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

stillwater-sc / universal / 43625651943
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 14 Nov 2026 06:35PM UTC
Jobs 1
Files 892
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

14 Nov 2026 06:16PM UTC coverage: 89.119% (+0.02%) from 89.098%
43625651943

push

github

web-flow
fix(posit): posit-to-posit and wide-integer conversion on the encoding, exact integer casts, numeric_limits, reciprocal/rsqrt (#2073)

* fix(posit): posit-to-posit and wide-integer conversion on the encoding, exact integer casts, numeric_limits, reciprocal/rsqrt

The QA sweep in #2047 found the posit conversions going through double and several
helpers that were wrong or did not compile.

- posit<n,e,b>(posit<nn,ee,bb>) decodes the source on its encoding (posit_limbs) and
  rounds once into the target. Through double it lost every bit past 53, rounded
  twice (posit<64,2>(1 + 2^-13 + 2^-58) -> posit<16,1> was 1), and gave NaR or 0 for
  values outside double's range, where a posit saturates.
- Integers into posits wider than 64 bits enter exactly and round once:
  posit<128,2>(2^60 + 1) - 2^60 was 0.
- operator int/long/long long truncate toward zero exactly at any nbits, saturate to
  the integer's range and give its minimum for NaR. They went through double (or long
  double): inexact past 53 bits, undefined behaviour out of range or for NaR.
- reciprocal() is 1/x on the division path; the old body used blockbinary helpers that
  no longer exist, so neither it nor rsqrt compiled.
- numeric_limits<posit>: max_exponent is maxpos's scale + 1 (posit<16,1>: 29, was 28),
  as min_exponent already was; has_infinity false, has_signaling_NaN false, is_bounded
  true, as for bposit.
- attributes.hpp: useed_value / useed<nbits,es> build 2^(2^es) without a shift (UB for
  es >= 6, useed_value<32,6> was 0); useed<es>() rejects es > 5, which a uint64_t cannot
  hold; fraction_value uses the clamped fbits (posit<5,3> did not compile).
- ulp(maxpos) is maxpos minus its predecessor, not NaR.
- positExponent == and < named members that do not exist; operator>> no longer writes
  to std::cerr on a failed parse (failbit only); parse() builds its regex once.
- debug.hpp: showLimbs printed one bit per block and constexprClassParameters passed
  to_binary's arguments swap... (continued)

62 of 62 new or added lines in 7 files covered. (100.0%)

3 existing lines in 2 files now uncovered.

63181 of 70895 relevant lines covered (89.12%)

11009491.11 hits per line

Coverage Regressions

Lines Coverage ∆ File
2
82.05
0.17% include/sw/universal/verification/posit_test_suite.hpp
1
35.02
0.32% include/sw/universal/verification/test_suite_randoms.hpp
Jobs
ID Job ID Ran Files Coverage
1 43625651943.1 14 Nov 2026 06:35PM UTC 892
89.12
GitHub Action Run
Source Files on build 43625651943
  • Tree
  • List 892
  • Changed 11
  • Source Changed 6
  • Coverage Changed 10
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #43625651943
  • fb72eae3 on github
  • Prev Build on main (#43601176504)
  • Next Build on main (#43635043952)
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