Jvq on SoftDevStudio.com

jvq Counts Instructions to Catch Small Slowdowns

Mon, 19 Oct 2026 00:00:00 +0000

jvq 0.6.0 adds a second way to measure. Instead of timing the benchmark, the new instructions adapter runs it under valgrind’s cachegrind and counts what it executes. A count hardly moves between runs and ignores how busy the machine is, so the budget drops from 5% to 1%.

On a shared 2-core VM, a real 2% slowdown passed the time gate as noise and came out at +1.9% counted, caught in 5 rounds. In the recorded scenarios, every 2% step was caught, every 0.5% step passed, and a no-change run never failed, beside a noisy neighbour too.

jvq Gates Benchmarks in Any Language

Mon, 19 Oct 2026 00:00:00 +0000

jvq, the studio’s CI gate that fails a pull request when it makes benchmarks slower than their budget, read only Go’s benchmark output in its first release. Version 0.2.0 opens it to any language. A time adapter times the whole command, so a Python script or a compiled binary can be gated with no change to its code, and a json adapter reads a line per result that any benchmark can print. A new jvq compare judges two saved outputs with the same statistics. The release has binaries for Linux, macOS and Windows.

jvq Gates GitLab Merge Requests

Mon, 19 Oct 2026 00:00:00 +0000

jvq 0.4.0 brings the studio’s CI benchmark gate to GitLab. A ready job ships with the release: include it, extend it, name the benchmark command, and merge requests get a gate that measures base and head in turns and fails the pipeline on a regression.

The work was in the details GitLab adds. A merged results pipeline, a merge train on a project that rebases and a plain merge request pipeline each want a different base commit, and jvq now reads GitLab’s own variables to pick the right one. A new JUnit report puts the verdict in the merge request’s test summary, and Jenkins reads the same file.

jvq Keeps a History of Every Run

Mon, 19 Oct 2026 00:00:00 +0000

jvq 0.5.0 goes after the one slowdown a per-change gate can’t see: many small ones, each inside the budget. Runs on the default branch now add a line to a history file, and jvq history shows how far each benchmark has moved in all, with an interval, run by run.

Each run’s change was measured on one machine, so the changes can be chained even when CI runners differ. Each run’s interval is chained along with it, so noise widens the answer instead of passing for a trend. In the GitHub Action it’s one input, history, kept in the Actions cache; the GitLab template has the same.

jvq Launches, a CI Gate That Fails Pull Requests Which Make Benchmarks Slower

Mon, 19 Oct 2026 00:00:00 +0000

jvq, short for just very quickly, is the studio’s nineteenth project: a performance budget gate for CI. Most teams that run benchmarks in CI compare today’s numbers with last week’s, taken on another machine, and shared runners can drift 5 to 20% between runs. The alarms stop meaning anything. jvq measures the base commit and the change in the same job, taking turns, so both sides see the same machine. Version 0.1.0 came out on 18 October for Go benchmarks, as one binary with no dependencies, with binaries for Linux, macOS and Windows and a GitHub Action that checks the download’s checksum. It’s open source under Apache 2.0.

jvq Reads Four Benchmark Harnesses and Gates File Sizes

Mon, 19 Oct 2026 00:00:00 +0000

jvq 0.3.0 takes the output of four popular benchmark harnesses as it is: Rust’s criterion from cargo bench, Google Benchmark’s JSON, pytest-benchmark and vitest bench. A project with one of them now needs only a line or three in jvq.toml to get a CI gate that runs base and head in turns. The release has binaries for Linux, macOS and Windows.

The adapters were tested on output captured from the real tools, and the recordings run each harness end to end through jvq. On a two-core cloud VM, criterion and vitest caught all 10 runs with 20% more work and raised no false alarm in 10 without. pytest-benchmark caught all 10 too, but raised one false alarm when given only a short time per test, and the write-up says so.