You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "Dandandan (via GitHub)" <gi...@apache.org> on 2026/04/16 18:41:21 UTC
[PR] perf(parquet): vectorise dict-index bounds check in RleDecoder::get_batch_with_dict [arrow-rs]
Dandandan opened a new pull request, #9746:
URL: https://github.com/apache/arrow-rs/pull/9746
## Summary
- Replace the short-circuiting `idx_chunk.iter().all(|&i| (i as usize) < dict_len)` in the bit-packed hot loop of `RleDecoder::get_batch_with_dict` with a u32 max-reduction. `.all` blocks autovectorisation; `fold(0u32, |acc, &i| acc.max(i as u32))` has no early exit, so LLVM lowers the check to a single SIMD max-reduction and reuses the loaded registers for the gather that follows.
- Adds `parquet/benches/rle_dict.rs`, a small targeted Criterion bench that drives `get_batch_with_dict` directly (i32 and `String` dictionaries, sizes 16/256/1024, 8192 values per batch).
## Why
On aarch64 the old code compiled to eight serialised `ldrsw` + `cmp` + `b.ls` pairs per 8-index chunk, followed by eight separate scalar gather loads — one lane at a time. After the change the bounds check is one SIMD reduction:
```
ldp q1, q0, [x11], #0x20 ; load 8 indices
umax.4s v2, v1, v0 ; lane-wise max
umaxv.4s s2, v2 ; horizontal max
fmov w13, s2
cmp x20, x13 ; one bounds check
b.ls <panic>
```
and `v1 / v0` are then reused for the gather, avoiding the reloads.
Negative `i32` values cast to `u32` become large, so the check still rejects them.
## Measurements
Apple Silicon (aarch64), `cargo bench --bench rle_dict`:
| case | before | after | Δ |
|---------------------|-------------|-------------|-------|
| str/dict=16 | 59.48 µs | 57.90 µs | −2.6% |
| i32/dict=16 | 3.28 µs | 3.34 µs | noise |
| str/dict=256 | 48.72 µs | 47.96 µs | noise |
| i32/dict=256 | 3.33 µs | 3.21 µs | −3.3% |
| str/dict=1024 | 34.29 µs | 33.01 µs | −4.2% |
| i32/dict=1024 | 3.79 µs | 3.74 µs | −1.7% |
## Test plan
- [x] `cargo test -p parquet --lib -- encodings::rle::`
- [x] `cargo bench -p parquet --bench rle_dict --features experimental` (results above)
- [ ] Verify CI
🤖 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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4266013757
run benchmarks
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "jhorstmann (via GitHub)" <gi...@apache.org>.
jhorstmann commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3098928650
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
Review Comment:
Hm, this sounds familiar, I opened an issue about this missed optimization a while ago (https://github.com/rust-lang/rust/issues/113789). The issue seems to still exist.
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4266025651
🤖 Benchmark running (GKE) | [trigger](https://github.com/apache/arrow-rs/pull/9746#issuecomment-4266013757)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4266013757-1406-v4lfd 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 rle-dict-max-reduction (0ca4c50121d761fbe7ae2115f55ef90aa4696413) to 89b1497 (merge-base) [diff](https://github.com/apache/arrow-rs/compare/89b1497484115eeffe2daaf6be11193aa1054153..0ca4c50121d761fbe7ae2115f55ef90aa4696413) using:
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): vectorise dict-index bounds check in RleDecoder::get_batch_with_dict (-2-6%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4263000224
run benchmark arrow_reader_clickbench
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "jhorstmann (via GitHub)" <gi...@apache.org>.
jhorstmann commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3098982962
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
+ // short-circuits and blocks autovectorisation. Negative
+ // i32 cast to u32 becomes a large value so the bounds
+ // check still rejects it.
+ let max_idx = idx_chunk.iter().fold(0u32, |acc, &i| acc.max(i as u32));
+ if (max_idx as usize) >= dict_len {
+ oob(max_idx, dict_len);
Review Comment:
Now I remember there was a draft pr about improving the error handling (#9365), so this could also be done separately.
