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

stillwater-sc / universal / 41381162763
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 02 Nov 2026 03:31AM UTC
Jobs 1
Files 773
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

02 Nov 2026 03:05AM UTC coverage: 88.467%. Remained the same
41381162763

push

github

web-flow
fix(cfloat): IEEE-754 infinity encoding when there are no max-exponent values (#1895)

cfloat encoded +-inf as s.11..11.11..10 in every configuration. Without
max-exponent values the all-ones exponent binade holds no values, so
that choice was free, and it made the IEEE-754 aliases (half, single,
duble, quad, bfloat_t, tf32, and the default cfloat<n,es>) disagree
with IEEE-754: 0x7c00 read as NaN, half(inf) was 0x7ffe.

One rule now covers all eight configurations: inf is the encoding
right after maxpos, and every NaN sits above it.
- without hasMaxExpValues: IEEE-754, inf = s.11..11.00..00, every other
  all-ones-exponent pattern a NaN
- with hasMaxExpValues: unchanged, inf = s.11..11.11..10, NaN = s.11..11.11..11
- with hasMaxExpValues and isSaturating: unchanged, no inf, s.11..11.11..10
  is maxpos (OCP E4M3FN)
Quiet/signalling NaN typing by the sign is unchanged (USER DECISION:
step 1 of the #1780 assessment only).

setinf()/isinf() use a per-limb IEEE inf pattern when !hasMaxExpValues,
and INF_ENCODING is 0 there, so the rounding carry past maxpos lands on
inf. The integer conversions caught overflow with isnan() alone, which
missed the carry once it became an inf: saturating cfloat<8,4> took
int 256 to inf. They now check isnan() || isinf().

New test static/float/cfloat/standard/ieee754_encoding.cpp: half vs
IEEE binary16 on all 2^16 patterns, single/duble native inf patterns,
cfloat<16,8,uint8_t> vs bfloat16 on all 2^16 patterns, quad, and the
unchanged max-exponent layouts. Against main it fails 26 checks.

Resolves #1780

BEHAVIOUR-CHANGE: cfloat configurations without max-exponent values (the default, half, single, duble, quad, bfloat_t, tf32, ...) encode +-inf as IEEE-754 does (exponent all ones, fraction 0); the old inf pattern s.11..11.11..10 is now a NaN


Docs state that NaN classification matches IEEE-754 but NaN type and
payload do not; the test uses std::bit_cast.

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

15 of 15 new or added lines in 1 file covered. (100.0%)

4 existing lines in 2 files now uncovered.

56082 of 63393 relevant lines covered (88.47%)

9012703.64 hits per line

Coverage Regressions

Lines Coverage ∆ File
2
97.14
0.03% include/sw/universal/number/cfloat/cfloat_impl.hpp
2
83.86
-0.3% include/sw/universal/number/qd/qd_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 41381162763.1 02 Nov 2026 03:31AM UTC 773
88.47
GitHub Action Run
Source Files on build 41381162763
  • Tree
  • List 773
  • Changed 2
  • Source Changed 1
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #41381162763
  • f9d25c50 on github
  • Prev Build on main (#41357904777)
  • Next Build on main (#41400317777)
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