stillwater-sc / universal / 23324727851

85%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

feat(tools): add ucalc mixed-precision REPL calculator (#596)

* feat(ucalc): add mixed-precision REPL calculator (#595)

Interactive calculator that replaces the workflow of writing, compiling,
and running C++ to compare arithmetic across Universal number types.

Supports 29 types (posit, cfloat, fixpnt, lns, integer, bfloat16,
dd, qd, hfloat, dfloat, cascades), infix expressions with variables
and math functions, and analysis commands (compare, sweep, faithful,
range, precision, ulp). Optional GNU Readline for line editing and
tab completion.

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

* docs(ucalc): add implementation plan for REPL calculator

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

* fix(number): replace typeid(bt).name() with type_tag(bt{}) in manipulators

The mangled typeid output ('j' for uint32_t, 'h' for uint8_t on GCC)
is not human-readable. Use the existing type_tag() overloads from
integer_type_tag.hpp which return proper names like 'uint32_t'.

Fixed in: fixpnt, dfloat, hfloat, dfixpnt, integer manipulators.

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

* feat(ucalc): add td_cascade and label cascades as Priest variants

Rename cascade type_tags to "double-double Priest", "triple-double
Priest", and "quad-double Priest" to distinguish them from the
Dekker-based dd/qd types.

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

* fix(posit): use standard es=2 config and drop namespace prefix from type_tag

The Posit Standard defines es=2 for all standard sizes (posit8 through
posit64). Updated ucalc registry to match. Removed the sw::universal::
prefix from posit type_tag to be consistent with all other number types.

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

* feat(ucalc): add fp8 DL formats and fp128 quad precision

Add fp8e2m5, fp8e3m4, fp8e4m3, fp8e5m2 Deep Learning 8-bit formats
and fp128 IEEE-754 quad precis... (continued)

Coverage Stats

50 of 54 new or added lines in 9 files covered. (92.59%)

1 existing line in 1 file now uncovered.

44315 of 52489 relevant lines covered (84.43%)

6071409.34 hits per line