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "jhorstmann (via GitHub)" <gi...@apache.org>.
jhorstmann commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3098956779
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
+ // short-circuits and blocks autovectorisation. Negative
+ // i32 cast to u32 becomes a large value so the bounds
+ // check still rejects it.
+ let max_idx = idx_chunk.iter().fold(0u32, |acc, &i| acc.max(i as u32));
+ if (max_idx as usize) >= dict_len {
+ oob(max_idx, dict_len);
Review Comment:
We should probably also return an error here instead of panicking, since the panic can be triggered by input provided by a user.
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3099285744
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
Review Comment:
Oof :)
I compared it now as well to `.fold(true, |a, &i| a & ((i as u32) < dict_len_u32))`, looks like computing the max-then-compare does also generate better code (and benchmarks) than compare-against-max:
```
max-reduce:
ldp q1, q0, [x13, #32]
ldp q3, q2, [x13]
umax.4s v4, v2, v0
umax.4s v3, v3, v1
umax.4s v3, v3, v4
umaxv.4s s3, v3
fmov w8, s3
cmp x21, x8
b.ls LBB2735_48
AND-reduce :
ldp q1, q0, [x8, #32]
ldp q3, q2, [x8]
cmhs.4s v4, v2, v6
cmhs.4s v3, v3, v6
uzp1.8h v3, v3, v4
cmhs.4s v4, v1, v6
cmhs.4s v5, v0, v6
uzp1.8h v4, v4, v5
uzp1.16b v3, v3, v4
umaxv.16b b3, v3
fmov w14, s3
tbnz w14, #0, LBB2735_48
```
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4275543964
@jhorstmann :thank you! Could you ✅ the 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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4276018035
🤖 Arrow criterion benchmark running (GKE) | [trigger](https://github.com/apache/arrow-rs/pull/9746#issuecomment-4276014403)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4276014403-1540-n8mhx 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 rle-dict-max-reduction (963f2a267e0ad1fa40c5dc4511525df7ae176caf) to 89b1497 (merge-base) [diff](https://github.com/apache/arrow-rs/compare/89b1497484115eeffe2daaf6be11193aa1054153..963f2a267e0ad1fa40c5dc4511525df7ae176caf)
BENCH_NAME=arrow_reader
BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental,object_store --bench arrow_reader
BENCH_FILTER=
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4276014403
run benchmark arrow_reader
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "jhorstmann (via GitHub)" <gi...@apache.org>.
jhorstmann commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4273913607
Thanks, looks good to me!
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3105262176
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
+ // short-circuits and blocks autovectorisation. Negative
+ // i32 cast to u32 becomes a large value so the bounds
+ // check still rejects it.
+ let max_idx = idx_chunk.iter().fold(0u32, |acc, &i| acc.max(i as u32));
+ if (max_idx as usize) >= dict_len {
+ oob(max_idx, dict_len);
Review Comment:
addressed
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): vectorise dict-index bounds check in RleDecoder::get_batch_with_dict (-2-6%) [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4263174943
🤖 Arrow criterion benchmark completed (GKE) | [trigger](https://github.com/apache/arrow-rs/pull/9746#issuecomment-4263000224)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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>
```
group main rle-dict-max-reduction
----- ---- ----------------------
