You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "pitrou (via GitHub)" <gi...@apache.org> on 2025/09/24 10:09:42 UTC
[PR] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
pitrou opened a new pull request, #47637:
URL: https://github.com/apache/arrow/pull/47637
Thanks for opening a pull request!
If this is your first pull request you can find detailed information on how to contribute here:
* [New Contributor's Guide](https://arrow.apache.org/docs/dev/developers/guide/step_by_step/pr_lifecycle.html#reviews-and-merge-of-the-pull-request)
* [Contributing Overview](https://arrow.apache.org/docs/dev/developers/overview.html)
Please remove this line and the above text before creating your pull request.
### Rationale for this change
### What changes are included in this PR?
### Are these changes tested?
### Are there any user-facing changes?
**This PR includes breaking changes to public APIs.** (If there are any breaking changes to public APIs, please explain which changes are breaking. If not, you can remove this.)
**This PR contains a "Critical Fix".** (If the changes fix either (a) a security vulnerability, (b) a bug that caused incorrect or invalid data to be produced, or (c) a bug that causes a crash (even when the API contract is upheld), please provide explanation. If not, you can remove this.)
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "voltrondatabot (via GitHub)" <gi...@apache.org>.
voltrondatabot commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3328524025
Benchmark runs are scheduled for commit bf2698a9024d657083ae947c132a5fb1c41e1b3a. Watch https://buildkite.com/apache-arrow and https://conbench.ursa.dev for updates. A comment will be posted here when the runs are complete.
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "conbench-apache-arrow[bot] (via GitHub)" <gi...@apache.org>.
conbench-apache-arrow[bot] commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3330757012
Thanks for your patience. Conbench analyzed the 3 benchmarking runs that have been run so far on PR commit bf2698a9024d657083ae947c132a5fb1c41e1b3a.
There were 33 benchmark results indicating a performance regression:
- Pull Request Run on `arm64-t4g-2xlarge-linux` at [2025-09-24 17:58:41Z](https://conbench.ursa.dev/compare/runs/637047ac713040b08721454986679af8...4539b757e96e4490b0f28b8fbb2b337e/)
- [`CopyEmptyVector` (C++) with params=<STATIC_VECTOR(std::string)>, source=cpp-micro, suite=arrow-small-vector-benchmark](https://conbench.ursa.dev/compare/benchmarks/068d3df7b3c7781a8000432e2c602812...068d431bddd87d7d80000c390e165831)
- [`CopyShortVector` (C++) with params=<STD_VECTOR(int)>, source=cpp-micro, suite=arrow-small-vector-benchmark](https://conbench.ursa.dev/compare/benchmarks/068d3df7b87f790a8000e59b6a690bb2...068d431be2de70138000d5e0caec18d4)
- and 31 more (see the report linked below)
The [full Conbench report](https://github.com/apache/arrow/runs/51177631486) has more 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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3327583765
@ursabot please benchmark
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3329542409
Also, I was surprised that, at least with gcc 14.3.0, build times do not seem to increase significantly with `CMAKE_INTERPROCEDURAL_OPTIMIZATION` enabled. So perhaps gcc is already using some equivalent of clang's "thin LTO".
