stillwater-sc / universal / 23119164012

85%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

feat(unum1): implement native type conversions (#573)

* feat(unum1): implement native type conversions for unum Type I

Implement operator=(double) that encodes IEEE-754 doubles into unum
Type I format by decomposing into sign, exponent, and fraction, then
finding the tightest esize/fsize representation. The ubit is set when
the value cannot be represented exactly.

Key design decisions:
- IEEE-754-style subnormal convention: biased_exp=0 means hidden bit
  is 0 (subnormal/zero), biased_exp>0 means hidden bit is 1 (normal)
- Encoder finds smallest esize where biased_exp >= 1 (avoids
  subnormal zone for normal values)
- Encoder finds smallest fsize that exactly captures the fraction;
  sets ubit=1 if no exact match within max fsize
- to_double() updated to handle subnormal decoding

Also implements:
- Integer assignment via double conversion
- Unary negation (sign bit flip)
- Conversion test suite covering exact round-trips (powers of 2,
  binary fractions), inexact ubit flagging (1/3, pi), negative
  values, integer assignment, and unary negation

Resolves #565

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

* fix(unum1): update education example to use unsigned template params

The unum template changed from size_t to unsigned parameters in Phase 1.
Update basic_operators.cpp to match.

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

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Coverage Stats

67 of 74 new or added lines in 1 file covered. (90.54%)

4 existing lines in 3 files now uncovered.

43216 of 51411 relevant lines covered (84.06%)

6130613.92 hits per line