Mixed PrecisionΒΆ

  • ginkgo types are templated

  • allows using objects with different precision in the same code

  • can convert vectors and matrices between different precisions

  • lin op apply can handle different precisions

    • matrix and any vector can have different precision

    • vectors are converted back and forth to matrix precision

    • some matrices can skip this step

      • ell and csr