Method configure

src/cpu/kernels/CpuSubKernel.cpp:129–158

} // namespace

void CpuSubKernel::configure(const ITensorInfo *src0, const ITensorInfo *src1, ITensorInfo *dst, ConvertPolicy policy)
{
    ARM_COMPUTE_TRACE_EVENT(ARM_COMPUTE_PROF_CAT_CPU, ARM_COMPUTE_PROF_LVL_CPU, "CpuSubKernel::configure");
    ARM_COMPUTE_ERROR_ON_NULLPTR(src0, src1, dst);
    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(*src0, *src1, *dst, policy));

    const TensorShape &out_shape = TensorShape::broadcast_shape(src0->tensor_shape(), src1->tensor_shape());

    // Auto initialize dst if not initialized
    set_shape_if_empty(*dst, out_shape);
    set_data_type_if_unknown(*dst, src0->data_type());

    const auto can_use_fixedpoint    = sub_q8_neon_fixedpoint_possible(src0, src1, dst);
    const auto can_use_sme2_add_impl = false;
    const auto uk =
        CpuSubKernel::get_implementation<CpuSubKernelDataTypeISASelectorData>(CpuSubKernelDataTypeISASelectorData{
            src0->data_type(), CPUInfo::get().get_isa(), can_use_fixedpoint, can_use_sme2_add_impl});

    ARM_COMPUTE_ERROR_ON_NULLPTR(uk);

    _policy     = policy;
    _run_method = uk->ukernel;
    _name       = std::string("CpuSubKernel").append("/").append(uk->name);

    // CpuSubKernel doesn't need padding so update_window_and_padding() can be skipped
    Window win;
    std::tie(win, _split_dimension) = calculate_squashed_or_max_window(*src0, *src1);

    ICpuKernel::configure(win);
}

size_t CpuSubKernel::get_mws(const CPUInfo &platform, size_t thread_count) const
{