You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "Reranko05 (via GitHub)" <gi...@apache.org> on 2026/04/24 12:23:26 UTC
Re: [I] [C++][Gandiva] Constructing LLVM module with only necessary functions for better performance [arrow]
Reranko05 commented on issue #40024:
URL: https://github.com/apache/arrow/issues/40024#issuecomment-4313106999
Hi @dmitry-chirkov-dremio, following your suggestion to look into GH-40031, I took a deeper look at this issue and the old PR #40031 (which seems to have already laid substantial groundwork for this optimization).
From what I can tell, the eager `AddGlobalMappings` path and unconditional `LoadPreCompiledIR` path still appear present on current main, and I reproduced `TimedTestExprCompilation` in `gandiva-micro-benchmarks` at roughly 43–52 ms CPU (~49 ms average) locally.
One thought I had is whether a reduced-scope approach might make sense to revisit this incrementally, for example starting only with selective C-function mappings first (leaving IR bitcode loading unchanged initially), and considering optional IR loading separately later.
Does that seem like a reasonable direction, or is there context around this issue/PR I may be missing?
--
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