Use SIMD vectorization for physics thorns

Merged
#7 · Created  · Last updated

Merged pull request

CarpetX: Reverse dimensions in grid structure output

ab26535·Author: ·Closed by: ·2021-07-13

Description

  • Experiment with SIMD vectorization

  • Experiment with SIMD vectorization (add missing file)

  • Arith: Avoid constexpr problems in div and dual

  • Arith: Improve constructors

  • Arith: Add mixed rational-integer operators

  • Arith: Support CUDA for simd types

  • CarpetX: Removed unused lambda

  • Loop: New type GF3D2index

  • CarpetX: Make GF3D5layout its own type

  • CarpetX: loop: Support SIMD masked load/store operations

  • CarpetX: Replace omp critical by C++ mutex

    LLVM 11 crashes on omp critical statements in lambdas.

  • CarpetX: Replace constructor by assignment

  • CarpetX: Omit "(void)" in empty argument list declarations

  • CarpetX: Circumvent constexpr problems in prolongation operators

  • CarpetX: Replace constructors by assignments

  • DGWaveToy: Reformat code

  • Hydro: Reformat code

  • HydroToyCarpetX: Move attributes after function arguments

  • HydroToyGPU: Move attributes after function arguments

  • WaveToyGPU: Move attributes after function arguments

  • Weyl: Move attributes after function arguments

  • Z4c: Move attributes after function arguments

  • ODESovlers: Add helper function make_array

  • ODESolvers: Allow computed why arguments to check_valid

  • ODESolvers: Implement a more efficient linear combination algorithm

    Avoid temporaries, parallelize via OpenMP and CUDA.

  • TwoPunctures: Reformat code

  • WaveToyMoL: Reformat code

  • Weyl: Only check assert conditions in debug mode

  • Weyl: Use symmetry for calculating Ricci tensor

  • Weyl: Experiment with optimizing calculating Riemann tensor

  • Weyl: Optimize array indexing

  • Z4c: Optimize array indexing

  • Arith: Update code to make it work with GCC, LLVM, and nvcc

  • Arith: Modernize tensor operations (vectors, tensors, Riemann tensors, etc.)

  • CarpetX: Modernize CUDA annotations

  • Hydro: Modernize SIMD annotations

  • HydroToyCarpetX: Modernize CUDA annotations

  • HydroToyGPU: Modernize CUDA annotations

  • WaveToyGPU: Modernize CUDA annotations

  • Weyl: Use SIMD

  • Z4c: Modernize tensor operations

  • Weyl: Remove unused file

  • Z4c: Remove unused file

  • Arith: Add alignment functions

  • Arith: Correct and simplify simd class

  • Weyl: Correct derivatives and tetrad

  • Z4c: Correct derivatives

  • Arith: Introduce anyisnan

  • Weyl: Improve handling coordinate singularities

  • Weyl: Clean up unused code

 

0 attachments

0 comments

Loading commits...