stillwater-sc / universal / 23342913519
85%
master: 84%

LAST BUILD BRANCH: main
DEFAULT BRANCH: master
Ran
Jobs 1
Files 627
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
coverage: 84.438% (+0.01%) from 84.426%
23342913519

push

github

web-flow 
feat(ucalc): replace double interchange with native-precision arithmetic (#601)

* feat(ucalc): replace double interchange with native-precision arithmetic (#598)

Add std::any to Value to store the native type alongside the double
approximation. The extract<T>(value) helper retrieves the native type
when available, falling back to double conversion for cross-type
operations.

All arithmetic lambdas in register_type<T> now use extract<T>()
instead of T(a.num), preserving full precision within the active type:

  qd:  phi*phi - phi - 1 = 1.99e-64 (was 0 via double cancellation)
  dd:  phi*phi - phi - 1 = -6.16e-33 (was 0 via double cancellation)
  fp32: phi*phi - phi - 1 = 1.19e-7 (correct, one ULP)

Cross-type fallback: when a Value's std::any doesn't hold the
expected type (e.g., variable assigned in qd used in posit32),
extract<T> falls back to T(v.num) as before.

Closes #598.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(docs): correct false posit closure claim in ucalc tutorial

Example 1 falsely claimed that 1/3 + 1/3 + 1/3 = 1 is a posit-specific
property due to tapered precision. In fact, IEEE float also produces 1.0
for this expression -- the rounding errors cancel in both systems.

Replace with a genuine precision comparison: (1 + 1e-8) - 1 returns
7.45e-9 in posit32 (preserving the perturbation) vs 0 in fp32 (lost
because 1e-8 is below fp32's ULP at 1.0). This correctly demonstrates
posit's smaller epsilon near 1.0 without making false claims.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* docs(ucalc): refresh all tutorial examples with current REPL output

Regenerate all 8 examples using the current ucalc with native-precision
rendering and native arithmetic. Key changes:

- All values now show at native precision (not hardcoded 17 digits)
- Example 3 (golden ratio) shows dd residual at -6.16e-33 (native dd
  arithmetic) instead of 0 (double cancellation)
- Example 2 mentions grouped... (continued)

44324 of 52493 relevant lines covered (84.44%)

6094661.92 hits per line

Coverage Regressions

Lines Coverage ∆ File
1
81.47
 0.0% include/sw/universal/number/posit1/specialized/posit_16_1.hpp
Jobs
ID Job ID Ran Files Coverage
1 23342913519.1 627
84.44
 GitHub Action Run
Source Files on build 23342913519