• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

Qiskit / coheriq / 41079206227
93%

Build:
DEFAULT BRANCH: main
Ran 30 Oct 2026 03:17PM UTC
Jobs 1
Files 5
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

30 Oct 2026 03:10PM UTC coverage: 86.239%. Remained the same
41079206227

push

github

web-flow
Document how to compare an engine against the reference implementation (#27)

* Document how to compare an engine against the reference

Someone trying to check an engine for numerical agreement with the
implementation it replaces runs into the activation contract: the
reference implementation and an engine cannot both be active in one
process, and there is no way to relax that.  The FAQ explained the
restriction from several angles but never answered the question a user
actually arrives with, which is what to do instead.

Add an entry to the activation contract section covering it, giving
three reasons one process per implementation is the right shape rather
than a consolation prize.  First, a same-process comparison would be a
confounded experiment: enabling an engine can change denormal flags,
the FPU control word, thread counts, or which BLAS is loaded, so code
running after activation is the reference implementation as observed in
a process an engine has already altered, and the tighter the bar, the
more that matters.  Second, benchmarking needs separate processes
regardless, since timings taken after an engine has warmed caches or
spun up a thread pool measure the state of the process.  Third, output
recorded as an artifact can be compared against a previous release,
another machine, or a build that cannot be installed locally.

The entry records why the restriction applies uniformly even though the
first reason is weakest for an engine touching no global numerical
state: Coheriq is built for the setting where non-Python code enters
the picture, and a contract holding only for pure-Python engines would
not be worth much in the HPC settings it targets.

Also discourages reaching for pytest-forked downstream.  Coheriq's own
suite needs it to exercise the activation machinery repeatedly, but a
domain or engine suite runs under a single activation and can share one
process; depending on os.fork() would only cost it Windows support.

Assisted-by: Cla... (continued)

188 of 218 relevant lines covered (86.24%)

0.86 hits per line

Jobs
ID Job ID Ran Files Coverage
1 41079206227.1 30 Oct 2026 03:17PM UTC 5
86.24
GitHub Action Run
Source Files on build 41079206227
  • Tree
  • List 5
  • Changed 0
  • Source Changed 0
  • Coverage Changed 0
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #41079206227
  • baa57cdb on github
  • Prev Build on gh-readonly-queue/main/pr-36-78ab1f4a5a8f51b483093db60dd4ef7d35a1a7da (#41071357699)
  • Next Build on main (#41090999527)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc