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

stillwater-sc / universal / 40525921004
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/issue-2111-fixpnt
DEFAULT BRANCH: master
Ran 27 Oct 2026 03:16PM UTC
Jobs 1
Files 764
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

27 Oct 2026 02:55PM UTC coverage: 88.219% (-0.05%) from 88.266%
40525921004

push

github

web-flow
feat(efloat): the math library and the float conversions in constant evaluation (#1847)

* feat(efloat): the math library and the float conversions in constant evaluation

Phase 3 of #1819: every function of the efloat math library runs in a constant
expression, with the run-time result bit for bit, and efloat converts from and
to float and double there.

- Constants cached per thread (ln 2, ln 10, pi, atan(1/2), 2/sqrt(pi),
  log(2 pi)/2, the log table, the Stirling coefficients) are computed in a
  constant evaluation by constant_value(); the series take their scratch from
  scratch_space, a vector of their own there.
- The double estimates (series terms, halvings, cancellations, lgamma's parts)
  use efloat/double_math.hpp instead of <cmath>: floor, ceil, trunc, round,
  nearbyint, fabs, ilogb and the power-of-2 scale by the fields, bit for bit
  <cmath>'s over 20M arguments; log2/log/log10 from constexpr_math; sin(pi f)
  and lgamma by series. sqrt is the instruction at run time and the integer
  root in a constant evaluation, the same value.
- float and double conversions read and write the encoding (extractFields,
  bit_cast) instead of std::fpclassify/signbit; one rounding routine,
  round_to_native<K, E_min, E_max>(), serves float, double, x87 and binary128.
  long double is taken apart and assembled by its encoding, double-double as
  hi + lo: no std::frexp or std::ldexp.
- Flags go through raise_flags(), the mode through rounding_mode() (cbrt, rint);
  log(0)'s -inf now carries x's precision in a constant evaluation too.

Run-time results are unchanged: 125656 results of 41 functions at 7 precisions
in the four modes, with their flags, and 1.6M float/double conversions, match
main under g++ and clang; speed within noise, lgamma 3-5% faster.

New test elastic/efloat/math/constexpr_mathlib.cpp; lgamma/tgamma need a raised
step limit on clang and MSVC, set for that test in CMakeLists.txt.

Relates to #1819

BEHAVIOUR-CHANGE: long double(efloat)... (continued)

510 of 587 new or added lines in 13 files covered. (86.88%)

2 existing lines in 1 file now uncovered.

55105 of 62464 relevant lines covered (88.22%)

8991754.54 hits per line

Uncovered Changes

Lines Coverage ∆ File
44
58.88
include/sw/universal/number/efloat/double_math.hpp
16
95.37
0.08% include/sw/universal/number/efloat/efloat_impl.hpp
10
96.48
-1.64% include/sw/universal/number/efloat/math/exponent.hpp
5
89.34
5.26% include/sw/universal/number/efloat/math/logarithm.hpp
2
98.97
-0.51% include/sw/universal/number/efloat/math/error_and_gamma.hpp

Coverage Regressions

Lines Coverage ∆ File
2
96.91
-0.13% include/sw/universal/number/cfloat/cfloat_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 40525921004.1 27 Oct 2026 03:16PM UTC 764
88.22
GitHub Action Run
Source Files on build 40525921004
  • Tree
  • List 764
  • Changed 15
  • Source Changed 13
  • Coverage Changed 13
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #40525921004
  • 2b1aa9b5 on github
  • Prev Build on main (#40442115766)
  • Next Build on main (#40596158543)
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