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

stillwater-sc / universal / 42383889947
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 07 Nov 2026 03:15PM UTC
Jobs 1
Files 868
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

07 Nov 2026 02:55PM UTC coverage: 88.691% (-0.008%) from 88.699%
42383889947

push

github

web-flow
fix(areal): numeric_limits from the encoding, a callable abs, fma ubits, and the ubit's promise (#1983)

Part 2 of #1970, from QA sweep #1962.

- numeric_limits<areal> follows the encoding. It was placeholders:
  min() was minpos, denorm_min() 1, epsilon() 0 with the ubit set, and
  digits counted the ubit; the exponents were off, and has_denorm and
  is_bounded denied the subnormals and the finite range. Now:
  - min() is the smallest normal value and denorm_min() minpos;
  - epsilon() is 2^-fbits, and round_error() 1 for truncation;
  - digits is fbits + 1, the significand of an exact value;
  - min_exponent and max_exponent follow the IEEE convention;
  - has_denorm is denorm_present and is_bounded is true.
  The powers of two are built bit by bit, so the limits stay constant
  expressions.
- abs(areal) returned areal<nbits, es> through a constructor areal does
  not have, and disagreed with its declaration, so no call compiled. It
  clears the sign bit, keeps the value bits and the ubit, and returns a
  NaN as is.
- fma took each input at its represented value and ignored its ubit:
  fma(areal<16,5>(0.1f), 1, 0) was exact while 0.1 * 1 + 0 was not. An
  inexact input now makes the result inexact.
- What the ubit promises is documented in areal_impl.hpp and
  docs/number-systems/areal.md: a single areal's ubit is a flag. For a
  conversion or an operation on exact operands, a set ubit means the
  exact result lies between the encoding and the next. An inexact
  operand makes the result inexact, but its interval need not enclose
  the exact interval result: (0, minpos) - minpos is flagged, yet
  (-minpos, 0) is not the interval it names. tile_interval is the
  enclosure.

Consumers of the old conventions:
- tools/ucalc/tiles.hpp took minpos from numeric_limits::min(). It now
  uses denorm_min() for areal, so ucalc's range output is unchanged.
- static/range/areal/math/fma.cpp expected an exact result from inexact
  inputs.
- static/utility/test_deci... (continued)

24 of 25 new or added lines in 3 files covered. (96.0%)

6 existing lines in 3 files now uncovered.

59382 of 66954 relevant lines covered (88.69%)

9720664.9 hits per line

Uncovered Changes

Lines Coverage ∆ File
1
96.0
-4.0% include/sw/universal/number/areal/numeric_limits.hpp

Coverage Regressions

Lines Coverage ∆ File
3
91.45
-2.56% include/sw/universal/verification/correctly_rounded_reference.hpp
2
34.7
-0.63% include/sw/universal/verification/test_suite_randoms.hpp
1
97.8
-0.06% include/sw/universal/number/cfloat/cfloat_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 42383889947.1 07 Nov 2026 03:15PM UTC 868
88.69
GitHub Action Run
Source Files on build 42383889947
  • Tree
  • List 868
  • Changed 6
  • Source Changed 3
  • Coverage Changed 6
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42383889947
  • 84399210 on github
  • Prev Build on main (#42379094381)
  • Next Build on main (#42406034972)
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