Function copy

src/backend/cuda/kernel/memcopy.hpp:172–207

template<typename inType, typename outType>
void copy(Param<outType> dst, CParam<inType> src, dim_t ondims,
          outType default_value, double factor) {
    const size_t totalSize{dst.elements() * sizeof(outType) +
                           src.elements() * sizeof(inType)};
    bool same_dims{true};
    for (dim_t i{0}; i < ondims; ++i) {
        if (src.dims[i] > dst.dims[i]) {
            src.dims[i] = dst.dims[i];
        } else if (src.dims[i] != dst.dims[i]) {
            same_dims = false;
        }
    }
    removeEmptyColumns(dst.dims, ondims, src.dims, src.strides);
    ondims = removeEmptyColumns(dst.dims, ondims, dst.dims, dst.strides);
    ondims =
        combineColumns(dst.dims, dst.strides, ondims, src.dims, src.strides);

    threadsMgt<dim_t> th(dst.dims, ondims);
    const dim3 threads{th.genThreads()};
    const dim3 blocks{th.genBlocks(threads, 1, 1, totalSize, sizeof(outType))};

    EnqueueArgs qArgs(blocks, threads, getActiveStream());

    auto copy{common::getKernel(
        th.loop0                 ? "arrayfire::cuda::scaledCopyLoop0"
        : (th.loop2 || th.loop3) ? "arrayfire::cuda::scaledCopyLoop123"
        : th.loop1               ? "arrayfire::cuda::scaledCopyLoop1"
                                 : "arrayfire::cuda::scaledCopy",
        {{copy_cuh_src}},
        TemplateArgs(TemplateTypename<inType>(), TemplateTypename<outType>(),
                     TemplateArg(same_dims), TemplateArg(factor != 1.0)))};

    copy(qArgs, dst, src, default_value, factor);

    POST_LAUNCH_CHECK();
}
}  // namespace kernel
}  // namespace cuda
}  // namespace arrayfire