arrow_reader_clickbench/async/Q1 1.00 1097.9±4.61µs ? ?/sec 1.00 1101.9±5.50µs ? ?/sec
arrow_reader_clickbench/async/Q10 1.01 6.7±0.16ms ? ?/sec 1.00 6.6±0.19ms ? ?/sec
arrow_reader_clickbench/async/Q11 1.00 7.6±0.26ms ? ?/sec 1.01 7.7±0.24ms ? ?/sec
arrow_reader_clickbench/async/Q12 1.00 14.4±0.29ms ? ?/sec 1.02 14.8±0.31ms ? ?/sec
arrow_reader_clickbench/async/Q13 1.00 17.0±0.32ms ? ?/sec 1.02 17.3±0.34ms ? ?/sec
arrow_reader_clickbench/async/Q14 1.00 16.1±0.25ms ? ?/sec 1.02 16.5±0.33ms ? ?/sec
arrow_reader_clickbench/async/Q19 1.00 3.0±0.06ms ? ?/sec 1.00 3.0±0.05ms ? ?/sec
arrow_reader_clickbench/async/Q20 1.00 72.9±0.45ms ? ?/sec 1.22 89.2±1.05ms ? ?/sec
arrow_reader_clickbench/async/Q21 1.00 81.2±0.63ms ? ?/sec 1.20 97.7±10.88ms ? ?/sec
arrow_reader_clickbench/async/Q22 1.08 128.7±6.49ms ? ?/sec 1.00 119.4±4.59ms ? ?/sec
arrow_reader_clickbench/async/Q23 1.00 245.4±5.69ms ? ?/sec 1.00 245.0±4.13ms ? ?/sec
arrow_reader_clickbench/async/Q24 1.00 19.5±0.38ms ? ?/sec 1.01 19.7±0.45ms ? ?/sec
arrow_reader_clickbench/async/Q27 1.02 58.1±0.73ms ? ?/sec 1.00 57.0±0.60ms ? ?/sec
arrow_reader_clickbench/async/Q28 1.00 58.0±0.60ms ? ?/sec 1.00 58.2±0.64ms ? ?/sec
arrow_reader_clickbench/async/Q30 1.00 18.4±0.15ms ? ?/sec 1.01 18.6±0.22ms ? ?/sec
arrow_reader_clickbench/async/Q36 1.05 15.6±0.67ms ? ?/sec 1.00 15.0±0.40ms ? ?/sec
arrow_reader_clickbench/async/Q37 1.02 5.5±0.09ms ? ?/sec 1.00 5.4±0.08ms ? ?/sec
arrow_reader_clickbench/async/Q38 1.04 13.9±0.39ms ? ?/sec 1.00 13.4±0.30ms ? ?/sec
arrow_reader_clickbench/async/Q39 1.06 25.3±0.61ms ? ?/sec 1.00 23.9±0.33ms ? ?/sec
arrow_reader_clickbench/async/Q40 1.00 5.6±0.10ms ? ?/sec 1.04 5.9±0.04ms ? ?/sec
arrow_reader_clickbench/async/Q41 1.00 5.0±0.07ms ? ?/sec 1.02 5.1±0.08ms ? ?/sec
arrow_reader_clickbench/async/Q42 1.00 3.5±0.05ms ? ?/sec 1.00 3.5±0.03ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q1 1.00 1065.0±6.85µs ? ?/sec 1.01 1070.4±4.59µs ? ?/sec
arrow_reader_clickbench/async_object_store/Q10 1.00 6.4±0.22ms ? ?/sec 1.01 6.5±0.22ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q11 1.00 7.4±0.24ms ? ?/sec 1.00 7.4±0.24ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q12 1.00 14.3±0.28ms ? ?/sec 1.04 14.8±0.27ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q13 1.01 17.3±0.09ms ? ?/sec 1.00 17.1±0.40ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q14 1.00 15.8±0.36ms ? ?/sec 1.02 16.1±0.33ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q19 1.00 2.9±0.03ms ? ?/sec 1.00 2.9±0.04ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q20 1.01 71.8±0.77ms ? ?/sec 1.00 71.3±0.51ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q21 1.01 80.6±0.63ms ? ?/sec 1.00 80.1±0.58ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q22 1.00 98.1±0.88ms ? ?/sec 1.00 98.4±1.01ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q23 1.10 232.6±5.63ms ? ?/sec 1.00 211.5±1.80ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q24 1.00 19.3±0.40ms ? ?/sec 1.01 19.4±0.42ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q27 1.01 57.4±0.75ms ? ?/sec 1.00 56.7±0.52ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q28 1.00 57.4±0.66ms ? ?/sec 1.01 57.7±0.64ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q30 1.00 18.1±0.16ms ? ?/sec 1.01 18.3±0.28ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q36 1.02 15.0±0.63ms ? ?/sec 1.00 14.7±0.52ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q37 1.01 5.3±0.08ms ? ?/sec 1.00 5.3±0.08ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q38 1.04 13.5±0.54ms ? ?/sec 1.00 13.0±0.40ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q39 1.02 24.2±0.61ms ? ?/sec 1.00 23.7±0.44ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q40 1.00 5.5±0.11ms ? ?/sec 1.03 5.7±0.04ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q41 1.00 4.7±0.07ms ? ?/sec 1.02 4.8±0.08ms ? ?/sec
