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

stillwater-sc / universal / 43785117344
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 16 Nov 2026 01:34AM UTC
Jobs 1
Files 894
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

16 Nov 2026 01:14AM UTC coverage: 89.131% (-0.03%) from 89.161%
43785117344

push

github

web-flow
fix(cfloat): numeric_limits, fdp with one fraction bit and guards in every build, polyroot, inline table, inf/NaN text; remove unused exceptions (#2092)

The limits, fdp and utility findings of sweep #2051:
- numeric_limits:
  - max_exponent was one short for layouts with max-exponent values
    (cfloat<8,4,uint8_t,true,true,false> reported 8; its max() scale is 8).
    It is maxpos's scale + 1, and max_exponent10 follows.
  - is_bounded was false. It is true.
  - is_iec559 is true for the IEEE layouts (subnormals, no max-exponent
    values, not saturating, and a quiet bit: fbits >= 2), as decided.
    round_style is unchanged.
  - epsilon was 0 where a layout without subnormals cannot hold 2^-fbits
    (cfloat<12,4> without subnormals); it reports min() there.
  The float and double layouts now report exactly what
  std::numeric_limits<float> and <double> do.
- quire_resolve ORed the sticky bit into bit 0, which with one fraction bit
  is the guard bit: cfloat<4,2> fdp of 0.5*0.5 + 3 was inf, and cfloat<5,3>
  got 4060 of 21952 three-term sums wrong. The quire MSB goes to the MUL
  blocktriple's overflow position, which leaves bit 0 for the sticky bit at
  every width.
- fdp_qc, fdp_stride and fdp relied on asserts: under NDEBUG a zero stride
  re-read x[0] and an index past a vector read out of bounds. They give a
  quiet NaN (fdp_qc a NaN quire) in every build, as posit and bposit do.
- polyroot did not compile (a cfloat does not convert to double
  implicitly), and a constant or empty polynomial read out of bounds.
- GenerateCfloatExponentBounds() is inline; GenerateTable's 1 << nbits is
  a size_t shift, capped at 2^24 rows.
- to_decimal_fixpnt_string prints inf/-inf/nan/snan instead of numbers;
  color_print puts its exponent nibble markers where to_binary does.
- parse builds its regex once: 3.75 -> 0.29 us for decimal text,
  4.0 -> 0.61 us for the native form.
- The exceptions that were never thrown are removed, as decided:
  cfloat_quire_ex... (continued)

31 of 31 new or added lines in 5 files covered. (100.0%)

48 existing lines in 2 files now uncovered.

63569 of 71321 relevant lines covered (89.13%)

12671130.66 hits per line

Coverage Regressions

Lines Coverage ∆ File
46
75.38
-15.96% include/sw/universal/number/cfloat/manipulators.hpp
2
93.75
-6.25% include/sw/universal/number/cfloat/mathext/horners.hpp
Jobs
ID Job ID Ran Files Coverage
1 43785117344.1 16 Nov 2026 01:34AM UTC 894
89.13
GitHub Action Run
Source Files on build 43785117344
  • Tree
  • List 894
  • Changed 9
  • Source Changed 8
  • Coverage Changed 9
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #43785117344
  • 0909ae8f on github
  • Prev Build on main (#43775712941)
  • Next Build on main (#43798278173)
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