stillwater-sc / universal / 22902740273

84%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

feat(quire): add generalized limb-based quire with blocktriple support (#551)

* feat(quire): add generalized limb-based quire with blocktriple support

Implement a generalized quire<NumberType, capacity, LimbType> class that
uses blockbinary (limb-based, uint32_t/uint64_t) for fast carry propagation,
replacing the legacy bitblock/std::bitset approach.

Key design:
- Single blockbinary<qbits, LimbType, Unsigned> for magnitude storage
- Uses quire_traits<NumberType> for sizing (from #545)
- operator+=(blocktriple) as the primary accumulation path
- MSB-based bit placement for robust blocktriple-to-accumulator mapping
- Works with cfloat, posit, fixpnt, lns, and dbns types
- Shared quire exception hierarchy (no type-specific coupling)

The existing specialized posit quire in number/posit/quire.hpp is retained
as-is for backward compatibility with posit-specific FDP workflows.

Also adds forward declaration includes to quire_traits.hpp so the traits
header can be used independently of the full number type headers.

Relates to #345, resolves #546

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix(quire): address CodeRabbit review feedback on PR #551

- Replace raw string throw in operator[] with std::out_of_range
- Fix operator< to use MSB-first unsigned magnitude comparison instead
  of delegating to blockbinary::operator< (which uses sign-aware logic
  incorrect for BinaryNumberType::Unsigned accumulators)
- Add clarifying comments on unsigned subtraction correctness via
  two's complement (valid when minuend >= subtrahend)
- Add precision limitation note to convert_to<TargetType>()

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix(quire): address second round of CodeRabbit review on PR #551

- Use quire_traits<NumberType>::capacity as default template parameter
  instead of hard-coded 30, so type-specific defaults are respected
- Fix undefined behavior when assigning INT64_MIN: compute magnitude
  in the unsigned domain to av... (continued)

Run Details

130 of 154 new or added lines in 1 file covered. (84.42%)

7 existing lines in 2 files now uncovered.

42406 of 50324 relevant lines covered (84.27%)

6242085.37 hits per line