arrow_reader_clickbench/async_object_store/Q42 1.00 3.4±0.05ms ? ?/sec 1.00 3.4±0.03ms ? ?/sec
arrow_reader_clickbench/sync/Q1 1.00 876.4±29.39µs ? ?/sec 1.01 883.5±4.41µs ? ?/sec
arrow_reader_clickbench/sync/Q10 1.00 5.1±0.08ms ? ?/sec 1.00 5.1±0.06ms ? ?/sec
arrow_reader_clickbench/sync/Q11 1.00 6.0±0.09ms ? ?/sec 1.01 6.0±0.08ms ? ?/sec
arrow_reader_clickbench/sync/Q12 1.00 21.8±0.45ms ? ?/sec 1.02 22.3±0.23ms ? ?/sec
arrow_reader_clickbench/sync/Q13 1.06 30.0±0.80ms ? ?/sec 1.00 28.4±0.77ms ? ?/sec
arrow_reader_clickbench/sync/Q14 1.00 22.9±0.19ms ? ?/sec 1.03 23.6±0.37ms ? ?/sec
arrow_reader_clickbench/sync/Q19 1.00 2.7±0.06ms ? ?/sec 1.00 2.7±0.04ms ? ?/sec
arrow_reader_clickbench/sync/Q20 1.00 123.5±0.99ms ? ?/sec 1.00 123.0±0.98ms ? ?/sec
arrow_reader_clickbench/sync/Q21 1.01 93.0±0.65ms ? ?/sec 1.00 92.6±0.69ms ? ?/sec
arrow_reader_clickbench/sync/Q22 1.00 136.7±1.00ms ? ?/sec 1.06 145.5±3.40ms ? ?/sec
arrow_reader_clickbench/sync/Q23 1.00 278.1±14.59ms ? ?/sec 1.01 279.8±15.08ms ? ?/sec
arrow_reader_clickbench/sync/Q24 1.00 27.1±0.43ms ? ?/sec 1.01 27.4±0.54ms ? ?/sec
arrow_reader_clickbench/sync/Q27 1.03 109.5±0.89ms ? ?/sec 1.00 106.3±0.79ms ? ?/sec
arrow_reader_clickbench/sync/Q28 1.01 105.8±0.71ms ? ?/sec 1.00 104.9±0.79ms ? ?/sec
arrow_reader_clickbench/sync/Q30 1.00 18.6±0.22ms ? ?/sec 1.01 18.8±0.26ms ? ?/sec
arrow_reader_clickbench/sync/Q36 1.01 22.9±0.07ms ? ?/sec 1.00 22.6±0.06ms ? ?/sec
arrow_reader_clickbench/sync/Q37 1.00 6.9±0.08ms ? ?/sec 1.00 6.9±0.07ms ? ?/sec
arrow_reader_clickbench/sync/Q38 1.01 11.6±0.14ms ? ?/sec 1.00 11.5±0.09ms ? ?/sec
arrow_reader_clickbench/sync/Q39 1.02 21.4±0.37ms ? ?/sec 1.00 20.9±0.32ms ? ?/sec
arrow_reader_clickbench/sync/Q40 1.00 5.1±0.10ms ? ?/sec 1.00 5.1±0.10ms ? ?/sec
arrow_reader_clickbench/sync/Q41 1.00 5.6±0.07ms ? ?/sec 1.01 5.7±0.07ms ? ?/sec
arrow_reader_clickbench/sync/Q42 1.00 4.4±0.06ms ? ?/sec 1.00 4.4±0.06ms ? ?/sec
```
</p>
</details>
<details><summary>Resource Usage</summary>
**base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 780.1s |
| Peak memory | 3.1 GiB |
| Avg memory | 2.9 GiB |
| CPU user | 707.7s |
| CPU sys | 72.5s |
| Peak spill | 0 B |
**branch**
| Metric | Value |
|--------|-------|
| Wall time | 782.5s |
| Peak memory | 3.3 GiB |
| Avg memory | 3.1 GiB |
| CPU user | 710.7s |
| CPU sys | 72.0s |
| 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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3106908692
##########
parquet/src/encodings/rle.rs:
##########
@@ -532,9 +555,14 @@ impl RleDecoder {
for (b, i) in out_chunks
.into_remainder()
.iter_mut()
- .zip(idx.chunks_exact(8).remainder().iter())
+ .zip(idx.chunks_exact(CHUNK).remainder().iter())
{
- b.clone_from(&dict[*i as usize]);
+ let dict_idx = *i as usize;
Review Comment:
it is fascinaing that the rust compiler doens't already do this
##########
parquet/src/encodings/rle.rs:
##########
@@ -484,7 +484,16 @@ impl RleDecoder {
if self.rle_left > 0 {
let num_values = cmp::min(max_values - values_read, self.rle_left as usize);
let dict_idx = self.current_value.unwrap() as usize;
- let dict_value = dict[dict_idx].clone();
