Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
many, many small and moderate-size changes:
* introduced scalar_t, which lists scalar type options: int, int8, ...int64, ... half, float, double, bool, char. The type replaces NN.Ast.nntyp_t. It can also be used in OpenCV bindings, but not yet * extended many NN operations to support FP16 * accelerated fp16 <=> float conversions in default C branch (non-NEON, non-AVX2) * accelerated vector exponent (currently available only in NEON configuration) * extended GEMM to support various combinations of input/output data types, including complete FP16 support. For example, alpha * A (F16) * B (F32) + beta * C (F32) => result (F16) is supported.
- Loading branch information