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

stillwater-sc / universal / 39053590561
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/issue-2132-rational-exact-saturating
DEFAULT BRANCH: master
Ran 17 Oct 2026 05:39PM UTC
Jobs 1
Files 730
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

17 Oct 2026 05:24PM UTC coverage: 86.758% (+0.07%) from 86.692%
39053590561

push

github

web-flow
feat(lindex): lindex<nbits, lbits, bt> core type and conversion (#1349 part 3) (#1708)

* feat(lindex): lindex<nbits, lbits, bt> core type and conversion (#1349 part 3)

The level-index type without arithmetic, layered as #1334 prescribes:
core.hpp (0 I/O-family headers), manipulators.hpp, iostream.hpp,
attributes.hpp and the umbrella lindex.hpp, plus traits/lindex_traits.hpp.

Encoding per docs/level-index-design.md: the magnitude is the fixed-point
number l + f (lbits integer bits, fbits = nbits - 1 - lbits fraction
bits), negatives are the two's complement of the whole pattern, zero is
all zeros, NaR is 1000...0, and there is no infinity. Comparison is a
signed-integer compare of the encoding; NaR equals itself and sorts
below every value (the posit convention).

Conversion rounds psi(|v|) to nearest even on the 2^-fbits grid and
saturates at maxpos; below minpos/2 it rounds to 0. Conversion to a
native float rounds once and gives +-inf past its range, as the design
doc's special-values table specifies (no IEEE type reaches level 5).
Integers saturate and truncate toward zero. Both directions evaluate
phi and psi in dd_cascade at every width, so the one rounding is
correct at fbits 60 as well; level 4 to native goes through
2^k e^r so exp never sees a large argument.

Tests check conversion for equality against
tools/generators/lindex_conversion_gen.py (mpmath, 600 bits) in six
configurations, never by round trip: doubles, floats, 64-bit signed and
unsigned integers to encodings, and encodings to correctly rounded
doubles and floats. Comparison is checked exhaustively at 8 bits and
sampled at 16/32/64 against the decoded signed psi and the double order.
A core-only test runs with no stream in the translation unit.

dd_cascade/math/functions/logarithm.hpp reported its domain errors
through std::cerr, which put <iostream> in any core that needs dd_cascade
log; it now uses fprintf(stderr, ...), as the #1334 cores do.

Resolves #1694

Co-Authored-By: C... (continued)

286 of 287 new or added lines in 5 files covered. (99.65%)

8 existing lines in 1 file now uncovered.

49765 of 57361 relevant lines covered (86.76%)

10430644.29 hits per line

Uncovered Changes

Lines Coverage ∆ File
1
99.49
include/sw/universal/number/lindex/lindex_impl.hpp

Coverage Regressions

Lines Coverage ∆ File
8
96.37
-0.52% include/sw/universal/number/cfloat/cfloat_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 39053590561.1 17 Oct 2026 05:39PM UTC 730
86.76
GitHub Action Run
Source Files on build 39053590561
  • Tree
  • List 730
  • Changed 4
  • Source Changed 1
  • Coverage Changed 4
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #39053590561
  • 8e420b3c on github
  • Prev Build on main (#39047295777)
  • Next Build on main (#39055757376)
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