import pyps
from pyps import module, Maker


class ompMaker(Maker):
    """
    A makefile builder for openmp
    """

    def get_ext(self):
        """"""
        return '.omp' + super(ompMaker, self).get_ext()

    def get_makefile_info(self):
        """"""
        return [('openmp', 'Makefile.omp')] + super(ompMaker, self).get_makefile_info()


def openmp(m, verbose=False, loop_distribution=True, loop_parallel_threshold_set=False, **props):
    """
    Parallelize module with opennmp
    """
    w = m.workspace
    # select most precise analysis
    w.activate(module.must_regions)
    w.activate(module.transformers_inter_full)
    w.activate(module.interprocedural_summary_precondition)
    w.activate(module.preconditions_inter_full)
    w.activate(module.region_chains)
    w.activate(module.rice_regions_dependence_graph)

    w.props.semantics_compute_transformers_in_context = True
    w.props.semantics_fix_point_operator = 'derivative'
    w.props.unspaghettify_test_restructuring = True
    w.props.unspaghettify_recursive_decomposition = True
    w.props.aliasing_across_io_streams = False
    w.props.constant_path_effects = False
    w.props.prettyprint_sequential_style = 'do'
    w.props.memory_effects_only = False
    w.props.parallelize_again_parallel_code = False

    m.loop_fusion()

    if loop_parallel_threshold_set:
        m.omp_loop_parallel_threshold_set(**props)
    m.split_initializations(**props)
    # privatize scalar variables
    m.privatize_module(**props)
    # openmp parallelization coarse grain
    if verbose:
        w.props.parallelization_statistics = True
    # custom functions
    m.coarse_grain_parallelization(**props)
    # do this **before** loop distribution
    m.flag_parallel_reduced_loops_with_openmp_directives(**props)
    if loop_distribution:
        m.internalize_parallel_code(**props)
        # some new reductions may have appeared
        m.flag_parallel_reduced_loops_with_openmp_directives(**props)
    m.ompify_code(**props)
    m.omp_merge_pragma(**props)
    if verbose:
        m.display(**props)


pyps.module.openmp = openmp
pyps.modules.openmp = lambda m, verbose=False, loop_distribution=True, loop_parallel_threshold_set=False, **props: [
    openmp(x, verbose, loop_distribution, loop_parallel_threshold_set, **props) for x in m]