ 302.27 / 315.45 ±12.34 / 338.89 ms │ 310.06 / 316.09 ±4.03 / 321.41 ms │ no change │
│ QQuery 31 │ 280.62 / 292.02 ±11.25 / 312.60 ms │ 288.45 / 294.51 ±4.73 / 302.39 ms │ no change │
│ QQuery 32 │ 953.09 / 970.99 ±13.56 / 990.06 ms │ 938.39 / 959.40 ±12.51 / 973.17 ms │ no change │
│ QQuery 33 │ 1444.36 / 1514.86 ±51.66 / 1603.85 ms │ 1486.30 / 1514.76 ±20.73 / 1537.84 ms │ no change │
│ QQuery 34 │ 1479.28 / 1507.12 ±24.83 / 1548.69 ms │ 1484.50 / 1535.22 ±57.13 / 1640.00 ms │ no change │
│ QQuery 35 │ 277.21 / 298.09 ±26.63 / 345.02 ms │ 293.77 / 315.43 ±20.58 / 352.93 ms │ 1.06x slower │
│ QQuery 36 │ 66.65 / 69.57 ±2.18 / 72.93 ms │ 69.86 / 71.79 ±1.23 / 73.71 ms │ no change │
│ QQuery 37 │ 35.36 / 41.59 ±5.53 / 50.29 ms │ 35.97 / 43.72 ±10.50 / 63.99 ms │ 1.05x slower │
│ QQuery 38 │ 40.29 / 43.11 ±1.44 / 44.27 ms │ 44.09 / 49.13 ±4.27 / 54.76 ms │ 1.14x slower │
│ QQuery 39 │ 133.89 / 149.55 ±10.42 / 163.32 ms │ 148.40 / 162.85 ±7.29 / 168.26 ms │ 1.09x slower │
│ QQuery 40 │ 13.90 / 16.84 ±3.95 / 24.57 ms │ 14.83 / 16.55 ±2.12 / 20.52 ms │ no change │
│ QQuery 41 │ 13.74 / 13.91 ±0.15 / 14.11 ms │ 14.65 / 17.16 ±3.11 / 22.74 ms │ 1.23x slower │
│ QQuery 42 │ 13.38 / 13.62 ±0.15 / 13.79 ms │ 14.25 / 14.44 ±0.18 / 14.69 ms │ 1.06x slower │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19548.19ms │
│ Total Time (claude_datafusion-issue-25116-75677f) │ 20219.87ms │
│ Average Time (HEAD) │ 454.61ms │
│ Average Time (claude_datafusion-issue-25116-75677f) │ 470.23ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 15 │
│ Queries with No Change │ 25 │
│ Queries with Failure │ 0 │
└─────────────────────────────────────────────────────┴────────────┘
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned — base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 11.5 GiB |
| Avg memory | 4.4 GiB |
| CPU user | 1002.1s |
| CPU sys | 69.4s |
| Peak spill | 0 B |
**clickbench_partitioned — branch**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 10.3 GiB |
| Avg memory | 4.0 GiB |
| CPU user | 1034.1s |
| CPU sys | 75.9s |
| Peak spill | 0 B |
</details>
---
[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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611481841
run benchmark clickbench_extended clickbench_partitioned external_aggr tpch
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5611482086
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5613943491
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5605293406
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/25123#issuecomment-5605256201)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5605256201-2273-rkhdp 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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 claude/datafusion-issue-25116-75677f (3238cd640c6028263dfe8689deae10080e7be8b9) to 4048898 (merge-base) [diff](https://github.com/apache/datafusion/compare/40488988ad596c9b093ad60e1453430d803ce33c..3238cd640c6028263dfe8689deae10080e7be8b9)
<details><summary>Run configuration</summary>
```yaml
run benchmark external_aggr
```
</details>
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: make GroupsAccumulatorAdapter cost proportional to the batch, not the group count [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5607825245
run benchmark clickbench_extended
baseline:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
changed:
ref: 40488988ad596c9b093ad60e1453430d803ce33c
--
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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5619491260
## Benchmark summary
All numbers are from the bench bot on a 12 vCPU `c4a-highmem-16`, against
baseline `40488988ad`. A ratio above 1.00 means this PR is faster.
Because a single run of any of these suites moves by several percent on its own,
every claim below is backed by repeat runs, and by **control runs that compare
`40488988ad` against itself**. The control tells you what a ratio of "no change"
actually looks like on this machine, which turned out to matter twice.
### `clickbench_extended` — where the adapter is exercised
| Query | control (3 runs) | this PR (6 runs) |
| --- | --- | --- |
| Q14 `COUNT(DISTINCT <string>)` by `SearchPhrase` | 0.97, 0.99, 1.02 | **1.16, 1.14, 1.14, 1.12, 1.09, 1.22** |
| Q5 `APPROX_PERCENTILE_CONT` by `ClientIP, WatchID` | 0.98, 1.00, 0.98 | **1.04, 1.06, 1.08, 1.08, 1.04, 1.03** |
| Q2 4x `COUNT(DISTINCT)` by `BrowserCountry` | 1.00, 1.00, 1.01 | 0.99, 0.99, 0.99, 0.97, 0.99, 0.99 |
Q14 and Q5 are the two queries in the suite whose aggregate has no
`GroupsAccumulator` and whose `GROUP BY` has many values. Both are faster in all
six runs, by roughly 14% and 5%. Suite totals: 1.032, 1.036, 1.026.
Q2 is the query that this change could hurt, because it groups by a column with
110 different values and has four `COUNT(DISTINCT)` aggregates, thus four
adapters that each route every batch. It stays level with its control.
### Every other suite: no change
| Suite | Runs | Result |
| --- | --- | --- |
| `clickbench_partitioned` | 3 + 2 controls | 0.969, 0.989, 0.999 against controls of 0.984 and 0.992 |
| `tpcds_sf1` | 1 | 0.994 total, 1 faster / 4 slower / 94 no change |
| `tpch_sf1` | 1 | 0.998 total, 0 faster / 0 slower / 22 no change |
| `external_aggr` | 1 | 0.992 total, 0 faster / 0 slower / 8 no change |
`clickbench_partitioned` is worth a word. Its first run came back at 0.969 with
13 queries marked slower, which looks like a regression. It is not one. Two more
runs gave 0.989 and 0.999, and the two control runs, which compare `40488988ad`
against itself, gave 0.984 and 0.992 with 5 and 6 queries marked slower. That
suite moves by that much on its own. No query in it is outside its control band
in all three runs.
Every aggregate in `tpch` has a `GroupsAccumulator`, so it never reaches the code
this PR changes. It is there as a control, and it is flat.
🤖 Generated with [Claude Code](https://claude.com/claude-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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767675986
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] perf: GroupsAccumulatorAdapter walks only the groups a batch has rows for [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #25123:
URL: https://github.com/apache/datafusion/pull/25123#issuecomment-5767676454
run benchmark 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: github-unsubscribe@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org