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

stillwater-sc / universal / 42150180977
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 06 Nov 2026 02:44AM UTC
Jobs 1
Files 845
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 02:24AM UTC coverage: 88.611% (-0.004%) from 88.615%
42150180977

push

github

web-flow
fix(numeric): implement faithful's comparisons, abs, ++/--, modifiers and classification (#1948)

Several faithful<Real> functions were stubs that compiled and returned a
wrong answer. == and < were always false, so a == a was false and <= and
>= always true. abs() returned zero. ++ and -- printed "TBD" and left the
value unchanged. reset() did nothing. isinf() and isnan() were always
false, and components() called a fraction() that faithful does not have.

- The comparisons order by the represented value, value + error. Both
  operands are renormalized, then the values compared and, on a tie, the
  errors. Rounding preserves order, so this is exact. A NaN compares
  unordered, as in IEEE-754.
- abs() negates both parts when the represented value is negative. ++
  and -- are compensated additions of one.
- reset() and the new setzero() clear both parts. convert() already
  called setzero() and setnan(), which did not exist. setinf() is new.
- isinf() and isnan() test the value. They use <cmath> for a native type
  and the type's own members for a Universal one, so faithful<cfloat>
  compiles with only the two cores included.
- sign(), isneg(), scale() and iszero() read value + error rounded. A pair
  that is not canonical, such as (1, -2), therefore reports the sign of
  the value it represents.
- An infinity survives arithmetic. twoSum(inf, x) has a NaN roundoff,
  which reached the error, and renormalizing then made the value NaN:
  faithful(inf) + 1 converted to NaN. A non-finite value now drops its
  error.
- manipulators.hpp was another type's copy, did not compile, and was not
  included. It is rewritten with type_tag() and components(), which
  prints the sign, scale and both parts, and the umbrella includes it.

numeric/faithful/logic_and_modifiers.cpp checks each function against
the exact double sum of a faithful<float>, covering random canonical and
non-canonical pairs, ties of the value, signed zeros, NaN and the
infinities.

Resolves #193... (continued)

238 of 270 new or added lines in 3 files covered. (88.15%)

3 existing lines in 2 files now uncovered.

57607 of 65011 relevant lines covered (88.61%)

8876089.34 hits per line

Uncovered Changes

Lines Coverage ∆ File
32
84.98
numeric/faithful/logic_and_modifiers.cpp

Coverage Regressions

Lines Coverage ∆ File
2
34.7
-0.63% include/sw/universal/verification/test_suite_randoms.hpp
1
93.16
-0.85% include/sw/universal/verification/correctly_rounded_reference.hpp
Jobs
ID Job ID Ran Files Coverage
1 42150180977.1 06 Nov 2026 02:44AM UTC 845
88.61
GitHub Action Run
Source Files on build 42150180977
  • Tree
  • List 845
  • Changed 4
  • Source Changed 2
  • Coverage Changed 4
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42150180977
  • 9b9aff92 on github
  • Prev Build on main (#42147205817)
  • Next Build on main (#42155020823)
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