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

stillwater-sc / universal / 41400317777
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 02 Nov 2026 06:55AM 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 06:30AM UTC coverage: 88.478% (+0.01%) from 88.467%
41400317777

push

github

web-flow
fix(cfloat): IEEE-754 quiet-bit NaN typing when there are no max-exponent values (#1897)

Step 2 of the #1780 assessment. Without max-exponent values the whole
all-ones-exponent binade except inf is NaN, and cfloat now tells quiet
from signalling by the fraction MSB, as IEEE-754 does, instead of by
the sign. A negative IEEE quiet NaN (half 0xfe00) was signalling, and
cfloat's own sNaN 0xffff was a quiet NaN to IEEE.

- ieeeNanTyping = !hasMaxExpValues && fbits >= 2. With max-exponent
  values, or a one-bit fraction, there are only two NaN patterns and
  the sign stays the type.
- setnan writes 0.11..11.10..00 (quiet, the native quiet_NaN pattern)
  or 0.11..11.01..00 (signalling); isnan reads the quiet bit.
- inf + -inf delivers a quiet NaN (IEEE-754 7.2), in all configurations.
- The NaN sign used to double as the overflow direction in the
  post-rounding projection. post_process now projects by the sign of
  the encoding and is shared with convert_ieee754.
- cfloat-to-cfloat, cfloat-to-native and the four blocktriple
  normalizations pass the NaN type, not the sign.
- With CFLOAT_THROW_ARITHMETIC_EXCEPTION, quiet NaN operands propagate
  through +, -, * and / (they reached the significand path and came out
  +0); signalling NaNs throw; qNaN / 0 is a qNaN, not divide_by_zero.
- operator++ past 0.11..11 steps to the raw 1.11..11 again.
- cfloat_test_suite's add/sub reference expects qNaN for inf - inf.

Tests: ieee754_encoding.cpp checks NaN type on all 2^16 half patterns,
native qNaN/sNaN patterns for single and duble, canonical quiet NaNs,
invalid operations, the max-exponent and one-bit-fraction layouts,
type through conversions and normalizations, and ++ past NaN.
exceptions.cpp now runs as a regression and checks NaN operands with
exceptions on.

Resolves #1896

BEHAVIOUR-CHANGE: cfloat configurations without max-exponent values (the default, half, single, duble, quad, bfloat_t, tf32, ...) type NaNs by the IEEE-754 quiet bit instead of the sign, ... (continued)

29 of 30 new or added lines in 2 files covered. (96.67%)

2 existing lines in 1 file now uncovered.

56096 of 63401 relevant lines covered (88.48%)

8983080.21 hits per line

Uncovered Changes

Lines Coverage ∆ File
1
97.4
0.27% include/sw/universal/number/cfloat/cfloat_impl.hpp

Coverage Regressions

Lines Coverage ∆ File
2
97.4
0.27% include/sw/universal/number/cfloat/cfloat_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 41400317777.1 02 Nov 2026 06:55AM UTC 773
88.48
GitHub Action Run
Source Files on build 41400317777
  • Tree
  • List 773
  • Changed 3
  • Source Changed 2
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #41400317777
  • afb41be4 on github
  • Prev Build on main (#41381162763)
  • Next Build on main (#41478103696)
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