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

stillwater-sc / universal / 38078833388
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: main
DEFAULT BRANCH: master
Ran 10 Oct 2026 07:56PM UTC
Jobs 1
Files 696
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

10 Oct 2026 07:11PM UTC coverage: 86.401% (+0.06%) from 86.339%
38078833388

push

github

web-flow
feat(ucalc): the tightest box and the overestimation factor in ubox (#1662)

Phase 3 of #1654: the tightest box a tile type could state for a
computation, and the overestimation factor of the box it computes.

- include/sw/universal/utility/tile_oracle.hpp:
  - exact einteger dyadics, and closed dyadic intervals whose / and
    sqrt round outward and detect exact results;
  - exact areal and poxel lattice values;
  - locate / box, a value's tile by bisection;
  - hull keeps open ends.
  It shares no code with the tile arithmetic, and agrees with it on every
  exact-operand + - * / and sqrt, exhaustively up to poxel<10,2>.
- tools/ucalc/oracle.hpp:
  - point inputs give one tile, located at rising precision;
  - set inputs give the tile hull of the image, with monotonicity proven
    per piece by interval forward AD (exact corners), and mean-value-form
    branch and bound for the rest;
  - open inputs exclude their ends;
  - the inner and outer bounds prove the box, or bracket it.
- ubox gains "tightest" and "over" columns, and JSON input_tiles and
  formula_tiles.
- For 3x^2 + 100x + 2 with ULP-wide inputs, both forms of r1 have the
  same proven tightest box (7/5/5/5/5/5), matching #1649. As written,
  r1 is 1093-3413x over; rearranged, 1-1.8x.
- quadratic_roots.cpp takes its dyadics from the header, and gains an
  assertion that the general oracle agrees with its polynomial test.

einteger (each in its own commit, with tests that fail on the old code):
- #1659: the quotient's sign was lost for single-limb divisors; the
  remainder had no sign; 1 / -2 gave -0.
- #1660: a - b with a < 0 and b > 0 was wrong (-5 - 1 = 4).
- #1661: multi-limb uint32 division was wrong: the Knuth D3 loop never
  broke, and a normalized divisor took an undefined shift.

Refs #1654
Fixes #1659
Fixes #1660
Fixes #1661

BEHAVIOUR-CHANGE: einteger division keeps the quotient's sign for single-limb divisors, and % of a negative dividend is now negative (truncated, as in C++)
B... (continued)

217 of 220 new or added lines in 2 files covered. (98.64%)

2 existing lines in 2 files now uncovered.

47498 of 54974 relevant lines covered (86.4%)

9952047.13 hits per line

Uncovered Changes

Lines Coverage ∆ File
3
98.49
include/sw/universal/utility/tile_oracle.hpp

Coverage Regressions

Lines Coverage ∆ File
1
96.3
-0.06% include/sw/universal/number/cfloat/cfloat_impl.hpp
1
97.52
1.28% include/sw/universal/number/einteger/einteger_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 38078833388.1 10 Oct 2026 07:56PM UTC 696
86.4
GitHub Action Run
Source Files on build 38078833388
  • Tree
  • List 696
  • Changed 2
  • Source Changed 1
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #38078833388
  • 2f7143aa on github
  • Prev Build on main (#38070258294)
  • Next Build on main (#38084891065)
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