Function transpose_conj

src/backend/cpu/kernel/transpose.hpp:105–133

template<typename T>
void transpose_conj(Param<T> output, CParam<T> input) {
    const af::dim4 odims    = output.dims();
    const af::dim4 ostrides = output.strides();
    const af::dim4 istrides = input.strides();

    T *out            = output.get();
    T const *const in = input.get();

    for (dim_t l = 0; l < odims[3]; ++l) {
        for (dim_t k = 0; k < odims[2]; ++k) {
            // Outermost loop handles batch mode
            // if input has no data along third dimension
            // this loop runs only once

            for (dim_t j = 0; j < odims[1]; ++j) {
                for (dim_t i = 0; i < odims[0]; ++i) {
                    // calculate array indices based on offsets and strides
                    // the helper getIdx takes care of indices
                    const dim_t inIdx  = getIdx(istrides, j, i, k, l);
                    const dim_t outIdx = getIdx(ostrides, i, j, k, l);
                    out[outIdx]        = getConjugate(in[inIdx]);
                }
            }
            // outData and inData pointers doesn't need to be
            // offset as the getIdx function is taking care
            // of the batch parameter
        }
    }
}

template<typename T>
void transpose(Param<T> out, CParam<T> in, const bool conjugate) {