+ let dict_value = dict
Review Comment:
This is a nice change to return errors rather than panic on input
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): vectorise dict-index bounds check in RleDecoder::get_batch_with_dict (-2-6%) [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4263012535
🤖 Arrow criterion benchmark running (GKE) | [trigger](https://github.com/apache/arrow-rs/pull/9746#issuecomment-4263000224)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4263000224-1400-f5cb4 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 rle-dict-max-reduction (b429fc1a4d670f1a86e29d4cc694daaca8a4d629) to 89b1497 (merge-base) [diff](https://github.com/apache/arrow-rs/compare/89b1497484115eeffe2daaf6be11193aa1054153..b429fc1a4d670f1a86e29d4cc694daaca8a4d629)
BENCH_NAME=arrow_reader_clickbench
BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental,object_store --bench arrow_reader_clickbench
BENCH_FILTER=
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3099158516
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
+ // short-circuits and blocks autovectorisation. Negative
+ // i32 cast to u32 becomes a large value so the bounds
+ // check still rejects it.
+ let max_idx = idx_chunk.iter().fold(0u32, |acc, &i| acc.max(i as u32));
+ if (max_idx as usize) >= dict_len {
+ oob(max_idx, dict_len);
Review Comment:
Ah yeah we could just as well return an error, I agree :)
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "jhorstmann (via GitHub)" <gi...@apache.org>.
jhorstmann commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3099406303
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is {dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read + num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) < dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`: `.all`
+ // short-circuits and blocks autovectorisation. Negative
+ // i32 cast to u32 becomes a large value so the bounds
+ // check still rejects it.
+ let max_idx = idx_chunk.iter().fold(0u32, |acc, &i| acc.max(i as u32));
+ if (max_idx as usize) >= dict_len {
+ oob(max_idx, dict_len);
Review Comment:
If you do, please also check the rle `dict_value` around line 487 and link #9434, otherwise I can do that afterwards.
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4266025712
Benchmark for [this request](https://github.com/apache/arrow-rs/pull/9746#issuecomment-4266013757) failed.
Last 20 lines of output:
<details><summary>Click to expand</summary>
```
Cloning into '/workspace/datafusion-branch'...
rle-dict-max-reduction
From https://github.com/apache/arrow-rs
* [new ref] refs/pull/9746/head -> rle-dict-max-reduction
* branch main -> FETCH_HEAD
Switched to branch 'rle-dict-max-reduction'
89b1497484115eeffe2daaf6be11193aa1054153
Cloning into '/workspace/datafusion-base'...
HEAD is now at 89b1497 Improve take performance on List arrays (#9643)
rustc 1.94.1 (e408947bf 2026-03-25)
0ca4c50121d761fbe7ae2115f55ef90aa4696413
89b1497484115eeffe2daaf6be11193aa1054153
```
</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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan merged PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): Vectorize dict-index bounds check in RleDecoder::get_batch_with_dict (up to -7.9%) [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#issuecomment-4281445775
🚀
--
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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org