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

stillwater-sc / universal / 42637949454
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 09 Nov 2026 03:15PM UTC
Jobs 1
Files 874
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

09 Nov 2026 02:46PM UTC coverage: 88.975% (+0.05%) from 88.926%
42637949454

push

github

web-flow
fix(math): compute complex abs, arg and the transcendentals in T for high-precision types (#2030)

The shared complex library computed abs, arg and every transcendental in
double, so complex<ereal<8>> stopped at 53 bits. A type marked
is_high_precision (ereal, efloat, dd, qd) now computes them in its own
arithmetic; every other type keeps the std::complex<double> path, whose abs
now uses std::hypot.

- Kahan's forms for asin/acos/asinh/acosh and tanh, hypot-scaled abs and
  sqrt, log1p near |z| = 1, an overflow-free scaled atanh; C99 branch cuts
  and Annex G special values, with non-finite cases handled explicitly.
- Zero-sign-safe helpers keep dd and qd (whose arithmetic drops the sign of
  zero) on the right side of the cuts; extreme atan2 arguments are scaled,
  since dd's and qd's atan2 square them.
- complex_functions_dd.hpp / _qd.hpp held naive specializations (the dd one
  did not compile); they now only register the trait.
- Measured against mpmath: ereal<8> >= 475 bits (holds 424), qd 212, dd 102
  of 106, efloat<8> 252 of 256. On glibc every function matches
  std::complex<double> on a 10x10 special-value grid, signs included, apart
  from 17 cases Annex G leaves unspecified.
- Tests: static/complex/math/precision.cpp and special_values.cpp (ereal,
  ereal<4, float>, dd, qd, efloat, cfloat double path); ereal and efloat
  complex tests now check full precision. 70 references from
  generate_reference_constants.py --complex.

Resolves #2026

BEHAVIOUR-CHANGE: complex<ereal>, complex<efloat>, complex<dd> and complex<qd> abs, arg and transcendentals now return values at the element type's precision instead of double's; abs on the double path uses std::hypot

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

311 of 317 new or added lines in 3 files covered. (98.11%)

61074 of 68642 relevant lines covered (88.97%)

9936789.77 hits per line

Uncovered Changes

Lines Coverage ∆ File
6
97.64
-2.36% include/sw/universal/math/complex/complex_functions.hpp
Jobs
ID Job ID Ran Files Coverage
1 42637949454.1 09 Nov 2026 03:15PM UTC 874
88.97
GitHub Action Run
Source Files on build 42637949454
  • Tree
  • List 874
  • Changed 8
  • Source Changed 4
  • Coverage Changed 8
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42637949454
  • 6560a549 on github
  • Prev Build on main (#42569178976)
  • Next Build on main (#42662405004)
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