You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Recently I checked Profile-Guided Optimization (PGO) improvements on many projects. All current results are available here. E.g. ClickHouse PGO results can be checked here, ClickHouse documentation about building with PGO - https://clickhouse.com/docs/en/operations/optimizing-performance/profile-guided-optimization . According to the multiple tests, PGO can help with improving performance in many cases (including databases). That's why I think trying to optimize chdb with PGO can be a good idea.
I can suggest the following action points:
Perform PGO benchmarks on chdb. And if it shows improvements - add a note about possible improvements in chdb performance with PGO.
Providing an easier way (e.g. a build option) to build scripts with PGO can be helpful for the end-users and maintainers since they will be able to optimize chdb according to their own workloads.
Optimize pre-built binaries
Since the chdb native part is the library, I think the Pydantic-core experience can be reused here. Also, Clang supports PGO for shared libraries.
Maybe testing Post-Link Optimization techniques (like LLVM BOLT) would be interesting too (Clang and Rustc already use BOLT as an addition to PGO) but I recommend starting from the usual PGO.
Here are some examples of how PGO optimization is integrated in other projects:
Description
Recently I checked Profile-Guided Optimization (PGO) improvements on many projects. All current results are available here. E.g. ClickHouse PGO results can be checked here, ClickHouse documentation about building with PGO - https://clickhouse.com/docs/en/operations/optimizing-performance/profile-guided-optimization . According to the multiple tests, PGO can help with improving performance in many cases (including databases). That's why I think trying to optimize chdb with PGO can be a good idea.
I can suggest the following action points:
Since the chdb native part is the library, I think the Pydantic-core experience can be reused here. Also, Clang supports PGO for shared libraries.
Maybe testing Post-Link Optimization techniques (like LLVM BOLT) would be interesting too (Clang and Rustc already use BOLT as an addition to PGO) but I recommend starting from the usual PGO.
Here are some examples of how PGO optimization is integrated in other projects:
configure
scriptThe text was updated successfully, but these errors were encountered: