zX41ZdbW

Thanks!

My colleague analyzed their ClickBench submission on May 3rd and concluded that:

- "This is probably the most sophisticated spam attempt in ClickBench so far"

- "Unless someone tells me otherwise, I assume ParticleDB is fake."

Previous attempts were strange, to say the least: https://github.com/ClickHouse/ClickBench/pulls?q=is%3Apr+par...

We are building our compiler and the entire LLVM toolchain with PGO, and it gives a solid 20..30% improvement in build speed*. At the same time, building ClickHouse with PGO does not give any meaningful improvement.

* https://github.com/ClickHouse/ClickHouse/pulls?q=is%3Apr+sta...

Also works fairly well for Hadamard-rotated embeddings, as demonstrated here: https://embeddings.info/

This is quite an old technique. In ClickHouse, it appeared before it was named "shredding": https://github.com/ClickHouse/ClickHouse/issues/54864

I have a similar project (which I created in 2024): https://adsb.exposed/

Among various datasets, it has OSM data (all of it): https://adsb.exposed/?dataset=OSM&zoom=4&lat=52.3756&lng=2.5...

And here are lighthouses: https://adsb.exposed/?dataset=OSM&zoom=3&lat=40.0444&lng=9.9...

lfalloc: https://github.com/ClickHouse/ClickHouse/pull/62826 hualloc: https://github.com/ClickHouse/ClickHouse/pull/31376

ClickHouse has been tested with jemalloc, mimalloc, tcmalloc (both variants), rpmalloc, lfalloc, hualloc, and ended up using jemalloc after a few patches and bug fixes.

Some of the oldest PRs could no longer be found.

For example, when I decided to continue my eight-year-old PR https://github.com/ClickHouse/ClickHouse/pull/104948, I couldn't see it - neither in search nor while navigating through pages.

Direct links still work.

Strange to see it here, the article is quite old.

Since pdqsort, vqsort, and glide sort, the current state-of-the-art are driftsort and ipnsort.

I've integrated them into ClickHouse: https://github.com/ClickHouse/ClickHouse/pull/106650

I'm interested. [email protected]

Maybe share my old article with Haiku developers: https://clickhouse.com/blog/agentic-coding

Could be a small chance, but it has already made many engineers reconsider their AI stance.

I've also built a visualizer into ClickHouse, so we can analyze its binary: https://play.clickhouse.com/binary?user=binary - which code takes more space, what dependencies are too big, or what template instantiations are bloated.

ClickHouse can read and visualize its own binary: https://play.clickhouse.com/binary?user=binary

It is useful to understand which dependencies, libraries, and template instantiations contribute to the size.

Let's add it to ClickBench: https://github.com/ClickHouse/ClickBench/

If you are interested in the AIS (ship telemetry) data, my team has created a visualization project: https://adsb.exposed/?dataset=Ships&zoom=5&lat=52.3890&lng=4...

This is one of the optimizations from ClickHouse - detect ASCII and go a fast path: https://github.com/ClickHouse/ClickHouse/blob/6cde32de1a2463...

All databases are hosted in VMs. Most of them only support amd64, so the host machine is also amd64 - this is the only limitation. It uses hardware virtualization, so nothing is emulated.

Thanks! There is a write-up: https://clickhouse.com/blog/clickbench-playground A lot of interesting things around hosting it.

Thank you for Ninja!

The speed (reading the build.ninja file) was never a concern for us. If I could share a wish-list, it will be:

- Fix the possibility of a segmentation fault when the build file is damaged;

- Use a better order of execution: https://github.com/ninja-build/ninja/issues/2157