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

stillwater-sc / universal / 42180824817
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 06 Nov 2026 07:41AM UTC
Jobs 1
Files 863
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

06 Nov 2026 07:25AM UTC coverage: 88.603% (+0.01%) from 88.592%
42180824817

push

github

web-flow
feat(numeric): compensated sin, cos, tan, asin, acos, atan and atan2 for faithful (#1960)

Part 3 of #1952, evaluated in faithful arithmetic.

- sin, cos and tan reduce the argument with the shared Payne-Hanek
  reduction of the multi-double types (#1723), so the reduced argument
  keeps the pair's precision for any finite x, up to max(), and next to
  a multiple of pi/2. The parts of a faithful serve as double limbs. A
  small adapter rebuilds the reduced argument, splitting each 53-bit
  chunk exactly into float parts on a float host. Taylor kernels for sin
  and cos on [-pi/4, pi/4] follow.
- atan2 takes one Newton step from Real's atan2, after scaling (x, y) by
  a common power of two. atan, asin and acos go through it, the last two
  with (1 - x)(1 + x).
- The hosts are float, double, and a long double that has double's
  format, as on MSVC and macOS ARM64. A wider long double part does not
  fit the reduction's double limbs, so there faithful<long double> has no
  trigonometric functions rather than wrong ones.
- Domain errors are NaN with EDOM, or throw the new faithful_domain_error
  with FAITHFUL_THROW_ARITHMETIC_EXCEPTION set.

numeric/faithful/trigonometry.cpp measures every function over 20000
arguments per range: [-pi, pi], [-1e4, 1e4], up to max(), next to
k pi/2, near +-1, and the four atan2 quadrants. The reference is dd for
faithful<float> and qd for faithful<double>. The worst error is 3.6 and
3.5 units of epsilon(), against a bound of 5. Mutation checks: sin
through Real is 10^10 to 10^51 units off, and a reduction with a pair
pi/2 is 10^4 units off on [-pi, pi]. The worst-error tracking keeps a
NaN once seen, here and in exponential.cpp.

Resolves #1956
Relates to #1952

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

244 of 269 new or added lines in 5 files covered. (90.71%)

1 existing line in 1 file now uncovered.

58688 of 66237 relevant lines covered (88.6%)

9343185.66 hits per line

Uncovered Changes

Lines Coverage ∆ File
22
85.23
numeric/faithful/trigonometry.cpp
3
97.35
include/sw/universal/number/faithful/math/functions/trigonometry.hpp

Coverage Regressions

Lines Coverage ∆ File
1
93.16
-0.85% include/sw/universal/verification/correctly_rounded_reference.hpp
Jobs
ID Job ID Ran Files Coverage
1 42180824817.1 06 Nov 2026 07:41AM UTC 863
88.6
GitHub Action Run
Source Files on build 42180824817
  • Tree
  • List 863
  • Changed 6
  • Source Changed 3
  • Coverage Changed 5
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42180824817
  • 79d40e9d on github
  • Prev Build on main (#42174431513)
  • Next Build on main (#42193918817)
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