mp-numerics

Mixed-precision numerical methods. mp-numerics composes two header-only libraries — MTL5 for linear algebra and Universal for parameterized number systems — to study classic numerical methods (quadrature, root finding, interpolation, optimization, approximation) under custom arithmetic (half precision, posits, ...).

MTL5 deliberately has no dependency on Universal: it is the general linear-algebra layer. mp-numerics is the integration layer where numerical methods meet Universal's number types.

Build

# Dependencies (MTL5 + Universal) are pulled automatically via FetchContent.
cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j

# Run the smoke test
ctest --test-dir build --output-on-failure

# Run the quadrature precision study
./build/applications/integration/simpson_precision/simpson_precision

# Run the benchmark (optional arg: subinterval count)
./build/benchmarks/integration/simpson_benchmark 1048576

Using local checkouts instead of fetching from GitHub:

cmake -B build \
  -DFETCHCONTENT_SOURCE_DIR_MTL5=../mtl5 \
  -DFETCHCONTENT_SOURCE_DIR_UNIVERSAL=../universal

Layout

The repo is organized by numerical-method category: integration, roots, interpolation, optimization, approximation.

include/sw/mp_numerics/
  integration.hpp                # composite Simpson / trapezoidal, generic over Real
include/mtl/math/
  quire_accumulator.hpp          # MTL5 accumulator_traits <- Universal quire bridge
applications/
  integration/simpson_precision/ # quadrature error across number precisions
benchmarks/
  integration/                   # accuracy-per-unit-time across number systems
tests/
  integration/                   # composite Simpson smoke test across types
  roots/ interpolation/ optimization/ approximation/   # (no tests yet -- roadmap)
docs/roadmap.md                  # milestones and migration plan

License

MIT — see LICENSE.