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3328912558
I'm trying this locally, also using conda-forge and therefore the same toolchain. One interesting aspect is that IPO seems to decrease total library size, perhaps by removing unused private functions and data:
* before:
```console
$ size -G /build/build-release/relwithdebinfo/*.so
text data bss total filename
1231288 436476 1080 1668844 /build/build-release/relwithdebinfo/libarrow_acero.so
11719000 2076782 46232 13842014 /build/build-release/relwithdebinfo/libarrow_compute.so
1029085 613682 3496 1646263 /build/build-release/relwithdebinfo/libarrow_dataset.so
11368776 3895932 2181849 17446557 /build/build-release/relwithdebinfo/libarrow.so
1010359 500136 1992 1512487 /build/build-release/relwithdebinfo/libarrow_testing.so
2675915 1510609 6328 4192852 /build/build-release/relwithdebinfo/libparquet.so
```
* after:
```console
$ size -G /build/build-release/relwithdebinfo/*.so
text data bss total filename
971243 380629 1552 1353424 /build/build-release/relwithdebinfo/libarrow_acero.so
11764472 1924230 46440 13735142 /build/build-release/relwithdebinfo/libarrow_compute.so
911885 500747 4328 1416960 /build/build-release/relwithdebinfo/libarrow_dataset.so
10306362 3473709 2183392 15963463 /build/build-release/relwithdebinfo/libarrow.so
954794 439139 3968 1397901 /build/build-release/relwithdebinfo/libarrow_testing.so
2446829 1344558 4128 3795515 /build/build-release/relwithdebinfo/libparquet.so
```
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3329200093
I ran two sets of benchmark locally and the results are roughly similar to those on the conbench machines:
* Parquet benchmarks: https://gist.github.com/pitrou/b20fae69d72933dd82f066f64af54e99
* Compute benchmarks: https://gist.github.com/pitrou/132a23457713775697253ca8d54de408
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3328812300
Note: a caveat of the conbench benchmarks results is that they all use the conda-forge toolchain, including the same compiler (gcc 14.3.0).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3385673183
Ok, this PR validates that IPO can be useful. I'm going to close now.
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "conbench-apache-arrow[bot] (via GitHub)" <gi...@apache.org>.
conbench-apache-arrow[bot] commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3329851688
Thanks for your patience. Conbench analyzed the 3 benchmarking runs that have been run so far on PR commit 41f441af34b8dbcc9e25c5e2a374bca853e0414e.
There were 31 benchmark results indicating a performance regression:
- Pull Request Run on `amd64-c6a-4xlarge-linux` at [2025-09-24 11:15:34Z](https://conbench.ursa.dev/compare/runs/e2d6041dfc2040a0b1c622bf3e8f6e48...dcd14a87a14647078e3fa6ef5b8f241b/)
- [`MaxElementWiseArrayScalarString` (C++) with params=32768/2, source=cpp-micro, suite=arrow-compute-scalar-compare-benchmark](https://conbench.ursa.dev/compare/benchmarks/068d3b1c805f747c8000fcc8a8cb8a53...068d3d37e871738c80006f09cb61c4fa)
- [`SetBitRunReader` (C++) with params=50, source=cpp-micro, suite=arrow-bit-util-benchmark](https://conbench.ursa.dev/compare/benchmarks/068d3b203f0970888000ab8c00f55234...068d3d3ce6b97fd88000f3d5c325c1e0)
- and 29 more (see the report linked below)
The [full Conbench report](https://github.com/apache/arrow/runs/51156330383) has more 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@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "WillAyd (via GitHub)" <gi...@apache.org>.
WillAyd commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3329519135
I see that CMake also has a "thin" LTO type that may offer improvements over the default LTO
https://clang.llvm.org/docs/ThinLTO.html
https://blog.llvm.org/2016/06/thinlto-scalable-and-incremental-lto.html
Do we know if this enabled by CMake through this option, or would it be an additional option to benchmark?
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3329530705
I have no idea, but the benchmarks here use gcc.
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "voltrondatabot (via GitHub)" <gi...@apache.org>.
voltrondatabot commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3327584216
Benchmark runs are scheduled for commit 41f441af34b8dbcc9e25c5e2a374bca853e0414e. Watch https://buildkite.com/apache-arrow and https://conbench.ursa.dev for updates. A comment will be posted here when the runs are complete.
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3327586538
@AntoinePrv This is trying to run the benchmarks with IPO to see if it makes a significant difference.
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou closed pull request #47637: [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION
URL: https://github.com/apache/arrow/pull/47637
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3328232725
Looks like we miss `ar` and `ranlib` on macOS. https://buildkite.com/apache-arrow/arrow-bci-benchmark-on-test-mac-arm/builds/7320/steps/canvas?sid=01997b83-cbd5-4f86-8d2c-dfe8315e630a#01997b83-cc32-46ad-b680-38fbc1e59ab1/34-1341
--
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] [C++][Benchmarks] Experiment with CMAKE_INTERPROCEDURAL_OPTIMIZATION [arrow]
Posted by "pitrou (via GitHub)" <gi...@apache.org>.
pitrou commented on PR #47637:
URL: https://github.com/apache/arrow/pull/47637#issuecomment-3328523358
@ursabot please benchmark lang=C++
--
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