pitrou commented on PR #47586:
URL: https://github.com/apache/arrow/pull/47586#issuecomment-3397945215
Hmm, the problem is that this PR is leading to a combinatorial explosion of
code generation that has very noticeable effects.
Here is library code size before this PR:
```console
$ size -G /build/build-release/relwithdebinfo/lib*.so
text data bss total filename
1231288 436476 1080 1668844
/build/build-release/relwithdebinfo/libarrow_acero.so
11741720 2078723 46256 13866699
/build/build-release/relwithdebinfo/libarrow_compute.so
1029085 613682 3496 1646263
/build/build-release/relwithdebinfo/libarrow_dataset.so
11380712 3898809 2184321 17463842
/build/build-release/relwithdebinfo/libarrow.so
1010359 500136 1992 1512487
/build/build-release/relwithdebinfo/libarrow_testing.so
2676427 1510937 5536 4192900
/build/build-release/relwithdebinfo/libparquet.so
```
And here it is after this PR:
```console
$ size -G /build/build-release/relwithdebinfo/lib*.so
text data bss total filename
1231288 436476 1080 1668844
/build/build-release/relwithdebinfo/libarrow_acero.so
11741720 2078723 46256 13866699
/build/build-release/relwithdebinfo/libarrow_compute.so
1029085 613682 3496 1646263
/build/build-release/relwithdebinfo/libarrow_dataset.so
15511528 4319214 2183753 22014495
/build/build-release/relwithdebinfo/libarrow.so
1010359 500136 1992 1512487
/build/build-release/relwithdebinfo/libarrow_testing.so
2676427 1510937 5536 4192900
/build/build-release/relwithdebinfo/libparquet.so
```
We cannot afford to add 4MB of code (a 37% increase for `libarrow.so`) just
to make sparse tensors more correct. So we need to find an approach that
generates much less code.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
