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

stillwater-sc / universal / 38897796048
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: main
DEFAULT BRANCH: master
Ran 16 Oct 2026 12:44PM UTC
Jobs 1
Files 723
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 Oct 2026 12:02PM UTC coverage: 86.692% (+0.01%) from 86.68%
38897796048

push

github

web-flow
feat(quire): accumulator_* protocol so linear-algebra libraries accumulate in a quire (#1699)

* feat(quire): accumulator_* protocol so linear-algebra libraries accumulate in a quire

Adds quire/accumulator.hpp, included from quire/core.hpp: five free
functions on quire<T, capacity, LimbType> -- accumulator_clear,
accumulator_assign, accumulator_add_product, accumulator_value(q,
std::type_identity<R>{}) and accumulator_merge -- defined generically
over every number type whose fdp.hpp provides quire_mul and quire_resolve
(posit, bposit, cfloat, fixpnt, lns, dbns).

MTL5's accumulator_traits finds these by argument-dependent lookup
(stillwater-sc/mtl5#581), so a quire is an accumulator for any MTL5
operation that takes one, with neither library including the other. This
replaces the quire_accumulator.hpp bridge copied into mp-blas, mp-ir,
mp-iterative, mp-numerics and mp-spice. accumulator_merge combines the
partial quires of a split reduction exactly.

The functions require the element type to be exactly T, and exclude the
native float/double quires, whose quire_mul takes built-in types that ADL
cannot reach.

Relates to stillwater-sc/mtl5#582

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(quire): round accumulator_value to float/double correctly; narrow the exact-product claim

Addresses the CodeRabbit review of #1699:

* accumulator_value to float or double went through convert_to<>(),
  which keeps the leading 53 bits and drops the rest: 1 + 3*2^-54 came
  back as 1 instead of 1 + 2^-52, and a float result rounded twice. It
  now rounds from the full quire (quire_round_to): to nearest, ties to
  even, with a sticky bit over everything below the guard, and gradual
  underflow and overflow handled by the target's limits. Other result
  types go through the correctly rounded double and say so.

* The header and docs claimed an exact product for every type. That holds
  for posit, bposit, cfloat and fixpnt. The lns and dbns quire_mul fo... (continued)

40 of 40 new or added lines in 1 file covered. (100.0%)

49136 of 56679 relevant lines covered (86.69%)

10159153.31 hits per line

Jobs
ID Job ID Ran Files Coverage
1 38897796048.1 16 Oct 2026 12:44PM UTC 723
86.69
GitHub Action Run
Source Files on build 38897796048
  • Tree
  • List 723
  • Changed 2
  • Source Changed 0
  • Coverage Changed 2
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #38897796048
  • 15198928 on github
  • Prev Build on main (#38845614918)
  • Next Build on main (#38930407030)
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