vortex-data/vortex

vortex

Performance History

Latest Results

feat: check VortexReadAt::read_at results in the I/O driver (#7783) This ensures misbehaving VortexReadAt implementations cannot trigger later bad behavior by the I/O driver. Instead, the driver raises an error as close as possible to the badly behaving VortexReadAt implementation. --------- Signed-off-by: Daniel King <dan@spiraldb.com>

develop

26 minutes ago

fix Signed-off-by: Daniel King <dan@spiraldb.com>

dk/verify-bytes-from-read-at-implementation

34 minutes ago

fix tests Signed-off-by: Daniel King <dan@spiraldb.com>

dk/verify-bytes-from-read-at-implementation

39 minutes ago

Merge branch 'bp/minmax-index' of github.com:vortex-data/vortex into bp/minmax-index

bp/minmax-index

1 hour ago

more Signed-off-by: Robert Kruszewski <github@robertk.io>

rk/intersect-by-rank

1 hour ago

pick nits Signed-off-by: Daniel King <dan@spiraldb.com>

dk/verify-bytes-from-read-at-implementation

1 hour ago

bench: CUDA host-to-device copy modes (#7815) Compare pageable host memory with `cuMemHostAlloc` pinned allocations using default flags and `WRITECOMBINED`. Benchmark results on a GH200: ``` cuda/load_to_device/memcpy_htod/pageable/1GiB time: [10.717 ms 10.754 ms 10.793 ms] thrpt: [92.649 GiB/s 92.989 GiB/s 93.306 GiB/s] cuda/load_to_device/memcpy_htod/pinned_default/1GiB time: [10.085 ms 10.265 ms 10.527 ms] thrpt: [94.992 GiB/s 97.423 GiB/s 99.159 GiB/s] cuda/load_to_device/memcpy_htod/pinned_write_combined/1GiB time: [21.043 ms 21.127 ms 21.204 ms] thrpt: [47.161 GiB/s 47.333 GiB/s 47.522 GiB/s] cuda/load_to_device/device_alloc_memcpy_htod/pageable/1GiB time: [42.625 ms 42.704 ms 42.781 ms] thrpt: [23.375 GiB/s 23.417 GiB/s 23.460 GiB/s] cuda/load_to_device/device_alloc_memcpy_htod/pinned_default/1GiB time: [41.864 ms 42.186 ms 42.592 ms] thrpt: [23.478 GiB/s 23.704 GiB/s 23.887 GiB/s] change: time: [+1.7580% +2.5859% +3.6570%] (p = 0.00 < 0.05) thrpt: [-3.5280% -2.5207% -1.7276%] cuda/load_to_device/device_alloc_memcpy_htod/pinned_write_combined/1GiB time: [51.986 ms 52.077 ms 52.166 ms] thrpt: [19.170 GiB/s 19.202 GiB/s 19.236 GiB/s] ``` The insights here being, `WRITECOMBINED` yields significantly slower copy performance, whilst pageable host memory is roughly on par with pinned host memory (without using `WRITECOMBINED`). Signed-off-by: Alexander Droste <alexander.droste@protonmail.com>

develop

1 hour ago

docs(benchmarks-website): address review feedback on planning-docs migration Doc-accuracy nits surfaced in review of #7810. No behavior changes. - `vortex-bench/src/datasets/mod.rs::Dataset::v3_dataset_dims`: drop the stale "Public-BI emits dataset = public-bi, dataset_variant = …" example. The only override (`PBIBenchmark::v3_dataset_dims`) actually returns `(&self.name, None)` to match the migrate classifier — the parent-namespace framing was historical. Reword to "Override only when a suite needs a different dataset name on the wire than its `name()` returns." - `query_idx` field doc on both `vortex-bench/src/v3.rs` and `server/src/records.rs`: drop the "1-based" claim. Query indices are whatever the producing bench loop happens to use (ClickBench is 0-based, others vary); both sides agree only because the migrate classifier parses literal digits out of `q07`-style v2 chart names. - `server/src/app.rs`: add `/api/group/{slug}` to the read-API line; the router wires it but the module doc had drifted. - `server/src/api/charts.rs::collect_group_charts`: re-introduce a `TODO(#7812):` marker so the N+1 follow-up is grep-discoverable. Tracking issue: https://github.com/vortex-data/vortex/issues/7812 - `benchmarks-website/README.md`: rewrite the cutover-plan bullet to reflect that v3 runs alongside v2 on the same EC2 host (v2 on `:80`, v3 on `:3001`), per `docker-compose.yml` and `ec2-init.txt`. The earlier "separate EC2 host" claim was stale. - `server/src/ingest.rs`: add the missing 400 row for the older-than- expected `schema_version` case to the HTTP matrix table. - `vortex-bench/src/v3.rs`: drop the `[`canonical_tpc_scale_factor`]` intra-doc link in the new dim-mapping table — the function is private, so `RUSTDOCFLAGS="-D warnings" cargo doc` rejected it. - `server/src/lib.rs`: rename "PNGs" to "logos" in the route table — typos parsed "PN" inside "PNGs" as a typo of "ON" and failed `Spell Check with Typos`. Verification: - `cargo +nightly fmt --all` - `cargo build -p vortex-bench-server -p vortex-bench-migrate` → clean. - `cargo test -p vortex-bench-server -p vortex-bench-migrate` → all 140+ tests pass. - `cargo clippy --all-targets --all-features -p vortex-bench-server -p vortex-bench-migrate -- -W dead_code -W unreachable_pub -W unused_imports` → 0 warnings. - `RUSTDOCFLAGS="-D warnings" cargo doc --no-deps -p vortex-bench -p vortex-bench-server -p vortex-bench-migrate` → clean. - `typos .` → 0 hits. Signed-off-by: Claude <noreply@anthropic.com>

claude/vortex-v3-quality-pass-jbkRO

1 hour ago

Latest Branches

+18%

feat: check VortexReadAt::read_at results in the I/O driver#7783

35 minutes ago

a7c4640

dk/verify-bytes-from-read-at-implementation

-32%

Cache min/max indices in patches#7753

1 hour ago

8bc3d8a

bp/minmax-index

×19

Improve intersect_by_rank performance#7744

1 day ago

a8488fc

rk/intersect-by-rank

Home Terms Privacy Docs